omnius 1.0.14 → 1.0.15

package/dist/index.js

@@ -1451,7 +1451,7 @@ var init_security_classifier = __esm({
       { match: /^(aiwg_setup|aiwg_health|aiwg_workflow)$/, info: CRITICAL_SENSITIVE },
       { match: /^(expose|sponsor|nexus_register|wallet_|x402|payment|spend)/, info: CRITICAL_SENSITIVE },
       // ── Hardware peripherals
-      { match: /^(camera_capture|audio_capture|audio_playback|audio_analyze|asr_listen)$/, info: HARDWARE_DEVICE },
+      { match: /^(camera_capture|audio_capture|audio_playback|play_sound|audio_analyze|asr_listen)$/, info: HARDWARE_DEVICE },
       { match: /^(wifi_control|bluetooth_scan|sdr_scan|flipper_zero|meshtastic|gps_location)$/, info: HARDWARE_DEVICE },
       { match: /^(desktop_click|desktop_describe|screenshot)$/, info: HARDWARE_DEVICE },
       { match: /^(jibberlink)$/, info: HARDWARE_DEVICE },
@@ -84452,7 +84452,7 @@ var require_mime_types = __commonJS({
   "../node_modules/mime-types/index.js"(exports) {
     "use strict";
     var db = require_mime_db();
-    var extname14 = __require("path").extname;
+    var extname15 = __require("path").extname;
     var EXTRACT_TYPE_REGEXP = /^\s*([^;\s]*)(?:;|\s|$)/;
     var TEXT_TYPE_REGEXP = /^text\//i;
     exports.charset = charset;
@@ -84506,7 +84506,7 @@ var require_mime_types = __commonJS({
       if (!path11 || typeof path11 !== "string") {
         return false;
       }
-      var extension4 = extname14("x." + path11).toLowerCase().substr(1);
+      var extension4 = extname15("x." + path11).toLowerCase().substr(1);
       if (!extension4) {
         return false;
       }
@@ -250538,6 +250538,19 @@ function trimProcessText(text, max = 1800) {
     return clean3;
   return clean3.slice(0, max - 20) + "\n... (truncated)";
 }
+function formatDiffusersFailure(stderrOrStdout) {
+  const text = trimProcessText(stderrOrStdout);
+  const lower = stderrOrStdout.toLowerCase();
+  const notes2 = [];
+  if (lower.includes("torchvision") && (lower.includes("not installed") || lower.includes("no module named"))) {
+    notes2.push("Missing torchvision was detected. The image-generation dependency set now includes torchvision; run /image prewarm again to repair the existing .omnius/image-gen/.venv.");
+  }
+  if (lower.includes("hf_token") || lower.includes("gated repo") || lower.includes("401") || lower.includes("unauthorized")) {
+    notes2.push("This model may require Hugging Face authentication or license acceptance. Set HF_TOKEN in the environment and accept the model license on Hugging Face, then prewarm again.");
+  }
+  return [text, ...notes2.map((note) => `
+Note: ${note}`)].filter(Boolean).join("");
+}
 function imageGenerationPythonEnv(repoRoot) {
   const root = imageGenerationDir(repoRoot);
   const hf = join36(root, "huggingface");
@@ -250595,7 +250608,7 @@ async function ensurePythonFor(repoRoot, kind, explicit, onProgress) {
 ${trimProcessText(created.stderr || created.stdout)}`);
     }
   }
-  const importCheck = kind === "diffusers" ? "import torch, diffusers, PIL\nfrom diffusers import AutoPipelineForText2Image\n" : "import stable_diffusion_cpp, PIL\n";
+  const importCheck = kind === "diffusers" ? "import torch, torchvision, diffusers, PIL\nfrom diffusers import AutoPipelineForText2Image\n" : "import stable_diffusion_cpp, PIL\n";
   if (await pythonCanImport(command, importCheck, repoRoot, pythonEnv)) {
     return { command, env: pythonEnv };
   }
@@ -250661,6 +250674,7 @@ var init_image_generate = __esm({
     DEFAULT_OLLAMA_IMAGE_MODEL = "x/z-image-turbo";
     DIFFUSERS_PYTHON_PACKAGES = [
       "torch",
+      "torchvision",
       "diffusers",
       "transformers",
       "accelerate",
@@ -250710,7 +250724,7 @@ var init_image_generate = __esm({
         id: "black-forest-labs/FLUX.1-dev",
         label: "FLUX.1 dev",
         backend: "diffusers",
-        install: 'python .omnius/image-gen/diffusers_text2image.py --model black-forest-labs/FLUX.1-dev --steps 28 --guidance 3.5 --width 1024 --height 1024 --prompt "..." --output .omnius/images/out.png',
+        install: 'python3 .omnius/image-gen/diffusers_text2image.py --model black-forest-labs/FLUX.1-dev --steps 28 --guidance 3.5 --width 1024 --height 1024 --prompt "..." --output .omnius/images/out.png',
         category: "Primary hyper-realistic baseline",
         sizeClass: "12B rectified-flow transformer",
         quality: "Top-tier open-weight photorealism, prompt adherence, texture detail, composition, and typography.",
@@ -250727,7 +250741,7 @@ var init_image_generate = __esm({
         id: "stabilityai/stable-diffusion-3.5-large",
         label: "Stable Diffusion 3.5 Large",
         backend: "diffusers",
-        install: 'python .omnius/image-gen/diffusers_text2image.py --model stabilityai/stable-diffusion-3.5-large --steps 28 --guidance 4.5 --width 1024 --height 1024 --prompt "..." --output .omnius/images/out.png',
+        install: 'python3 .omnius/image-gen/diffusers_text2image.py --model stabilityai/stable-diffusion-3.5-large --steps 28 --guidance 4.5 --width 1024 --height 1024 --prompt "..." --output .omnius/images/out.png',
         category: "Primary hyper-realistic baseline",
         sizeClass: "8B MMDiT",
         quality: "Serious open Stable Diffusion ecosystem baseline with strong realism, complex prompt understanding, typography, and controllability.",
@@ -250744,7 +250758,7 @@ var init_image_generate = __esm({
         id: "black-forest-labs/FLUX.1-schnell",
         label: "FLUX.1 schnell",
         backend: "diffusers",
-        install: 'python .omnius/image-gen/diffusers_text2image.py --model black-forest-labs/FLUX.1-schnell --steps 4 --guidance 0 --width 1024 --height 1024 --prompt "..." --output .omnius/images/out.png',
+        install: 'python3 .omnius/image-gen/diffusers_text2image.py --model black-forest-labs/FLUX.1-schnell --steps 4 --guidance 0 --width 1024 --height 1024 --prompt "..." --output .omnius/images/out.png',
         category: "Fast large-model iteration",
         sizeClass: "12B rectified-flow transformer",
         quality: "FLUX-style output with fewer steps; better for rapid iteration than absolute peak quality.",
@@ -250761,7 +250775,7 @@ var init_image_generate = __esm({
         id: "stabilityai/stable-diffusion-3.5-large-turbo",
         label: "SD3.5 Large Turbo",
         backend: "diffusers",
-        install: 'python .omnius/image-gen/diffusers_text2image.py --model stabilityai/stable-diffusion-3.5-large-turbo --steps 4 --guidance 0 --width 1024 --height 1024 --prompt "..." --output .omnius/images/out.png',
+        install: 'python3 .omnius/image-gen/diffusers_text2image.py --model stabilityai/stable-diffusion-3.5-large-turbo --steps 4 --guidance 0 --width 1024 --height 1024 --prompt "..." --output .omnius/images/out.png',
         category: "Fast large-model iteration",
         sizeClass: "8B distilled MMDiT",
         quality: "SD3.5-family quality optimized for fewer inference steps; throughput over peak fidelity.",
@@ -250778,7 +250792,7 @@ var init_image_generate = __esm({
         id: "Tencent-Hunyuan/HunyuanDiT-v1.2-Diffusers",
         label: "HunyuanDiT v1.2",
         backend: "diffusers",
-        install: 'python .omnius/image-gen/diffusers_text2image.py --model Tencent-Hunyuan/HunyuanDiT-v1.2-Diffusers --steps 30 --guidance 7.5 --width 1024 --height 1024 --prompt "..." --output .omnius/images/out.png',
+        install: 'python3 .omnius/image-gen/diffusers_text2image.py --model Tencent-Hunyuan/HunyuanDiT-v1.2-Diffusers --steps 30 --guidance 7.5 --width 1024 --height 1024 --prompt "..." --output .omnius/images/out.png',
         category: "Large multilingual diffusion",
         sizeClass: "Large DiT text-to-image",
         quality: "Strong bilingual English/Chinese prompt understanding with detailed, realistic multi-resolution output.",
@@ -250795,7 +250809,7 @@ var init_image_generate = __esm({
         id: "Tongyi-MAI/Z-Image-Turbo",
         label: "Z-Image-Turbo",
         backend: "diffusers",
-        install: 'python .omnius/image-gen/diffusers_text2image.py --model Tongyi-MAI/Z-Image-Turbo --steps 8 --width 1024 --height 1024 --prompt "..." --output .omnius/images/out.png',
+        install: 'python3 .omnius/image-gen/diffusers_text2image.py --model Tongyi-MAI/Z-Image-Turbo --steps 8 --width 1024 --height 1024 --prompt "..." --output .omnius/images/out.png',
         category: "Modern deployable",
         sizeClass: "6B image generation model",
         quality: "Efficient newer large-model quality; useful below full FLUX/SD3.5 hardware budgets.",
@@ -250811,7 +250825,7 @@ var init_image_generate = __esm({
         id: "black-forest-labs/FLUX.2-klein-4B",
         label: "FLUX.2 Klein 4B",
         backend: "diffusers",
-        install: 'python .omnius/image-gen/diffusers_text2image.py --model black-forest-labs/FLUX.2-klein-4B --steps 8 --width 1024 --height 1024 --prompt "..." --output .omnius/images/out.png',
+        install: 'python3 .omnius/image-gen/diffusers_text2image.py --model black-forest-labs/FLUX.2-klein-4B --steps 8 --width 1024 --height 1024 --prompt "..." --output .omnius/images/out.png',
         category: "Modern deployable",
         sizeClass: "4B compact FLUX-family",
         quality: "Bridge between practical deployment and modern FLUX-family visual quality.",
@@ -250843,7 +250857,7 @@ var init_image_generate = __esm({
         id: "segmind/tiny-sd",
         label: "Segmind Tiny-SD",
         backend: "diffusers",
-        install: 'python .omnius/image-gen/diffusers_text2image.py --model segmind/tiny-sd --prompt "..." --output .omnius/images/out.png',
+        install: 'python3 .omnius/image-gen/diffusers_text2image.py --model segmind/tiny-sd --prompt "..." --output .omnius/images/out.png',
         category: "Lightweight smoke test",
         sizeClass: "Small SD-compatible",
         quality: "Fast validation model; not a serious photorealism baseline.",
@@ -250860,7 +250874,7 @@ var init_image_generate = __esm({
         id: "nota-ai/bk-sdm-tiny-2m",
         label: "BK-SDM Tiny 2M",
         backend: "diffusers",
-        install: 'python .omnius/image-gen/diffusers_text2image.py --model nota-ai/bk-sdm-tiny-2m --prompt "..." --output .omnius/images/out.png',
+        install: 'python3 .omnius/image-gen/diffusers_text2image.py --model nota-ai/bk-sdm-tiny-2m --prompt "..." --output .omnius/images/out.png',
         category: "Lightweight smoke test",
         sizeClass: "Compressed SD-compatible",
         quality: "Very small and practical; quality is mainly for tests and rough drafts.",
@@ -250877,7 +250891,7 @@ var init_image_generate = __esm({
         id: "nota-ai/bk-sdm-small-2m",
         label: "BK-SDM Small 2M",
         backend: "diffusers",
-        install: 'python .omnius/image-gen/diffusers_text2image.py --model nota-ai/bk-sdm-small-2m --prompt "..." --output .omnius/images/out.png',
+        install: 'python3 .omnius/image-gen/diffusers_text2image.py --model nota-ai/bk-sdm-small-2m --prompt "..." --output .omnius/images/out.png',
         category: "Lightweight smoke test",
         sizeClass: "Compressed SD-compatible",
         quality: "Slightly better compressed-SD quality than tiny variants; still not a high-fidelity baseline.",
@@ -250894,7 +250908,7 @@ var init_image_generate = __esm({
         id: "SimianLuo/LCM_Dreamshaper_v7",
         label: "LCM DreamShaper v7",
         backend: "diffusers",
-        install: 'python .omnius/image-gen/diffusers_text2image.py --model SimianLuo/LCM_Dreamshaper_v7 --steps 4 --prompt "..." --output .omnius/images/out.png',
+        install: 'python3 .omnius/image-gen/diffusers_text2image.py --model SimianLuo/LCM_Dreamshaper_v7 --steps 4 --prompt "..." --output .omnius/images/out.png',
         category: "Fast iteration",
         sizeClass: "Few-step SD-compatible",
         quality: "Good for low-latency concepting; below SDXL/SD3.5/FLUX for photoreal detail.",
@@ -250911,7 +250925,7 @@ var init_image_generate = __esm({
         id: "stabilityai/sd-turbo",
         label: "SD-Turbo",
         backend: "diffusers",
-        install: 'python .omnius/image-gen/diffusers_text2image.py --model stabilityai/sd-turbo --steps 1 --guidance 0 --prompt "..." --output .omnius/images/out.png',
+        install: 'python3 .omnius/image-gen/diffusers_text2image.py --model stabilityai/sd-turbo --steps 1 --guidance 0 --prompt "..." --output .omnius/images/out.png',
         category: "Fast iteration",
         sizeClass: "One-to-four-step SD",
         quality: "Fast SD-family output; useful for iteration but lower ceiling than SDXL Turbo and large baselines.",
@@ -250928,7 +250942,7 @@ var init_image_generate = __esm({
         id: DEFAULT_DIFFUSERS_IMAGE_MODEL,
         label: "SDXL-Turbo",
         backend: "diffusers",
-        install: 'python .omnius/image-gen/diffusers_text2image.py --model stabilityai/sdxl-turbo --steps 1 --guidance 0 --width 1024 --height 1024 --prompt "..." --output .omnius/images/out.png',
+        install: 'python3 .omnius/image-gen/diffusers_text2image.py --model stabilityai/sdxl-turbo --steps 1 --guidance 0 --width 1024 --height 1024 --prompt "..." --output .omnius/images/out.png',
         category: "Default local generation",
         sizeClass: "Few-step SDXL",
         quality: "Strong fast default for local image generation; not as realistic as FLUX.1 dev or SD3.5 Large, but much more practical.",
@@ -250945,7 +250959,7 @@ var init_image_generate = __esm({
         id: "Efficient-Large-Model/Sana_Sprint_0.6B_1024px_diffusers",
         label: "Sana Sprint 0.6B",
         backend: "diffusers",
-        install: 'python .omnius/image-gen/diffusers_text2image.py --model Efficient-Large-Model/Sana_Sprint_0.6B_1024px_diffusers --steps 4 --guidance 0 --width 1024 --height 1024 --prompt "..." --output .omnius/images/out.png',
+        install: 'python3 .omnius/image-gen/diffusers_text2image.py --model Efficient-Large-Model/Sana_Sprint_0.6B_1024px_diffusers --steps 4 --guidance 0 --width 1024 --height 1024 --prompt "..." --output .omnius/images/out.png',
         category: "Modern efficient",
         sizeClass: "0.6B efficient diffusion",
         quality: "Modern efficient output under smaller compute budgets; below primary large baselines.",
@@ -250962,7 +250976,7 @@ var init_image_generate = __esm({
         id: "sdcpp:local",
         label: "stable-diffusion.cpp local checkpoint",
         backend: "sdcpp",
-        install: 'python .omnius/image-gen/sdcpp_text2image.py --model-path /path/to/model.gguf --prompt "..." --output .omnius/images/out.png',
+        install: 'python3 .omnius/image-gen/sdcpp_text2image.py --model-path /path/to/model.gguf --prompt "..." --output .omnius/images/out.png',
         category: "Local checkpoint/GGUF",
         sizeClass: "Depends on checkpoint",
         quality: "Quality depends entirely on the local checkpoint or GGUF variant.",
@@ -251417,15 +251431,17 @@ ${errText.slice(0, 1200)}`,
         });
         if (result.code !== 0) {
           const plan = imageGenerationSetupPlan("diffusers", this.cwd, args.model);
+          const output = [
+            `Diffusers model prewarm failed with exit code ${result.code ?? "unknown"}.`,
+            formatDiffusersFailure(result.stderr || result.stdout),
+            "",
+            "Setup path:",
+            ...plan.commands.map((cmd) => `  ${cmd}`)
+          ].filter(Boolean).join("\n");
           return {
             success: false,
-            output: [
-              `Diffusers model prewarm failed with exit code ${result.code ?? "unknown"}.`,
-              trimProcessText(result.stderr || result.stdout),
-              "",
-              "Setup path:",
-              ...plan.commands.map((cmd) => `  ${cmd}`)
-            ].filter(Boolean).join("\n"),
+            output,
+            error: output,
             durationMs: performance.now() - args.start
           };
         }
@@ -251600,15 +251616,17 @@ ${errText.slice(0, 800)}`,
         });
         if (result.code !== 0 || !existsSync23(filepath)) {
           const plan = imageGenerationSetupPlan("diffusers", this.cwd, args.model);
+          const output2 = [
+            `Diffusers image generation failed with exit code ${result.code ?? "unknown"}.`,
+            formatDiffusersFailure(result.stderr || result.stdout),
+            "",
+            "Setup path:",
+            ...plan.commands.map((cmd) => `  ${cmd}`)
+          ].filter(Boolean).join("\n");
           return {
             success: false,
-            output: [
-              `Diffusers image generation failed with exit code ${result.code ?? "unknown"}.`,
-              trimProcessText(result.stderr || result.stdout),
-              "",
-              "Setup path:",
-              ...plan.commands.map((cmd) => `  ${cmd}`)
-            ].filter(Boolean).join("\n"),
+            output: output2,
+            error: output2,
             durationMs: performance.now() - args.start
           };
         }
@@ -251796,6 +251814,8 @@ function audioOutputDir(repoRoot = ".") {
   return join37(repoRoot, ".omnius", "audio");
 }
 function backendPackages(backend) {
+  if (backend === "transformers")
+    return TRANSFORMERS_AUDIO_PACKAGES;
   if (backend === "audiocraft")
     return AUDIOCRAFT_PACKAGES;
   if (backend === "stable-audio")
@@ -251805,6 +251825,8 @@ function backendPackages(backend) {
   return DIFFUSERS_AUDIO_PACKAGES;
 }
 function backendImportCheck(backend) {
+  if (backend === "transformers")
+    return "import torch, torchaudio, transformers, scipy\nfrom transformers import AutoProcessor, MusicgenForConditionalGeneration\n";
   if (backend === "audiocraft")
     return "import torch, torchaudio, audiocraft\nfrom audiocraft.models import MusicGen, AudioGen\n";
   if (backend === "stable-audio")
@@ -251933,8 +251955,23 @@ function trimProcessText2(text, max = 1800) {
   return clean3.slice(0, max - 20) + "\n... (truncated)";
 }
 async function pythonCanImport2(command, code8, repoRoot, env2) {
-  const result = await runProcess3(command, ["-c", code8], { cwd: repoRoot, timeoutMs: 6e4, env: env2 });
-  return result.code === 0;
+  return (await pythonImportResult(command, code8, repoRoot, env2)).code === 0;
+}
+async function pythonImportResult(command, code8, repoRoot, env2) {
+  return await runProcess3(command, ["-c", code8], { cwd: repoRoot, timeoutMs: 6e4, env: env2 });
+}
+function formatAudioSetupFailure(backend, text) {
+  const body = trimProcessText2(text);
+  const lowered = text.toLowerCase();
+  const notes2 = [];
+  if (backend === "audiocraft" && (lowered.includes("libavformat") || lowered.includes("pkg-config") || lowered.includes("pyav") || lowered.includes(" av"))) {
+    notes2.push("AudioCraft pulls PyAV. If a matching PyAV wheel is unavailable, the system must provide FFmpeg development headers.", "Ubuntu/Debian fix: sudo apt install -y pkg-config ffmpeg libavformat-dev libavcodec-dev libavdevice-dev libavutil-dev libavfilter-dev libswscale-dev libswresample-dev", "For MusicGen, prefer the default Transformers backend: /music setup transformers or /music --backend transformers <prompt>.");
+  }
+  if (lowered.includes("cuda") && lowered.includes("not available")) {
+    notes2.push("CUDA was not available to the selected Python environment; install a Torch build matching this machine's CUDA runtime or use CPU-compatible settings.");
+  }
+  return [body, ...notes2.map((note) => `
+${note}`)].filter(Boolean).join("");
 }
 async function ensurePythonForAudio(repoRoot, backend, explicit, onProgress) {
   const pythonEnv = audioGenerationPythonEnv(repoRoot);
@@ -251965,7 +252002,19 @@ ${trimProcessText2(created.stderr || created.stdout)}`);
   }
   const packages = backendPackages(backend);
   onProgress?.({ stage: "setup", message: `Installing ${backend} audio-generation Python packages` });
-  const pip = await runProcess3(command, ["-m", "pip", "install", "--progress-bar", "on", "-U", "pip", ...packages], {
+  const pipArgs = [
+    "-m",
+    "pip",
+    "install",
+    "--progress-bar",
+    "on",
+    "--prefer-binary",
+    ...backend === "audiocraft" ? ["--only-binary", "av"] : [],
+    "-U",
+    "pip",
+    ...packages
+  ];
+  const pip = await runProcess3(command, pipArgs, {
     cwd: repoRoot,
     timeoutMs: 18e5,
     env: pythonEnv,
@@ -251974,18 +252023,20 @@ ${trimProcessText2(created.stderr || created.stdout)}`);
   });
   if (pip.code !== 0) {
     throw new Error(`Failed to install ${backend} audio-generation packages into ${venvDir}.
-${trimProcessText2(pip.stderr || pip.stdout)}`);
+${formatAudioSetupFailure(backend, pip.stderr || pip.stdout)}`);
   }
-  if (!await pythonCanImport2(command, backendImportCheck(backend), repoRoot, pythonEnv)) {
-    throw new Error(`Audio-generation Python environment at ${venvDir} was created, but required ${backend} imports still fail.`);
+  const importCheck = await pythonImportResult(command, backendImportCheck(backend), repoRoot, pythonEnv);
+  if (importCheck.code !== 0) {
+    throw new Error(`Audio-generation Python environment at ${venvDir} was created, but required ${backend} imports still fail.
+${formatAudioSetupFailure(backend, importCheck.stderr || importCheck.stdout)}`);
   }
   return { command, env: pythonEnv };
 }
 async function ensureAudioRunner(repoRoot, backend) {
   const dir = audioGenerationDir(repoRoot);
   await mkdir12(dir, { recursive: true });
-  const script = backend === "audiocraft" ? join37(dir, "audiocraft_audio.py") : backend === "stable-audio" ? join37(dir, "stable_audio.py") : backend === "tangoflux" ? join37(dir, "tangoflux_audio.py") : join37(dir, "diffusers_audio.py");
-  const body = backend === "audiocraft" ? AUDIOCRAFT_RUNNER : DIFFUSERS_AUDIO_RUNNER;
+  const script = backend === "transformers" ? join37(dir, "transformers_audio.py") : backend === "audiocraft" ? join37(dir, "audiocraft_audio.py") : backend === "stable-audio" ? join37(dir, "stable_audio.py") : backend === "tangoflux" ? join37(dir, "tangoflux_audio.py") : join37(dir, "diffusers_audio.py");
+  const body = backend === "transformers" ? TRANSFORMERS_AUDIO_RUNNER : backend === "audiocraft" ? AUDIOCRAFT_RUNNER : DIFFUSERS_AUDIO_RUNNER;
   await writeFile17(script, body, "utf8");
   await chmod4(script, 493).catch(() => {
   });
@@ -252040,7 +252091,7 @@ function getAudioGenerationPreset(model, kind) {
 }
 function inferAudioGenerationBackend(model, requested) {
   if (requested && requested !== "auto") {
-    if (requested === "diffusers" || requested === "audiocraft" || requested === "stable-audio" || requested === "tangoflux" || requested === "project")
+    if (requested === "diffusers" || requested === "transformers" || requested === "audiocraft" || requested === "stable-audio" || requested === "tangoflux" || requested === "project")
       return requested;
   }
   const preset = getAudioGenerationPreset(model);
@@ -252049,7 +252100,9 @@ function inferAudioGenerationBackend(model, requested) {
   if (!model)
     return "auto";
   const lowered = model.toLowerCase();
-  if (lowered.includes("musicgen") || lowered.includes("audiogen"))
+  if (lowered.includes("musicgen"))
+    return "transformers";
+  if (lowered.includes("audiogen"))
     return "audiocraft";
   if (lowered.includes("stable-audio"))
     return "stable-audio";
@@ -252071,7 +252124,7 @@ function audioGenerationSetupPlan(kind, backend, repoRoot = ".", model) {
       title: `Diffusers ${kind} generation`,
       commands: [
         `python3 -m venv ${venvDir}`,
-        `${venvPython(venvDir)} -m pip install -U pip ${DIFFUSERS_AUDIO_PACKAGES.join(" ")}`,
+        `${venvPython(venvDir)} -m pip install --prefer-binary -U pip ${DIFFUSERS_AUDIO_PACKAGES.join(" ")}`,
         `omnius /${commandName} "cinematic rain on a neon street" --backend diffusers --model ${chosen}`
       ],
       notes: [
@@ -252081,6 +252134,24 @@ function audioGenerationSetupPlan(kind, backend, repoRoot = ".", model) {
       ]
     };
   }
+  if (backend === "transformers") {
+    const venvDir = audioGenerationVenvDir(repoRoot, "transformers");
+    return {
+      kind,
+      backend,
+      title: `Transformers ${kind} generation`,
+      commands: [
+        `python3 -m venv ${venvDir}`,
+        `${venvPython(venvDir)} -m pip install --prefer-binary -U pip ${TRANSFORMERS_AUDIO_PACKAGES.join(" ")}`,
+        `omnius /${commandName} "warm analog synth arpeggio, slow drums" --backend transformers --model ${chosen}`
+      ],
+      notes: [
+        "Use this path for MusicGen without requiring AudioCraft/PyAV/libavformat system headers.",
+        "The venv, Hugging Face cache, Torch cache, and pip cache stay under .omnius/audio-gen.",
+        "First generation downloads model weights and then immediately generates the requested audio."
+      ]
+    };
+  }
   if (backend === "audiocraft") {
     const venvDir = audioGenerationVenvDir(repoRoot, "audiocraft");
     return {
@@ -252089,12 +252160,14 @@ function audioGenerationSetupPlan(kind, backend, repoRoot = ".", model) {
       title: `AudioCraft ${kind} generation`,
       commands: [
         `python3 -m venv ${venvDir}`,
-        `${venvPython(venvDir)} -m pip install -U pip ${AUDIOCRAFT_PACKAGES.join(" ")}`,
+        `${venvPython(venvDir)} -m pip install --prefer-binary --only-binary av -U pip ${AUDIOCRAFT_PACKAGES.join(" ")}`,
         `omnius /${commandName} "warm analog synth arpeggio, slow drums" --backend audiocraft --model ${chosen}`
       ],
       notes: [
-        "Use this path for MusicGen and AudioGen models.",
+        "Use this path for AudioCraft AudioGen models or explicit upstream AudioCraft testing.",
         "AudioCraft is a larger install; prefer CUDA and enough VRAM for the selected checkpoint.",
+        "If PyAV has no wheel for this Python version, install FFmpeg development headers first: sudo apt install -y pkg-config ffmpeg libavformat-dev libavcodec-dev libavdevice-dev libavutil-dev libavfilter-dev libswscale-dev libswresample-dev.",
+        "For MusicGen, the default Transformers backend avoids the PyAV/libavformat system-header path.",
         "First generation downloads model weights and saves WAV files under .omnius/audio."
       ]
     };
@@ -252107,7 +252180,7 @@ function audioGenerationSetupPlan(kind, backend, repoRoot = ".", model) {
       title: `Stable Audio ${kind} generation`,
       commands: [
         `python3 -m venv ${venvDir}`,
-        `${venvPython(venvDir)} -m pip install -U pip ${STABLE_AUDIO_PACKAGES.join(" ")}`,
+        `${venvPython(venvDir)} -m pip install --prefer-binary -U pip ${STABLE_AUDIO_PACKAGES.join(" ")}`,
         `omnius /${commandName} "high fidelity stereo ${kind} bed, detailed and clean" --backend stable-audio --model ${chosen}`
       ],
       notes: [
@@ -252124,7 +252197,7 @@ function audioGenerationSetupPlan(kind, backend, repoRoot = ".", model) {
       title: "TangoFlux sound generation",
       commands: [
         `python3 -m venv ${venvDir}`,
-        `${venvPython(venvDir)} -m pip install -U pip ${TANGOFLUX_PACKAGES.join(" ")}`,
+        `${venvPython(venvDir)} -m pip install --prefer-binary -U pip ${TANGOFLUX_PACKAGES.join(" ")}`,
         `omnius /sound "fast whoosh impact with metallic tail" --backend tangoflux --model ${chosen}`
       ],
       notes: [
@@ -252147,7 +252220,7 @@ function audioGenerationSetupPlan(kind, backend, repoRoot = ".", model) {
     ]
   };
 }
-var DEFAULT_SOUND_MODEL, DEFAULT_MUSIC_MODEL, DIFFUSERS_AUDIO_PACKAGES, AUDIOCRAFT_PACKAGES, STABLE_AUDIO_PACKAGES, TANGOFLUX_PACKAGES, AUDIO_GENERATION_MODEL_PRESETS, DIFFUSERS_AUDIO_RUNNER, AUDIOCRAFT_RUNNER, AudioGenerateTool;
+var DEFAULT_SOUND_MODEL, DEFAULT_MUSIC_MODEL, DIFFUSERS_AUDIO_PACKAGES, TRANSFORMERS_AUDIO_PACKAGES, AUDIOCRAFT_PACKAGES, STABLE_AUDIO_PACKAGES, TANGOFLUX_PACKAGES, AUDIO_GENERATION_MODEL_PRESETS, DIFFUSERS_AUDIO_RUNNER, AUDIOCRAFT_RUNNER, TRANSFORMERS_AUDIO_RUNNER, AudioGenerateTool;
 var init_audio_generate = __esm({
   "packages/execution/dist/tools/audio-generate.js"() {
     "use strict";
@@ -252163,6 +252236,16 @@ var init_audio_generate = __esm({
       "soundfile",
       "librosa"
     ];
+    TRANSFORMERS_AUDIO_PACKAGES = [
+      "torch",
+      "torchaudio",
+      "transformers",
+      "accelerate",
+      "scipy",
+      "soundfile",
+      "sentencepiece",
+      "protobuf"
+    ];
     AUDIOCRAFT_PACKAGES = [
       "torch",
       "torchaudio",
@@ -252190,7 +252273,7 @@ var init_audio_generate = __esm({
         label: "AudioLDM S-Full v2",
         kind: "sound",
         backend: "diffusers",
-        install: 'python .omnius/audio-gen/diffusers_audio.py --kind sound --model cvssp/audioldm-s-full-v2 --duration 8 --prompt "..." --output .omnius/audio/out.wav',
+        install: 'python3 .omnius/audio-gen/diffusers_audio.py --kind sound --model cvssp/audioldm-s-full-v2 --duration 8 --prompt "..." --output .omnius/audio/out.wav',
         category: "Default practical sound effects",
         sizeClass: "Small text-to-audio latent diffusion",
         quality: "Good practical baseline for SFX, ambience, foley, and quick local validation.",
@@ -252208,7 +252291,7 @@ var init_audio_generate = __esm({
         label: "AudioLDM 2",
         kind: "sound",
         backend: "diffusers",
-        install: 'python .omnius/audio-gen/diffusers_audio.py --kind sound --model cvssp/audioldm2 --duration 10 --prompt "..." --output .omnius/audio/out.wav',
+        install: 'python3 .omnius/audio-gen/diffusers_audio.py --kind sound --model cvssp/audioldm2 --duration 10 --prompt "..." --output .omnius/audio/out.wav',
         category: "General sound effects",
         sizeClass: "Text-to-audio latent diffusion",
         quality: "More capable general sound synthesis than AudioLDM S; slower at high step counts.",
@@ -252226,7 +252309,7 @@ var init_audio_generate = __esm({
         label: "AudioLDM 2 Large",
         kind: "sound",
         backend: "diffusers",
-        install: 'python .omnius/audio-gen/diffusers_audio.py --kind sound --model cvssp/audioldm2-large --duration 10 --prompt "..." --output .omnius/audio/out.wav',
+        install: 'python3 .omnius/audio-gen/diffusers_audio.py --kind sound --model cvssp/audioldm2-large --duration 10 --prompt "..." --output .omnius/audio/out.wav',
         category: "High quality sound effects",
         sizeClass: "Large text-to-audio latent diffusion",
         quality: "Higher ceiling than AudioLDM 2 base, with slower sampling and more VRAM pressure.",
@@ -252244,7 +252327,7 @@ var init_audio_generate = __esm({
         label: "AudioGen Medium",
         kind: "sound",
         backend: "audiocraft",
-        install: 'python .omnius/audio-gen/audiocraft_audio.py --kind sound --model facebook/audiogen-medium --duration 8 --prompt "..." --output .omnius/audio/out.wav',
+        install: 'python3 .omnius/audio-gen/audiocraft_audio.py --kind sound --model facebook/audiogen-medium --duration 8 --prompt "..." --output .omnius/audio/out.wav',
         category: "AudioCraft sound effects",
         sizeClass: "AudioCraft text-to-sound",
         quality: "Strong text-conditioned non-music sound effects; good for foley and environments.",
@@ -252261,7 +252344,7 @@ var init_audio_generate = __esm({
         label: "TANGO 2",
         kind: "sound",
         backend: "project",
-        install: "git clone https://github.com/declare-lab/tango .omnius/audio-gen/projects/tango && cd .omnius/audio-gen/projects/tango && python -m pip install -r requirements.txt",
+        install: "git clone https://github.com/declare-lab/tango .omnius/audio-gen/projects/tango && cd .omnius/audio-gen/projects/tango && python3 -m pip install -r requirements.txt",
         category: "Research sound effects",
         sizeClass: "DPO-aligned text-to-audio research stack",
         quality: "Good prompt alignment target, but less turnkey than Diffusers or AudioCraft.",
@@ -252278,7 +252361,7 @@ var init_audio_generate = __esm({
         label: "TangoFlux",
         kind: "sound",
         backend: "tangoflux",
-        install: 'python .omnius/audio-gen/tangoflux_audio.py --model declare-lab/TangoFlux --duration 8 --prompt "..." --output .omnius/audio/out.wav',
+        install: 'python3 .omnius/audio-gen/tangoflux_audio.py --model declare-lab/TangoFlux --duration 8 --prompt "..." --output .omnius/audio/out.wav',
         category: "Fast sound effects",
         sizeClass: "Fast text-to-audio generation",
         quality: "Fast SFX and ambience iteration, with newer-stack stability risks.",
@@ -252295,7 +252378,7 @@ var init_audio_generate = __esm({
         label: "Stable Audio Open 1.0",
         kind: "sound",
         backend: "stable-audio",
-        install: 'python .omnius/audio-gen/stable_audio.py --kind sound --model stabilityai/stable-audio-open-1.0 --duration 20 --prompt "..." --output .omnius/audio/out.wav',
+        install: 'python3 .omnius/audio-gen/stable_audio.py --kind sound --model stabilityai/stable-audio-open-1.0 --duration 20 --prompt "..." --output .omnius/audio/out.wav',
         category: "High-quality long audio",
         sizeClass: "Stereo 44.1 kHz diffusion audio",
         quality: "High-fidelity stereo generation, useful for longer designed sounds and music-like clips.",
@@ -252312,16 +252395,16 @@ var init_audio_generate = __esm({
         id: DEFAULT_MUSIC_MODEL,
         label: "MusicGen Small",
         kind: "music",
-        backend: "audiocraft",
-        install: 'python .omnius/audio-gen/audiocraft_audio.py --kind music --model facebook/musicgen-small --duration 20 --prompt "..." --output .omnius/audio/out.wav',
+        backend: "transformers",
+        install: 'python3 .omnius/audio-gen/transformers_audio.py --kind music --model facebook/musicgen-small --duration 20 --prompt "..." --output .omnius/audio/out.wav',
         category: "Default practical music",
-        sizeClass: "Small AudioCraft music model",
+        sizeClass: "Small Transformers MusicGen model",
         quality: "Fast, practical music sketches with lower fidelity and complexity than medium/large models.",
         output: "Short mono/stereo music sketches depending on checkpoint.",
         bestUse: "Default /music smoke test and quick ideas.",
         minVramGB: 6,
         recommendedVramGB: 8,
-        deployment: "AudioCraft path; fastest usable MusicGen option.",
+        deployment: "Transformers path; fastest usable MusicGen option.",
         defaultDurationSec: 20,
         note: "Practical default music generation model."
       },
@@ -252329,16 +252412,16 @@ var init_audio_generate = __esm({
         id: "facebook/musicgen-medium",
         label: "MusicGen Medium",
         kind: "music",
-        backend: "audiocraft",
-        install: 'python .omnius/audio-gen/audiocraft_audio.py --kind music --model facebook/musicgen-medium --duration 20 --prompt "..." --output .omnius/audio/out.wav',
+        backend: "transformers",
+        install: 'python3 .omnius/audio-gen/transformers_audio.py --kind music --model facebook/musicgen-medium --duration 20 --prompt "..." --output .omnius/audio/out.wav',
         category: "Higher quality MusicGen",
-        sizeClass: "Medium AudioCraft music model",
+        sizeClass: "Medium Transformers MusicGen model",
         quality: "Better arrangement and texture than small; more VRAM and latency.",
         output: "Short music clips.",
         bestUse: "Local music generation when 12-16GB VRAM is available.",
         minVramGB: 10,
         recommendedVramGB: 16,
-        deployment: "AudioCraft path; good quality/footprint balance.",
+        deployment: "Transformers path; good quality/footprint balance.",
         defaultDurationSec: 20,
         note: "Balanced MusicGen quality target."
       },
@@ -252346,16 +252429,16 @@ var init_audio_generate = __esm({
         id: "facebook/musicgen-large",
         label: "MusicGen Large",
         kind: "music",
-        backend: "audiocraft",
-        install: 'python .omnius/audio-gen/audiocraft_audio.py --kind music --model facebook/musicgen-large --duration 20 --prompt "..." --output .omnius/audio/out.wav',
+        backend: "transformers",
+        install: 'python3 .omnius/audio-gen/transformers_audio.py --kind music --model facebook/musicgen-large --duration 20 --prompt "..." --output .omnius/audio/out.wav',
         category: "Higher quality MusicGen",
-        sizeClass: "Large AudioCraft music model",
+        sizeClass: "Large Transformers MusicGen model",
         quality: "Stronger musical structure and detail, with high memory pressure.",
         output: "Higher-quality short music clips.",
         bestUse: "High-VRAM local GPU or remote GPU music generation.",
         minVramGB: 16,
         recommendedVramGB: 24,
-        deployment: "AudioCraft path; expect large downloads and slower inference.",
+        deployment: "Transformers path; expect large downloads and slower inference.",
         defaultDurationSec: 20,
         note: "Large MusicGen baseline."
       },
@@ -252363,16 +252446,16 @@ var init_audio_generate = __esm({
         id: "facebook/musicgen-melody",
         label: "MusicGen Melody",
         kind: "music",
-        backend: "audiocraft",
-        install: 'python .omnius/audio-gen/audiocraft_audio.py --kind music --model facebook/musicgen-melody --duration 20 --prompt "..." --output .omnius/audio/out.wav',
+        backend: "transformers",
+        install: 'python3 .omnius/audio-gen/transformers_audio.py --kind music --model facebook/musicgen-melody --duration 20 --prompt "..." --output .omnius/audio/out.wav',
         category: "Conditioned music",
-        sizeClass: "Melody-conditioned AudioCraft music model",
+        sizeClass: "Melody-conditioned Transformers MusicGen model",
         quality: "Useful when a reference melody path is added to the generation flow.",
         output: "Music clips guided by text and optional melody conditioning.",
         bestUse: "Future melody-conditioned workflows; useful to pre-deploy now.",
         minVramGB: 10,
         recommendedVramGB: 16,
-        deployment: "AudioCraft path; melody conditioning needs an additional reference audio argument.",
+        deployment: "Transformers path; melody conditioning needs an additional reference audio argument.",
         defaultDurationSec: 20,
         note: "MusicGen variant for melody-conditioned generation."
       },
@@ -252380,16 +252463,16 @@ var init_audio_generate = __esm({
         id: "facebook/musicgen-stereo-large",
         label: "MusicGen Stereo Large",
         kind: "music",
-        backend: "audiocraft",
-        install: 'python .omnius/audio-gen/audiocraft_audio.py --kind music --model facebook/musicgen-stereo-large --duration 20 --prompt "..." --output .omnius/audio/out.wav',
+        backend: "transformers",
+        install: 'python3 .omnius/audio-gen/transformers_audio.py --kind music --model facebook/musicgen-stereo-large --duration 20 --prompt "..." --output .omnius/audio/out.wav',
         category: "Higher quality MusicGen",
-        sizeClass: "Large stereo AudioCraft music model",
+        sizeClass: "Large stereo Transformers MusicGen model",
         quality: "Stereo large MusicGen path; best MusicGen quality listed here but heavy.",
         output: "Stereo music clips.",
         bestUse: "High-VRAM machines where stereo output matters.",
         minVramGB: 20,
         recommendedVramGB: 32,
-        deployment: "AudioCraft path; prefer high-VRAM local GPUs or remote workers.",
+        deployment: "Transformers path; prefer high-VRAM local GPUs or remote workers.",
         defaultDurationSec: 20,
         note: "Heavy stereo MusicGen option."
       },
@@ -252398,7 +252481,7 @@ var init_audio_generate = __esm({
         label: "Stable Audio Open 1.0",
         kind: "music",
         backend: "stable-audio",
-        install: 'python .omnius/audio-gen/stable_audio.py --kind music --model stabilityai/stable-audio-open-1.0 --duration 30 --prompt "..." --output .omnius/audio/out.wav',
+        install: 'python3 .omnius/audio-gen/stable_audio.py --kind music --model stabilityai/stable-audio-open-1.0 --duration 30 --prompt "..." --output .omnius/audio/out.wav',
         category: "Primary serious music baseline",
         sizeClass: "Stereo 44.1 kHz diffusion audio",
         quality: "Primary serious open music/audio baseline for higher-fidelity stereo clips.",
@@ -252416,7 +252499,7 @@ var init_audio_generate = __esm({
         label: "MAGNeT AudioCraft Profile",
         kind: "music",
         backend: "project",
-        install: "python -m pip install -U audiocraft && inspect the AudioCraft MAGNeT examples before enabling generation",
+        install: "python3 -m pip install -U audiocraft && inspect the AudioCraft MAGNeT examples before enabling generation",
         category: "Research music",
         sizeClass: "Non-autoregressive AudioCraft research model",
         quality: "Interesting fast music research path; less standard than MusicGen for production.",
@@ -252433,7 +252516,7 @@ var init_audio_generate = __esm({
         label: "JASCO AudioCraft Profile",
         kind: "music",
         backend: "project",
-        install: "python -m pip install -U audiocraft && inspect the AudioCraft JASCO examples before enabling generation",
+        install: "python3 -m pip install -U audiocraft && inspect the AudioCraft JASCO examples before enabling generation",
         category: "Conditioned music",
         sizeClass: "Chord/melody/drum conditioned research model",
         quality: "Promising controlled music path when conditioning inputs are available.",
@@ -252467,7 +252550,7 @@ var init_audio_generate = __esm({
         label: "Riffusion v1",
         kind: "music",
         backend: "diffusers",
-        install: 'python .omnius/audio-gen/diffusers_audio.py --kind music --model riffusion/riffusion-model-v1 --duration 8 --prompt "..." --output .omnius/audio/out.wav',
+        install: 'python3 .omnius/audio-gen/diffusers_audio.py --kind music --model riffusion/riffusion-model-v1 --duration 8 --prompt "..." --output .omnius/audio/out.wav',
         category: "Legacy/specialized music",
         sizeClass: "Spectrogram diffusion",
         quality: "Historically important and fun, but below MusicGen and Stable Audio Open for general quality.",
@@ -252641,19 +252724,100 @@ def main():
     torchaudio.save(str(out), wav, model.sample_rate)
     print(json.dumps({"ok": True, "path": str(out), "model": args.model, "backend": "audiocraft", "sample_rate": model.sample_rate, "seconds": round(time.perf_counter() - t0, 3)}))
 
+if __name__ == "__main__":
+    main()
+`;
+    TRANSFORMERS_AUDIO_RUNNER = String.raw`#!/usr/bin/env python3
+import argparse, json, sys, time
+from pathlib import Path
+
+def _progress(stage, message, percent=None):
+    payload = {"omnius_progress": True, "stage": stage, "message": message}
+    if percent is not None:
+        payload["percent"] = percent
+    print(json.dumps(payload), file=sys.stderr, flush=True)
+
+def _device():
+    import torch
+    if torch.cuda.is_available():
+        return "cuda"
+    if hasattr(torch.backends, "mps") and torch.backends.mps.is_available():
+        return "mps"
+    return "cpu"
+
+def _write_wav(path, sample_rate, audio):
+    import numpy as np
+    from scipy.io.wavfile import write
+    arr = np.asarray(audio)
+    if arr.ndim > 2:
+        arr = arr.squeeze()
+    if arr.ndim == 2 and arr.shape[0] < arr.shape[1]:
+        arr = arr.T
+    arr = np.nan_to_num(arr)
+    peak = float(np.max(np.abs(arr))) if arr.size else 0.0
+    if peak > 1.0:
+        arr = arr / peak
+    arr_i16 = (np.clip(arr, -1.0, 1.0) * 32767).astype(np.int16)
+    write(str(path), sample_rate, arr_i16)
+
+def main():
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--kind", choices=["sound", "music"], required=True)
+    parser.add_argument("--model", required=True)
+    parser.add_argument("--prompt", required=True)
+    parser.add_argument("--output", required=True)
+    parser.add_argument("--duration", type=float, default=20)
+    parser.add_argument("--seed", type=int, default=None)
+    parser.add_argument("--prewarm", action="store_true")
+    args = parser.parse_args()
+
+    t0 = time.perf_counter()
+    import torch
+    from transformers import AutoProcessor, MusicgenForConditionalGeneration
+
+    device = _device()
+    dtype = torch.float16 if device == "cuda" else torch.float32
+    _progress("load", f"loading MusicGen model {args.model}")
+    processor = AutoProcessor.from_pretrained(args.model)
+    model = MusicgenForConditionalGeneration.from_pretrained(args.model, torch_dtype=dtype)
+    model = model.to(device)
+    sample_rate = int(getattr(model.config.audio_encoder, "sampling_rate", 32000))
+    _progress("load", f"model loaded on {device}")
+
+    if args.prewarm:
+        _progress("load", f"prewarmed {args.model} on {device}", 100)
+        print(json.dumps({"ok": True, "path": "", "model": args.model, "backend": "transformers", "sample_rate": sample_rate, "prewarm": True, "seconds": round(time.perf_counter() - t0, 3)}))
+        return
+
+    if args.seed is not None:
+        torch.manual_seed(args.seed)
+    inputs = processor(text=[args.prompt], padding=True, return_tensors="pt")
+    inputs = {k: v.to(device) for k, v in inputs.items()}
+    # MusicGen uses roughly 50 generated tokens per second of audio.
+    max_new_tokens = max(32, min(1536, int(args.duration * 50)))
+    _progress("generate", f"generating about {args.duration}s audio with {max_new_tokens} tokens")
+    with torch.no_grad():
+        audio_values = model.generate(**inputs, max_new_tokens=max_new_tokens)
+    audio = audio_values[0, 0].detach().cpu().float().numpy()
+    out = Path(args.output)
+    out.parent.mkdir(parents=True, exist_ok=True)
+    _progress("save", f"saving audio to {out}")
+    _write_wav(out, sample_rate, audio)
+    print(json.dumps({"ok": True, "path": str(out), "model": args.model, "backend": "transformers", "sample_rate": sample_rate, "seconds": round(time.perf_counter() - t0, 3)}))
+
 if __name__ == "__main__":
     main()
 `;
     AudioGenerateTool = class {
       name = "generate_audio";
-      description = "Generate a sound effect or music clip from a text prompt using local audio-generation backends. Supports Diffusers AudioLDM/AudioLDM2, AudioCraft MusicGen/AudioGen, Stable Audio Open deployment paths, and explicit research-project profiles. Saves WAV files under .omnius/audio and returns the file path.";
+      description = "Generate a sound effect or music clip from a text prompt using local audio-generation backends. Supports Diffusers AudioLDM/AudioLDM2, Transformers MusicGen, AudioCraft AudioGen, Stable Audio Open deployment paths, and explicit research-project profiles. Saves WAV files under .omnius/audio and returns the file path.";
       parameters = {
         type: "object",
         properties: {
           kind: { type: "string", enum: ["sound", "music"], description: "Generate a sound effect/ambience clip or a music clip" },
           prompt: { type: "string", description: "Text description of the audio to generate" },
           model: { type: "string", description: "Audio model id, e.g. cvssp/audioldm2 or facebook/musicgen-small" },
-          backend: { type: "string", enum: ["auto", "diffusers", "audiocraft", "stable-audio", "tangoflux", "project"] },
+          backend: { type: "string", enum: ["auto", "diffusers", "transformers", "audiocraft", "stable-audio", "tangoflux", "project"] },
           duration: { type: "number", description: "Clip length in seconds" },
           steps: { type: "number", description: "Diffusion sampling steps when supported" },
           seed: { type: "number", description: "Optional random seed" }
@@ -252764,9 +252928,9 @@ if __name__ == "__main__":
           };
         }
         if (action === "setup") {
-          const requested = String(args["backend"] ?? (kind === "music" ? this.defaults.musicBackend : this.defaults.soundBackend) ?? "diffusers");
+          const requested = String(args["backend"] ?? (kind === "music" ? this.defaults.musicBackend : this.defaults.soundBackend) ?? (kind === "music" ? "transformers" : "diffusers"));
           const backend2 = inferAudioGenerationBackend(typeof args["model"] === "string" ? args["model"] : void 0, requested);
-          const resolvedBackend = backend2 === "auto" ? "diffusers" : backend2;
+          const resolvedBackend = backend2 === "auto" ? kind === "music" ? "transformers" : "diffusers" : backend2;
           const plan = audioGenerationSetupPlan(kind, resolvedBackend, this.cwd, typeof args["model"] === "string" ? args["model"] : void 0);
           return {
             success: true,
@@ -252787,7 +252951,7 @@ if __name__ == "__main__":
           const requestedModel2 = rawModel2 === "auto" ? void 0 : rawModel2;
           let backend2 = inferAudioGenerationBackend(requestedModel2, args["backend"] ? String(args["backend"]) : defaultBackend2);
           if (backend2 === "auto")
-            backend2 = kind === "music" ? "audiocraft" : "diffusers";
+            backend2 = kind === "music" ? "transformers" : "diffusers";
           const model2 = requestedModel2 ?? (kind === "music" ? DEFAULT_MUSIC_MODEL : DEFAULT_SOUND_MODEL);
           const preset2 = getAudioGenerationPreset(model2, kind);
           const duration2 = numberArg2(args["duration"], preset2?.defaultDurationSec ?? (kind === "music" ? 20 : 8));
@@ -252810,7 +252974,7 @@ if __name__ == "__main__":
           return await this.prewarmPythonBackend({
             kind,
             backend: backend2,
-            runnerBackend: backend2 === "audiocraft" ? "audiocraft" : backend2 === "stable-audio" ? "stable-audio" : backend2 === "tangoflux" ? "tangoflux" : "diffusers",
+            runnerBackend: backend2,
             model: model2,
             duration: duration2,
             start: start2,
@@ -252827,7 +252991,7 @@ if __name__ == "__main__":
         const requestedModel = rawModel === "auto" ? void 0 : rawModel;
         let backend = inferAudioGenerationBackend(requestedModel, args["backend"] ? String(args["backend"]) : defaultBackend);
         if (backend === "auto")
-          backend = kind === "music" ? "audiocraft" : "diffusers";
+          backend = kind === "music" ? "transformers" : "diffusers";
         const model = requestedModel ?? (kind === "music" ? DEFAULT_MUSIC_MODEL : DEFAULT_SOUND_MODEL);
         const preset = getAudioGenerationPreset(model, kind);
         const duration = numberArg2(args["duration"], preset?.defaultDurationSec ?? (kind === "music" ? 20 : 8));
@@ -252853,6 +253017,9 @@ if __name__ == "__main__":
           if (backend === "tangoflux") {
             return await this.generateWithPythonBackend({ kind, backend, runnerBackend: "tangoflux", prompt, model, duration, steps, seed, start: start2, python: args["python"] });
           }
+          if (backend === "transformers") {
+            return await this.generateWithPythonBackend({ kind, backend, runnerBackend: "transformers", prompt, model, duration, steps, seed, start: start2, python: args["python"] });
+          }
           if (backend === "audiocraft") {
             return await this.generateWithPythonBackend({ kind, backend, runnerBackend: "audiocraft", prompt, model, duration, steps, seed, start: start2, python: args["python"] });
           }
@@ -252902,8 +253069,9 @@ if __name__ == "__main__":
           "--duration",
           String(args.duration)
         ];
-        if (args.steps !== void 0 && args.runnerBackend !== "audiocraft")
+        if (args.steps !== void 0 && (args.runnerBackend === "diffusers" || args.runnerBackend === "stable-audio" || args.runnerBackend === "tangoflux")) {
           argv.push("--steps", String(args.steps));
+        }
         if (args.seed !== void 0)
           argv.push("--seed", String(args.seed));
         this.emitProgress({ stage: "load", message: `Starting ${args.kind} generation with ${args.model}` });
@@ -476611,7 +476779,7 @@ var require_path_browserify = __commonJS({
           return path11.slice(start2, end);
         }
       },
-      extname: function extname14(path11) {
+      extname: function extname15(path11) {
         assertPath(path11);
         var startDot = -1;
         var startPart = 0;
@@ -506772,10 +506940,82 @@ Saved to: ${tempFile}`,
 });
 
 // packages/execution/dist/tools/audio-playback.js
-import { execSync as execSync29, spawn as spawn16 } from "node:child_process";
+import { execFileSync as execFileSync2, execSync as execSync29, spawn as spawn16 } from "node:child_process";
 import { existsSync as existsSync40, statSync as statSync18, writeFileSync as writeFileSync16, readFileSync as readFileSync31, unlinkSync as unlinkSync8, mkdirSync as mkdirSync16 } from "node:fs";
-import { join as join58 } from "node:path";
+import { extname as extname10, join as join58 } from "node:path";
 import { homedir as homedir14, tmpdir as tmpdir11 } from "node:os";
+function hasCommand3(command) {
+  try {
+    if (process.platform === "win32") {
+      execFileSync2("where", [command], { stdio: "ignore", timeout: 2e3 });
+    } else {
+      execFileSync2("command", ["-v", command], { stdio: "ignore", timeout: 2e3 });
+    }
+    return true;
+  } catch {
+    if (process.platform !== "win32") {
+      try {
+        execFileSync2("which", [command], { stdio: "ignore", timeout: 2e3 });
+        return true;
+      } catch {
+        return false;
+      }
+    }
+    return false;
+  }
+}
+function playbackCommandFor(file, device = "default") {
+  const ext = extname10(file).toLowerCase();
+  if (process.platform === "darwin" && hasCommand3("afplay")) {
+    return { command: "afplay", args: [file], label: "afplay" };
+  }
+  if (hasCommand3("ffplay")) {
+    return { command: "ffplay", args: ["-nodisp", "-autoexit", "-loglevel", "error", file], label: "ffplay" };
+  }
+  if (hasCommand3("mpv")) {
+    return { command: "mpv", args: ["--no-video", "--really-quiet", file], label: "mpv" };
+  }
+  if (process.platform === "win32") {
+    return {
+      command: "powershell.exe",
+      args: [
+        "-NoProfile",
+        "-Command",
+        `Add-Type -AssemblyName presentationCore; $p=New-Object System.Windows.Media.MediaPlayer; $p.Open([Uri]::new(${JSON.stringify(file)})); Start-Sleep -Milliseconds 200; while($p.NaturalDuration.HasTimeSpan -eq $false){Start-Sleep -Milliseconds 100}; $p.Play(); Start-Sleep -Milliseconds ([int]$p.NaturalDuration.TimeSpan.TotalMilliseconds + 250)`
+      ],
+      label: "powershell MediaPlayer"
+    };
+  }
+  if ((ext === ".wav" || ext === ".flac" || ext === ".oga" || ext === ".ogg") && hasCommand3("pw-play")) {
+    return { command: "pw-play", args: [file], label: "pw-play" };
+  }
+  if ((ext === ".wav" || ext === ".flac" || ext === ".oga" || ext === ".ogg") && hasCommand3("paplay")) {
+    return { command: "paplay", args: [file], label: "paplay" };
+  }
+  if (ext === ".wav" && hasCommand3("aplay")) {
+    return { command: "aplay", args: ["-D", device, "-q", file], label: "aplay" };
+  }
+  return null;
+}
+function playSoundFile(file, opts = {}) {
+  if (!file)
+    return { ok: false, error: "Missing file path." };
+  if (!existsSync40(file))
+    return { ok: false, error: `Audio file not found: ${file}` };
+  const command = playbackCommandFor(file, opts.device || "default");
+  if (!command) {
+    return {
+      ok: false,
+      error: "No terminal audio player found. Install ffmpeg/ffplay or mpv; WAV fallback can use pipewire (pw-play), PulseAudio (paplay), or ALSA (aplay)."
+    };
+  }
+  try {
+    execFileSync2(command.command, command.args, { timeout: opts.timeoutMs ?? 3e5, stdio: "pipe" });
+    return { ok: true, player: command.label };
+  } catch (err) {
+    return { ok: false, error: `Playback via ${command.label} failed: ${err instanceof Error ? err.message.slice(0, 300) : String(err).slice(0, 300)}` };
+  }
+}
 function ensureLuxttsDaemon() {
   if (_luxttsDaemon && !_luxttsDaemon.killed && _luxttsReady)
     return Promise.resolve(true);
@@ -506886,7 +507126,7 @@ function luxttsSynthesize(text, cloneRef) {
     _luxttsDaemon.stdin.write(req2 + "\n");
   });
 }
-var _luxttsDaemon, _luxttsReady, _luxttsRequestId, _luxttsPending, _luxttsBuffer, _luxttsStarting, AudioPlaybackTool;
+var _luxttsDaemon, _luxttsReady, _luxttsRequestId, _luxttsPending, _luxttsBuffer, _luxttsStarting, AudioPlaybackTool, SoundPlaybackTool;
 var init_audio_playback = __esm({
   "packages/execution/dist/tools/audio-playback.js"() {
     "use strict";
@@ -506965,21 +507205,13 @@ var init_audio_playback = __esm({
         const device = args["device"] || "default";
         const size = statSync18(file).size;
         const ext = file.split(".").pop()?.toLowerCase() || "";
-        let cmd;
-        if (ext === "wav") {
-          cmd = `aplay -D ${device} -q "${file}"`;
-        } else {
-          cmd = `ffplay -nodisp -autoexit -loglevel error "${file}"`;
-        }
-        try {
-          execSync29(cmd, { timeout: 3e5, stdio: "pipe" });
-        } catch (err) {
-          const msg = err instanceof Error ? err.message : String(err);
-          return { success: false, output: "", error: `Playback failed: ${msg.slice(0, 300)}`, durationMs: performance.now() - start2 };
+        const played = playSoundFile(file, { device, timeoutMs: 3e5 });
+        if (!played.ok) {
+          return { success: false, output: "", error: played.error, durationMs: performance.now() - start2 };
         }
         return {
           success: true,
-          output: `Played ${file} (${Math.round(size / 1024)}KB ${ext.toUpperCase()}) on ${device}`,
+          output: `Played ${file} (${Math.round(size / 1024)}KB ${ext.toUpperCase()}) via ${played.player}`,
           durationMs: performance.now() - start2
         };
       }
@@ -507146,6 +507378,51 @@ ${devices.join("\n")}`,
         };
       }
     };
+    SoundPlaybackTool = class {
+      name = "play_sound";
+      description = "Play any local sound/audio file through the system speakers from the terminal. Supports generated WAV files and common audio formats such as MP3, OGG, OPUS, M4A, FLAC, and AAC when ffplay/mpv or system audio players are available. Use this immediately after generate_audio, /sound, or /music when the user wants to hear the generated file.";
+      parameters = {
+        type: "object",
+        properties: {
+          file: {
+            type: "string",
+            description: "Path to the local audio file to play"
+          },
+          device: {
+            type: "string",
+            description: "Optional output device. ALSA aplay accepts values like default or hw:0,0; other players ignore this."
+          },
+          timeout_ms: {
+            type: "number",
+            description: "Optional playback timeout in milliseconds; default 300000."
+          }
+        },
+        required: ["file"]
+      };
+      async execute(args) {
+        const start2 = performance.now();
+        const file = typeof args["file"] === "string" ? args["file"] : "";
+        const device = typeof args["device"] === "string" ? args["device"] : "default";
+        const timeoutMs = typeof args["timeout_ms"] === "number" && Number.isFinite(args["timeout_ms"]) ? Math.max(1e3, Math.min(36e5, Math.round(args["timeout_ms"]))) : 3e5;
+        if (!file) {
+          return { success: false, output: "", error: "Missing 'file' parameter. Provide a local sound/audio file path.", durationMs: performance.now() - start2 };
+        }
+        if (!existsSync40(file)) {
+          return { success: false, output: "", error: `Audio file not found: ${file}`, durationMs: performance.now() - start2 };
+        }
+        const size = statSync18(file).size;
+        const ext = extname10(file).replace(/^\./, "").toUpperCase() || "audio";
+        const played = playSoundFile(file, { device, timeoutMs });
+        if (!played.ok) {
+          return { success: false, output: "", error: played.error, durationMs: performance.now() - start2 };
+        }
+        return {
+          success: true,
+          output: `Played sound: ${file} (${Math.round(size / 1024)}KB ${ext}) via ${played.player}`,
+          durationMs: performance.now() - start2
+        };
+      }
+    };
   }
 });
 
@@ -514787,6 +515064,7 @@ __export(dist_exports, {
   SkillBuildTool: () => SkillBuildTool,
   SkillExecuteTool: () => SkillExecuteTool,
   SkillListTool: () => SkillListTool,
+  SoundPlaybackTool: () => SoundPlaybackTool,
   StdioTransport: () => StdioTransport,
   StructuredFileTool: () => StructuredFileTool,
   StructuredReadTool: () => StructuredReadTool,
@@ -514906,6 +515184,7 @@ __export(dist_exports, {
   packetPath: () => packetPath,
   parseMcpMarkdown: () => parseMcpMarkdown,
   parseMcpToolName: () => parseMcpToolName,
+  playSoundFile: () => playSoundFile,
   promoteWorkingNotes: () => promoteWorkingNotes,
   readPacket: () => readPacket,
   readProvenanceFile: () => readProvenanceFile,
@@ -530271,6 +530550,8 @@ var init_agenticRunner = __esm({
       "transcribe_file",
       "transcribe_url",
       "audio_playback",
+      "play_sound",
+      "generate_audio",
       "youtube_download"
     ]);
     SOCIAL_TOOLS = /* @__PURE__ */ new Set([
@@ -552084,7 +552365,7 @@ var init_command_registry = __esm({
       ["/music", "Open music-generation model/setup menu"],
       ["/music <prompt>", "Generate a music clip from a prompt"],
       ["/music --model <model> <prompt>", "Generate music with an explicit music model"],
-      ["/music setup <audiocraft|stable-audio|diffusers>", "Show setup commands for a music-generation backend"],
+      ["/music setup <transformers|audiocraft|stable-audio|diffusers>", "Show setup commands for a music-generation backend"],
       ["/music list", "List music models by category, quality, size, and hardware fit"],
       ["/call", "Start voice call session (cloudflared tunnel + ASR/TTS)"],
       ["/hangup", "End active call session"],
@@ -569380,7 +569661,7 @@ var init_platforms = __esm({
 
 // packages/cli/src/tui/workspace-explorer.ts
 import { existsSync as existsSync88, readdirSync as readdirSync26, readFileSync as readFileSync72, statSync as statSync31 } from "node:fs";
-import { basename as basename15, extname as extname11, join as join104, relative as relative10, resolve as resolve35 } from "node:path";
+import { basename as basename15, extname as extname12, join as join104, relative as relative10, resolve as resolve35 } from "node:path";
 function exploreWorkspace(root, options2 = {}) {
   const query = (options2.query ?? "").trim().toLowerCase();
   const maxResults = options2.maxResults ?? 80;
@@ -569493,7 +569774,7 @@ function previewWorkspaceFile(root, relPath, options2 = {}) {
 }
 function classifyWorkspaceFile(path11) {
   const lower = path11.toLowerCase();
-  const ext = extname11(lower);
+  const ext = extname12(lower);
   if (lower.includes(".test.") || lower.includes(".spec.") || lower.includes("/tests/")) return "test";
   if (SOURCE_EXT.has(ext)) return "source";
   if (DOC_EXT2.has(ext)) return "doc";
@@ -569569,7 +569850,7 @@ var init_workspace_explorer = __esm({
 
 // packages/cli/src/tui/drop-panel.ts
 import { existsSync as existsSync89 } from "node:fs";
-import { extname as extname12, resolve as resolve36 } from "node:path";
+import { extname as extname13, resolve as resolve36 } from "node:path";
 function ansi4(code8, text) {
   return isTTY4 ? `\x1B[${code8}m${text}\x1B[0m` : text;
 }
@@ -569695,7 +569976,7 @@ function showDropPanel(opts) {
         return;
       }
       if (allowedExtensions.length > 0) {
-        const ext = extname12(filePath).toLowerCase();
+        const ext = extname13(filePath).toLowerCase();
         if (!allowedExtensions.includes(ext)) {
           errorMsg = `Invalid file type: ${ext}. Expected: ${allowedExtensions.join(", ")}`;
           render2();
@@ -574002,7 +574283,7 @@ __export(image_ascii_preview_exports, {
   extractSavedImagePath: () => extractSavedImagePath,
   formatImageAsciiContext: () => formatImageAsciiContext
 });
-import { execFileSync as execFileSync2 } from "node:child_process";
+import { execFileSync as execFileSync3 } from "node:child_process";
 import { createRequire as createRequire4 } from "node:module";
 import { existsSync as existsSync94, readFileSync as readFileSync75, statSync as statSync32 } from "node:fs";
 import { resolve as resolve37 } from "node:path";
@@ -574139,7 +574420,7 @@ function convertWithFfmpeg(imagePath, width, height, timeoutMs) {
       `scale=${width}:${height}`,
       "format=gray"
     ].join(",");
-    const raw = execFileSync2(
+    const raw = execFileSync3(
       "ffmpeg",
       [
         "-hide_banner",
@@ -584829,10 +585110,10 @@ function defaultAudioModel(kind) {
   return kind === "music" ? DEFAULT_MUSIC_MODEL : DEFAULT_SOUND_MODEL;
 }
 function defaultAudioBackend(kind) {
-  return kind === "music" ? "audiocraft" : "diffusers";
+  return kind === "music" ? "transformers" : "diffusers";
 }
 function normalizeAudioBackend(value2, kind) {
-  if (value2 === "diffusers" || value2 === "audiocraft" || value2 === "stable-audio" || value2 === "tangoflux" || value2 === "project") return value2;
+  if (value2 === "diffusers" || value2 === "transformers" || value2 === "audiocraft" || value2 === "stable-audio" || value2 === "tangoflux" || value2 === "project") return value2;
   return defaultAudioBackend(kind);
 }
 function audioBackendForModel(model, requested, kind) {
@@ -584920,7 +585201,8 @@ async function showAudioGenerationMenu(ctx3, hasLocal, kind) {
   const activeModel = activeAudioModel(settings, kind);
   const title = kind === "music" ? "Music Generation" : "Sound Generation";
   const setupItems = kind === "music" ? [
-    { key: "setup:audiocraft", label: "Setup AudioCraft", detail: "MusicGen small/medium/large" },
+    { key: "setup:transformers", label: "Setup Transformers", detail: "Default MusicGen path; avoids AudioCraft/PyAV headers" },
+    { key: "setup:audiocraft", label: "Setup AudioCraft", detail: "Optional MusicGen/AudioGen upstream runtime" },
     { key: "setup:stable-audio", label: "Setup Stable Audio", detail: "Stable Audio Open 1.0 serious stereo baseline" },
     { key: "setup:diffusers", label: "Setup Diffusers", detail: "AudioLDM/Riffusion-style paths" }
   ] : [
@@ -585050,7 +585332,17 @@ async function handleAudioGenerationCommand(ctx3, arg, hasLocal, kind) {
   }
   renderInfo(result.output);
   const fileMatch = result.output.match(/(?:Sound|Music) generated:\s+(.+)/);
-  if (fileMatch?.[1]) renderInfo(`File: ${fileMatch[1].trim()}`);
+  if (fileMatch?.[1]) {
+    const audioPath = fileMatch[1].trim();
+    renderInfo(`File: ${audioPath}`);
+    renderInfo(`Playing generated ${kind}...`);
+    const playback = playSoundFile(audioPath);
+    if (playback.ok) {
+      renderInfo(`Audio playback complete via ${playback.player}.`);
+    } else {
+      renderWarning(`Audio playback failed: ${playback.error}`);
+    }
+  }
   return "handled";
 }
 function formatAudioGenerationProgress(event) {
@@ -595599,12 +595891,12 @@ __export(vision_ingress_exports, {
   queryVisionModel: () => queryVisionModel,
   runVisionIngress: () => runVisionIngress
 });
-import { execFileSync as execFileSync3 } from "node:child_process";
+import { execFileSync as execFileSync4 } from "node:child_process";
 import { existsSync as existsSync105, readFileSync as readFileSync86, unlinkSync as unlinkSync20 } from "node:fs";
 import { join as join120 } from "node:path";
 function isTesseractAvailable() {
   try {
-    execFileSync3("tesseract", ["--version"], { stdio: "ignore", timeout: 3e3 });
+    execFileSync4("tesseract", ["--version"], { stdio: "ignore", timeout: 3e3 });
     return true;
   } catch {
     return false;
@@ -595645,7 +595937,7 @@ function advancedOcr(imagePath) {
   for (const psm of psmModes) {
     const outFile = `${tmpBase}_psm${psm}`;
     try {
-      execFileSync3("tesseract", [
+      execFileSync4("tesseract", [
         imagePath,
         outFile,
         "--psm",
@@ -623381,7 +623673,7 @@ var clipboard_media_exports = {};
 __export(clipboard_media_exports, {
   pasteClipboardImageToFile: () => pasteClipboardImageToFile
 });
-import { execFileSync as execFileSync4, execSync as execSync58 } from "node:child_process";
+import { execFileSync as execFileSync5, execSync as execSync58 } from "node:child_process";
 import { mkdirSync as mkdirSync72, readFileSync as readFileSync99, rmSync as rmSync5, writeFileSync as writeFileSync67 } from "node:fs";
 import { join as join136 } from "node:path";
 function pasteClipboardImageToFile(repoRoot) {
@@ -623398,7 +623690,7 @@ function readClipboardImage() {
     try {
       execSync58("command -v pngpaste", { stdio: "ignore", timeout: 1e3 });
       const tmp = `/tmp/omnius-clipboard-${Date.now()}.png`;
-      execFileSync4("pngpaste", [tmp], { timeout: 3e3 });
+      execFileSync5("pngpaste", [tmp], { timeout: 3e3 });
       const buffer2 = readFileSync99(tmp);
       try {
         rmSync5(tmp);
@@ -623418,7 +623710,7 @@ function readClipboardImage() {
     ];
     for (const attempt of attempts) {
       try {
-        const buffer2 = execFileSync4(attempt.cmd, attempt.args, { timeout: 3e3, maxBuffer: 25 * 1024 * 1024 });
+        const buffer2 = execFileSync5(attempt.cmd, attempt.args, { timeout: 3e3, maxBuffer: 25 * 1024 * 1024 });
         if (buffer2.length > 0) return { buffer: buffer2, mime: attempt.mime, ext: attempt.ext };
       } catch {
         continue;
@@ -623435,7 +623727,7 @@ function readClipboardImage() {
         "$img.Save($ms,[Drawing.Imaging.ImageFormat]::Png);",
         "[Console]::OpenStandardOutput().Write($ms.ToArray(),0,$ms.Length)"
       ].join("");
-      const buffer2 = execFileSync4("powershell.exe", ["-NoProfile", "-Command", ps], {
+      const buffer2 = execFileSync5("powershell.exe", ["-NoProfile", "-Command", ps], {
         timeout: 5e3,
         maxBuffer: 25 * 1024 * 1024
       });
@@ -623454,7 +623746,7 @@ var init_clipboard_media = __esm({
 
 // packages/cli/src/tui/interactive.ts
 import { cwd } from "node:process";
-import { resolve as resolve44, join as join137, dirname as dirname38, extname as extname13, relative as relative14 } from "node:path";
+import { resolve as resolve44, join as join137, dirname as dirname38, extname as extname14, relative as relative14 } from "node:path";
 import { createRequire as createRequire7 } from "node:module";
 import { fileURLToPath as fileURLToPath18 } from "node:url";
 import {
@@ -623839,6 +624131,7 @@ function buildSubAgentTools(repoRoot, config) {
     new StructuredFileTool(repoRoot),
     // Audio
     new AudioPlaybackTool(),
+    new SoundPlaybackTool(),
     new AudioCaptureTool(),
     new AudioAnalyzeTool(),
     new AsrListenTool(),
@@ -623993,6 +624286,7 @@ function buildTools(repoRoot, config, contextWindowSize, modelTier) {
     new CameraCaptureTool(),
     new AudioCaptureTool(),
     new AudioPlaybackTool(),
+    new SoundPlaybackTool(),
     new WifiControlTool(),
     new BluetoothScanTool(),
     new SdrScanTool(),
@@ -624542,6 +624836,24 @@ async function renderAsciiPreviewForToolResult(toolName, output, repoRoot, write
     }
   }
 }
+function extractGeneratedAudioPath(output, repoRoot) {
+  const match = output.match(/(?:Sound|Music) generated:\s+([^\n\r]+)/i);
+  const raw = match?.[1]?.trim().replace(/^["']|["']$/g, "");
+  if (!raw) return null;
+  return raw.startsWith("/") || raw.startsWith("~") ? raw.replace(/^~(?=\/)/, homedir46()) : join137(repoRoot, raw);
+}
+async function playGeneratedAudioForToolResult(toolName, output, repoRoot, writer) {
+  if (toolName !== "generate_audio" || !output) return;
+  const audioPath = extractGeneratedAudioPath(output, repoRoot);
+  if (!audioPath) return;
+  writer(() => renderInfo(`Playing generated audio: ${relative14(repoRoot, audioPath).startsWith("..") ? audioPath : relative14(repoRoot, audioPath)}`));
+  const result = playSoundFile(audioPath);
+  if (result.ok) {
+    writer(() => renderInfo(`Audio playback complete via ${result.player}.`));
+  } else {
+    writer(() => renderWarning(`Audio playback failed: ${result.error}`));
+  }
+}
 async function runSelfImprovementCycle(repoRoot) {
   try {
     const {
@@ -625697,8 +626009,9 @@ ${entry.fullContent}`
             }
           });
         }
-        if (event.success || event.toolName === "generate_image") {
+        if (event.success) {
           void renderAsciiPreviewForToolResult(event.toolName, event.content ?? "", repoRoot, contentWrite);
+          void playGeneratedAudioForToolResult(event.toolName, event.content ?? "", repoRoot, contentWrite);
         }
         if (voice?.enabled && voice.voiceMode === "voicechat" && _voiceChatSession2?.isActive && event.toolName === "task_complete") {
           const emoStateFinal = emotionEngine?.getState();
@@ -630719,7 +631032,7 @@ Execute this skill now. Follow the behavioral guidance above.`;
           const imgPath = resolve44(repoRoot, cleanPath);
           const imgBuffer = readFileSync100(imgPath);
           const base642 = imgBuffer.toString("base64");
-          const ext = extname13(cleanPath).toLowerCase();
+          const ext = extname14(cleanPath).toLowerCase();
           const mime = ext === ".png" ? "image/png" : ext === ".gif" ? "image/gif" : ext === ".webp" ? "image/webp" : "image/jpeg";
           const asciiContext = await renderAsciiPreviewForImage(
             imgPath,

package/npm-shrinkwrap.json

@@ -1,12 +1,12 @@
 {
   "name": "omnius",
-  "version": "1.0.14",
+  "version": "1.0.15",
   "lockfileVersion": 3,
   "requires": true,
   "packages": {
     "": {
       "name": "omnius",
-      "version": "1.0.14",
+      "version": "1.0.15",
       "bundleDependencies": [
         "image-to-ascii"
       ],

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "omnius",
-  "version": "1.0.14",
+  "version": "1.0.15",
   "description": "AI coding agent powered by open-source models (Ollama/vLLM) — interactive TUI with agentic tool-calling loop",
   "type": "module",
   "main": "./dist/index.js",