@mevdragon/vidfarm-devcli 0.19.1 → 0.20.1

package/dist/src/primitive-registry.js

@@ -429,7 +429,11 @@ const ttsPayloadSchema = z.preprocess((raw) => {
     text: z.string().min(1).max(8000),
     voice: z.string().min(1).max(120).optional(),
     instructions: z.string().min(1).max(4000).optional(),
-    output_format: z.enum(["mp3", "wav"]).default("mp3")
+    output_format: z.enum(["mp3", "wav"]).default("mp3"),
+    // Default TRUE: great ElevenLabs narration out of the box on the platform key,
+    // billed to the wallet. Set false to use your own ElevenLabs key or a BYOK
+    // openai/gemini/openrouter key instead. See voice options via /audio/voices.
+    use_wallet_credits: z.boolean().default(true)
 }).superRefine((value, ctx) => {
     // Reject an explicit provider/voice contradiction at enqueue (400) instead
     // of queueing a job that can only fail at the provider.
@@ -465,7 +469,36 @@ const sttPayloadSchema = z.preprocess((raw) => {
     // output and downstream callers estimate word windows.
     word_timestamps: z.boolean().default(false),
     language: z.string().min(2).max(16).optional(),
-    prompt: z.string().min(1).max(2000).optional()
+    prompt: z.string().min(1).max(2000).optional(),
+    // Default TRUE: platform ElevenLabs Scribe (native diarization + real word
+    // timestamps), billed to the wallet. Set false to use your own ElevenLabs
+    // key or a BYOK gemini/openai/openrouter key.
+    use_wallet_credits: z.boolean().default(true)
+}));
+// Music generation (ElevenLabs). Prompt-based, or a hand-built composition plan.
+const musicPayloadSchema = z.preprocess((raw) => {
+    if (!raw || typeof raw !== "object" || Array.isArray(raw))
+        return raw;
+    const payload = raw;
+    return {
+        ...payload,
+        prompt: payload.prompt ?? payload.text ?? payload.description,
+        music_length_ms: payload.music_length_ms ?? payload.length_ms ?? payload.duration_ms
+    };
+}, z.object({
+    prompt: z.string().min(1).max(2000).optional(),
+    // ElevenLabs music caps at 5 minutes per generation.
+    music_length_ms: z.number().int().min(3000).max(300000).default(30000),
+    // Optional ElevenLabs composition plan for granular section-by-section control.
+    composition_plan: z.record(z.string(), z.any()).optional(),
+    model: z.string().min(1).optional(),
+    output_format: z.string().min(1).optional(),
+    // Default TRUE: platform ElevenLabs key, billed to the wallet. Set false to
+    // use your own saved ElevenLabs key (no wallet charge).
+    use_wallet_credits: z.boolean().default(true)
+}).refine((value) => Boolean(value.prompt) || Boolean(value.composition_plan), {
+    message: "Provide either a prompt or a composition_plan.",
+    path: ["prompt"]
 }));
 // Animated captions from speech: STT with word-level timings, paged into
 // ready-to-apply cue layers. Either transcribe a source_url or page plain text.
@@ -714,6 +747,7 @@ const PRIMITIVE_BRAINSTORM_PRODUCT_PLACEMENT_ID = "primitive:brainstorm_product_
 const PRIMITIVE_MEDIA_DEDUPE_ID = "primitive:media_dedupe";
 const PRIMITIVE_TTS_ID = "primitive:tts";
 const PRIMITIVE_STT_ID = "primitive:stt";
+const PRIMITIVE_MUSIC_ID = "primitive:music";
 const PRIMITIVE_CAPTIONS_ID = "primitive:captions";
 const PRIMITIVE_SPEECH_REGENERATE_ID = "primitive:speech_regenerate";
 // Gemini inline audio caps out around 20MB; at the 64kbps mono mp3 we extract,
@@ -2018,20 +2052,34 @@ const ttsPrimitive = definePrimitive({
                 text: payload.text,
                 voice: payload.voice,
                 instructions: payload.instructions,
-                responseFormat: payload.output_format
-            });
-            // Gemini TTS answers in wav regardless of the requested format; name the
-            // artifact after what actually came back.
+                responseFormat: payload.output_format,
+                useWalletCredits: payload.use_wallet_credits
+            });
+            // Route may have resolved to a different provider (e.g. platform ElevenLabs
+            // when use_wallet_credits is on); report what actually ran.
+            const effectiveProvider = generated.effectiveProvider;
+            const effectiveModel = generated.effectiveModel;
+            if (generated.billing) {
+                await ctx.billing.record({
+                    type: "ai_generation",
+                    costUsd: generated.billing.costUsd,
+                    costCenterSlug: generated.billing.costCenterSlug,
+                    metadata: { ...generated.billing.metadata, primitive: "tts", voice: payload.voice ?? null }
+                });
+            }
+            // Gemini/ElevenLabs TTS may answer in wav regardless of the requested
+            // format; name the artifact after what actually came back.
             const extension = generated.contentType.includes("wav") ? "wav" : "mp3";
             const stored = await ctx.storage.putBuffer(`speech.${extension}`, generated.bytes, {
                 contentType: generated.contentType,
                 kind: "audio",
                 metadata: {
                     primitive: "tts",
-                    provider,
-                    model,
+                    provider: effectiveProvider,
+                    model: effectiveModel,
                     voice: payload.voice ?? null,
-                    styleInstructions: payload.instructions ?? null
+                    styleInstructions: payload.instructions ?? null,
+                    wallet_billed: Boolean(generated.billing)
                 }
             });
             ctx.logger.progress(1, "Speech generation complete");
@@ -2044,11 +2092,72 @@ const ttsPrimitive = definePrimitive({
                     audio: {
                         file_url: stored.url,
                         content_type: generated.contentType,
-                        provider,
-                        model,
+                        provider: effectiveProvider,
+                        model: effectiveModel,
                         voice: payload.voice ?? null,
                         style_instructions: payload.instructions ?? null,
-                        text_length: payload.text.length
+                        text_length: payload.text.length,
+                        wallet_billed: Boolean(generated.billing)
+                    }
+                }
+            };
+        }
+    }
+});
+const musicPrimitive = definePrimitive({
+    id: PRIMITIVE_MUSIC_ID,
+    kind: "music",
+    operations: {
+        run: {
+            description: "Generate music (instrumental or with lyrics) from a prompt or a composition plan via ElevenLabs, persisted as a reusable platform audio URL.",
+            inputSchema: musicPayloadSchema,
+            workflow: "run"
+        }
+    },
+    jobs: {
+        async run(ctx, input) {
+            const payload = musicPayloadSchema.parse(input);
+            ctx.logger.progress(0.12, "Composing music");
+            const generated = await ctx.providers.generateMusic({
+                prompt: payload.prompt,
+                musicLengthMs: payload.music_length_ms,
+                compositionPlan: payload.composition_plan,
+                model: payload.model,
+                outputFormat: payload.output_format,
+                useWalletCredits: payload.use_wallet_credits
+            });
+            if (generated.billing) {
+                await ctx.billing.record({
+                    type: "ai_generation",
+                    costUsd: generated.billing.costUsd,
+                    costCenterSlug: generated.billing.costCenterSlug,
+                    metadata: { ...generated.billing.metadata, primitive: "music", prompt: payload.prompt ?? null }
+                });
+            }
+            const stored = await ctx.storage.putBuffer("music.mp3", generated.bytes, {
+                contentType: generated.contentType,
+                kind: "audio",
+                metadata: {
+                    primitive: "music",
+                    provider: "elevenlabs",
+                    music_length_ms: generated.musicLengthMs,
+                    wallet_billed: Boolean(generated.billing)
+                }
+            });
+            ctx.logger.progress(1, "Music generation complete");
+            return {
+                progress: 1,
+                output: {
+                    files: compactUrls([stored.url]),
+                    primary_file_url: stored.url,
+                    audioUrl: stored.url,
+                    audio: {
+                        file_url: stored.url,
+                        content_type: generated.contentType,
+                        provider: "elevenlabs",
+                        music_length_ms: generated.musicLengthMs,
+                        prompt: payload.prompt ?? null,
+                        wallet_billed: Boolean(generated.billing)
                     }
                 }
             };
@@ -2122,18 +2231,29 @@ const sttPrimitive = definePrimitive({
             const payload = sttPayloadSchema.parse(input);
             const { audioBytes, audioContentType, extractionMetadata } = await loadSpeechAudioForPrimitive(ctx, payload);
             ctx.logger.progress(0.45, "Transcribing speech");
-            const provider = payload.provider ?? (payload.word_timestamps ? "openai" : "gemini");
-            const model = payload.model ?? defaultSpeechModelFor("stt", provider) ?? "";
+            const requestedProvider = payload.provider ?? (payload.word_timestamps ? "openai" : "gemini");
+            const requestedModel = payload.model ?? defaultSpeechModelFor("stt", requestedProvider) ?? "";
             const transcription = await ctx.providers.transcribeSpeech({
-                provider,
-                model,
+                provider: requestedProvider,
+                model: requestedModel,
                 audio: audioBytes,
                 contentType: audioContentType,
                 prompt: payload.prompt,
                 language: payload.language,
                 diarize: payload.diarize,
-                wordTimestamps: payload.word_timestamps
-            });
+                wordTimestamps: payload.word_timestamps,
+                useWalletCredits: payload.use_wallet_credits
+            });
+            const provider = transcription.effectiveProvider;
+            const model = transcription.effectiveModel;
+            if (transcription.billing) {
+                await ctx.billing.record({
+                    type: "ai_generation",
+                    costUsd: transcription.billing.costUsd,
+                    costCenterSlug: transcription.billing.costCenterSlug,
+                    metadata: { ...transcription.billing.metadata, primitive: "stt", source_url: payload.source_url }
+                });
+            }
             ctx.logger.progress(0.85, "Storing transcript artifacts");
             const speakers = [...new Set(transcription.segments.map((segment) => segment.speaker))];
             const srt = buildSrtFromSegments(transcription.segments);
@@ -3114,6 +3234,7 @@ class PrimitiveRegistry {
         [audioNormalizePrimitive.id, audioNormalizePrimitive],
         [ttsPrimitive.id, ttsPrimitive],
         [sttPrimitive.id, sttPrimitive],
+        [musicPrimitive.id, musicPrimitive],
         [captionsPrimitive.id, captionsPrimitive],
         [speechRegeneratePrimitive.id, speechRegeneratePrimitive],
         [hyperframesRenderPrimitive.id, hyperframesRenderPrimitive],
@@ -3521,6 +3642,7 @@ function normalizeProductPlacementPayload(raw) {
     };
 }
 function preferredTextProvider(availableProviders) {
+    // Text-only priority — the audio-only elevenlabs provider is never a candidate.
     const priority = ["openai", "gemini", "openrouter", "perplexity"];
     return priority.find((provider) => availableProviders.includes(provider)) ?? "openai";
 }

package/dist/src/reskin/chat-page.js

@@ -59,8 +59,8 @@ export function renderReskinChat(input = DEFAULT_INPUT, opts = {}) {
             <select class="rk-chat-mode-select" id="rk-chat-mode">
               <option value="/chat" selected>Chat</option>
               <option value="/tools/image">Create Image</option>
-              <option value="/tools/clipper">Clip a Video</option>
               <option value="/editor/original/fork/new">Create Video</option>
+              <option value="/tools/clipper">Clip Media</option>
             </select>
             <span class="rk-chat-mode-chev" aria-hidden="true">&#9662;</span>
           </label>