@drakulavich/kesha-voice-kit 1.4.3 → 1.5.0

package/NOTICES.md

@@ -27,19 +27,13 @@ Pinned to the `v6.2.1` tag.
 
 Text-to-speech (English). Apache 2.0. ONNX export by [onnx-community/Kokoro-82M-v1.0-ONNX](https://huggingface.co/onnx-community/Kokoro-82M-v1.0-ONNX).
 
-### Piper voices (Russian: `ru-denis`)
+### Vosk-TTS (Russian, multi-speaker)
 
-Text-to-speech (multilingual, Piper VITS). MIT. Voice packs from [rhasspy/piper-voices](https://huggingface.co/rhasspy/piper-voices).
+Text-to-speech (Russian, 5 baked-in speakers). Apache 2.0. Model: [alphacep/vosk-tts](https://huggingface.co/alphacep/vosk-tts) (`vosk-model-tts-ru-0.9-multi`). Integrated via [`vosk-tts-rs`](https://github.com/alphacep/vosk-tts-rs).
 
-### CharsiuG2P ByT5-tiny (ONNX export)
+### misaki-rs (English G2P)
 
-Grapheme-to-phoneme conversion for Kokoro and Piper pipelines (#123).
-
-**License: CC-BY 4.0** (declared on the ONNX-export repo). Attribution required.
-
-- ONNX export: [klebster/g2p_multilingual_byT5_tiny_onnx](https://huggingface.co/klebster/g2p_multilingual_byT5_tiny_onnx) by Kleber Noel — CC-BY 4.0.
-- Base checkpoint: [charsiu/g2p_multilingual_byT5_tiny_16_layers_100](https://huggingface.co/charsiu/g2p_multilingual_byT5_tiny_16_layers_100) — **no explicit license declared on HuggingFace**; usage here follows the CC-BY 4.0 declared by the downstream export and the paper's terms.
-- Paper: Zhu, J., Zhang, C., & Jurgens, D. (2022). "ByT5 model for massively multilingual grapheme-to-phoneme conversion." [arXiv:2204.03067](https://arxiv.org/abs/2204.03067) · Interspeech 2022.
+Embedded grapheme-to-phoneme for English (Kokoro pipeline, #207). MIT. Source: [`misaki-rs`](https://github.com/MicheleYin/misaki-rs).
 
 ## System libraries linked by the engine binary
 

package/README.md

@@ -14,7 +14,7 @@
 <p align="center"><b>Open-source voice toolkit.</b> Optimized for Apple Silicon (CoreML), works on any platform (ONNX fallback).<br>A collection of small, fast, open-source audio models — packaged as CLI tools and an <a href="https://github.com/openclaw/openclaw">OpenClaw</a> skill for LLM agents.</p>
 
 - **Speech-to-text** — 25 languages, ~15x faster than Whisper on Apple Silicon, ~2.5x on CPU
-- **Text-to-speech** — Kokoro (EN) + Piper (RU) + macOS system voices, SSML preview
+- **Text-to-speech** — Kokoro (EN) + Vosk-TTS (RU) + macOS system voices, SSML preview
 - **Rust engine** — single 20MB binary, no ffmpeg, no Python, no native Node addons
 - **OpenClaw-ready** — plug into your LLM agent as a voice processing skill
 
@@ -61,12 +61,12 @@ For long / silence-heavy audio, use `--vad` (auto-on past 120 s). Details: [docs
 
 ## Text-to-speech
 
-Kesha speaks back via Kokoro-82M (English) and Piper (Russian) — voice auto-picks from the text's language:
+Kesha speaks back via Kokoro-82M (English) and Vosk-TTS (Russian) — voice auto-picks from the text's language:
 
 ```bash
-kesha install --tts                      # ~490MB (Kokoro + Piper RU + ONNX G2P, opt-in)
+kesha install --tts                      # ~990MB (Kokoro + Vosk-TTS RU, opt-in)
 kesha say "Hello, world" > hello.wav
-kesha say "Привет, мир" > privet.wav     # auto-routes to ru-denis
+kesha say "Привет, мир" > privet.wav     # auto-routes (Milena on darwin, ru-vosk-m02 elsewhere)
 ```
 
 macOS system voices, SSML, voice listing, and the full voice catalogue: [docs/tts.md](docs/tts.md).
@@ -89,7 +89,7 @@ See [BENCHMARK.md](BENCHMARK.md) for the full per-file breakdown (Russian + Engl
 | SpeechBrain ECAPA-TDNN | Audio language detection | ~86MB | [HuggingFace](https://huggingface.co/speechbrain/lang-id-voxlingua107-ecapa) |
 | Apple NLLanguageRecognizer | Text language detection | built-in | macOS system framework |
 | Silero VAD v5 (opt-in) | Voice activity detection | ~2.3MB | [snakers4/silero-vad](https://github.com/snakers4/silero-vad) |
-| Kokoro-82M / Piper (opt-in) | Text-to-speech | ~490MB | [Kokoro](https://huggingface.co/hexgrad/Kokoro-82M) · [Piper](https://github.com/rhasspy/piper) |
+| Kokoro-82M / Vosk-TTS (opt-in) | Text-to-speech | ~990MB | [Kokoro](https://huggingface.co/hexgrad/Kokoro-82M) · [Vosk-TTS](https://github.com/alphacep/vosk-tts) |
 
 All models run through `kesha-engine` — a Rust binary using [FluidAudio](https://github.com/FluidInference/FluidAudio) (CoreML) on Apple Silicon and [ort](https://github.com/pykeio/ort) (ONNX Runtime) on other platforms.
 
@@ -117,7 +117,8 @@ const text = await transcribe("audio.ogg");  // transcribe
 ## Requirements
 
 - [Bun](https://bun.sh) >= 1.3
-- macOS arm64, Linux x64, or Windows x64
+- macOS arm64 or Linux x64
+- Windows x64 temporarily unsupported in v1.5.0 — Vosk-TTS native deps trigger MSVC link conflicts. Tracked in [#216](https://github.com/drakulavich/kesha-voice-kit/issues/216).
 
 ## Contributing
 

package/SKILL.md

@@ -1,15 +1,14 @@
 ---
 name: kesha-voice-kit
-description: Local multilingual voice toolkit — speech-to-text (STT), text-to-speech (TTS), and language detection. Runs entirely offline on Apple Silicon, Linux, and Windows. No API keys, no cloud. NVIDIA Parakeet TDT for STT across 25 European languages, Kokoro-82M + Piper VITS for TTS, plus macOS AVSpeechSynthesizer for ~180 system voices with zero install.
+description: Local multilingual voice toolkit — speech-to-text (STT), text-to-speech (TTS), and language detection. Runs entirely offline on Apple Silicon, Linux, and Windows. No API keys, no cloud. NVIDIA Parakeet TDT for STT across 25 European languages, Kokoro-82M + Vosk-TTS for TTS, plus macOS AVSpeechSynthesizer for ~180 system voices with zero install.
 emoji: 🎙️
 
 requires:
   bins: [kesha]
 
 install:
-  - kind: npm
-    packages: [@drakulavich/kesha-voice-kit]
-    flags: [--global]
+  - kind: bash
+    cmd: bun add -g "@drakulavich/kesha-voice-kit"
   - kind: bash
     cmd: kesha install
 ---
@@ -23,7 +22,7 @@ Local voice toolkit: transcribe voice messages to text, synthesize speech, detec
 ## When to use
 
 - **Voice memo arrived** (Telegram, WhatsApp, Slack, Signal .ogg/.opus/.m4a): transcribe with `kesha --json <path>` and branch on the detected language.
-- **Need to reply with audio**: synthesize with `kesha say "<text>" > reply.wav`. Auto-routes by detected language (Kokoro-82M for English, Piper for Russian). For other languages and ~180 more voices use `--voice macos-*` on macOS (zero model download).
+- **Need to reply with audio**: synthesize with `kesha say "<text>" > reply.wav`. Auto-routes by detected language (Kokoro-82M for English, Vosk-TTS for Russian). For other languages and ~180 more voices use `--voice macos-*` on macOS (zero model download).
 - **Need to detect what language a file is in** before choosing a pipeline: `kesha --json audio.ogg` returns both audio-based and text-based language detection with confidence scores.
 
 ## STT: transcribe audio
@@ -57,9 +56,9 @@ Use `lang` (or the more detailed `audioLanguage`/`textLanguage`) to decide how t
 
 ```bash
 kesha say "Hello, world" > hello.wav               # auto-routes en → Kokoro-82M
-kesha say "Привет, мир" > privet.wav              # auto-routes ru → Piper
+kesha say "Привет, мир" > privet.wav              # auto-routes ru → Vosk-TTS
 kesha say --voice macos-de-DE "Guten Tag" > de.wav # any macOS system voice — German, French, Italian, ...
-kesha say --list-voices                            # Kokoro + Piper + ~180 macos-* voices
+kesha say --list-voices                            # Kokoro + Vosk-TTS + ~180 macos-* voices
 ```
 
 Output: WAV mono float32. `--out <path>` writes to a file instead of stdout.
@@ -71,18 +70,18 @@ Output: WAV mono float32. `--out <path>` writes to a file instead of stdout.
 ## Install
 
 ```bash
-bun add --global @drakulavich/kesha-voice-kit    # or: npm i -g @drakulavich/kesha-voice-kit
+bun add -g @drakulavich/kesha-voice-kit          # global CLI install
 kesha install                                    # downloads engine (~350 MB)
-kesha install --tts                              # adds Kokoro + Piper RU + ONNX G2P (~490 MB more, for TTS)
+kesha install --tts                              # adds Kokoro + Vosk-TTS RU (~990 MB more, for TTS)
 ```
 
-No system deps — G2P runs as ONNX alongside Kokoro/Piper. `macos-*` voices need no install either — they use voices already on the Mac.
+No system deps — English G2P is embedded (`misaki-rs`); Russian G2P is bundled inside Vosk-TTS. `macos-*` voices need no install either — they use voices already on the Mac.
 
 ## Supported languages
 
 **Speech-to-text (25):** Bulgarian, Croatian, Czech, Danish, Dutch, English, Estonian, Finnish, French, German, Greek, Hungarian, Italian, Latvian, Lithuanian, Maltese, Polish, Portuguese, Romanian, Russian, Slovak, Slovenian, Spanish, Swedish, Ukrainian.
 
-**Text-to-speech:** English (Kokoro-82M, ~70 voices), Russian (Piper `ru-denis`), plus any macOS system voice via `--voice macos-*`.
+**Text-to-speech:** English (Kokoro-82M, ~70 voices), Russian (Vosk-TTS, 5 baked-in speakers — default `ru-vosk-m02`), plus any macOS system voice via `--voice macos-*`.
 
 ## Performance
 

package/package.json

@@ -1,7 +1,7 @@
 {
   "name": "@drakulavich/kesha-voice-kit",
-  "version": "1.4.3",
-  "description": "Open-source voice toolkit. Speech-to-text (25 langs, ~19x faster than Whisper on Apple Silicon, ONNX fallback on CPU), text-to-speech (Kokoro + Piper + ~180 macOS system voices, SSML), voice-activity detection, language detection (107 langs). Rust engine, OpenClaw skill.",
+  "version": "1.5.0",
+  "description": "Open-source voice toolkit. Speech-to-text (25 langs, ~19x faster than Whisper on Apple Silicon, ONNX fallback on CPU), text-to-speech (Kokoro + Vosk-TTS + ~180 macOS system voices, SSML), voice-activity detection, language detection (107 langs). Rust engine, OpenClaw skill.",
   "type": "module",
   "bin": {
     "kesha": "bin/kesha.js",
@@ -30,7 +30,7 @@
     ]
   },
   "keshaEngine": {
-    "version": "1.4.1"
+    "version": "1.5.0"
   },
   "scripts": {
     "test": "bun test",

package/src/cli/install.ts

@@ -92,7 +92,7 @@ export const installCommand = defineCommand({
     },
     tts: {
       type: "boolean",
-      description: "Also install TTS models (Kokoro EN + Piper RU + ONNX G2P, ~490MB)",
+      description: "Also install TTS models (Kokoro EN + Vosk-TTS RU, ~990MB)",
       default: false,
     },
     vad: {

package/src/cli/say.ts

@@ -3,17 +3,25 @@ import { detectTextLanguageEngine, getEngineBinPath } from "../engine";
 import { log } from "../log";
 import { say, SayError } from "../say";
 
+/**
+ * Darwin defaults to AVSpeech Milena — zero install, no model download required.
+ * Linux/Windows fall through to Vosk-TTS `ru-vosk-m02` (male, per CLAUDE.md
+ * "DEFAULT TTS VOICES MUST BE MALE"; replaces Piper-ruslan as of #213).
+ */
+const RU_DARWIN_FALLBACK_VOICE = "macos-com.apple.voice.compact.ru-RU.Milena";
+
 /** Map a detected language code to a default voice id. Unknown / low-confidence → undefined. */
 export function pickVoiceForLang(
   code: string | undefined,
   confidence: number,
+  platform: NodeJS.Platform = process.platform,
 ): string | undefined {
   if (!code || confidence < 0.5) return undefined;
   switch (code) {
     case "en":
-      return "en-af_heart";
+      return "en-am_michael";
     case "ru":
-      return "ru-denis";
+      return platform === "darwin" ? RU_DARWIN_FALLBACK_VOICE : "ru-vosk-m02";
     default:
       return undefined;
   }
@@ -50,7 +58,7 @@ export const sayCommand = defineCommand({
   },
   args: {
     text: { type: "positional", required: false, description: "Text to speak (stdin if omitted)" },
-    voice: { type: "string", description: "Voice id, e.g. en-af_heart" },
+    voice: { type: "string", description: "Voice id, e.g. en-am_michael" },
     lang: { type: "string", description: "BCP 47 language code (default en-us)" },
     out: { type: "string", description: "Write WAV to file instead of stdout" },
     rate: { type: "string", description: "Speaking rate 0.5–2.0", default: "1.0" },

package/src/engine-install.ts

@@ -36,7 +36,14 @@ function getEngineBinaryName(): string {
 
   if (platform === "darwin" && arch === "arm64") return "kesha-engine-darwin-arm64";
   if (platform === "linux" && arch === "x64") return "kesha-engine-linux-x64";
-  if (platform === "win32" && arch === "x64") return "kesha-engine-windows-x64.exe";
+  if (platform === "win32" && arch === "x64") {
+    throw new Error(
+      "Windows x64 is temporarily unsupported in v1.5.0 — the Vosk-TTS engine has " +
+        "native deps that trip MSVC at link time. Tracked at " +
+        "https://github.com/drakulavich/kesha-voice-kit/issues/216. " +
+        "Use v1.4.x as a workaround until the fix lands.",
+    );
+  }
 
   throw new Error(`Unsupported platform: ${platform} ${arch}`);
 }
@@ -51,7 +58,7 @@ function getEngineBinaryName(): string {
  * Best-effort: 404s (older engine versions predate the sidecar) and
  * network errors log a warning and return — macos-* voices simply won't
  * be available, which is a graceful degradation. The user keeps Kokoro +
- * Piper.
+ * Vosk-TTS.
  */
 async function downloadAVSpeechSidecar(binPath: string, engineVersion: string): Promise<void> {
   if (process.platform !== "darwin" || process.arch !== "arm64") return;
@@ -95,7 +102,7 @@ async function downloadAVSpeechSidecar(binPath: string, engineVersion: string):
 }
 
 export interface InstallOptions {
-  /** Also install Kokoro + Piper + ONNX G2P TTS models. */
+  /** Also install Kokoro + Vosk-TTS models. */
   tts?: boolean;
   /** Also install Silero VAD model for long-audio preprocessing. */
   vad?: boolean;

package/src/say.ts

@@ -8,7 +8,7 @@ export interface SayOptions {
    * arg) handles stdin separately before invoking `say()`.
    */
   text?: string;
-  /** Voice id, e.g. `en-af_heart`. Defaults to engine default. */
+  /** Voice id, e.g. `en-am_michael`. Defaults to engine default. */
   voice?: string;
   /** Override the voice's default BCP 47 language code (e.g. `en-us`, `ru`). */
   lang?: string;

package/src/status.ts

@@ -104,8 +104,7 @@ function showDiskUsage(binPath: string): void {
     { label: "Language ID", path: join(cache, "models/lang-id-ecapa") },
     { label: "VAD (Silero)", path: join(cache, "models/silero-vad") },
     { label: "TTS (Kokoro)", path: join(cache, "models/kokoro-82m") },
-    { label: "TTS (Piper)", path: join(cache, "models/piper-ru") },
-    { label: "TTS (G2P)", path: join(cache, "models/g2p") },
+    { label: "TTS (Vosk)", path: join(cache, "models/vosk-ru") },
   ];
 
   const rows: Array<{ label: string; size: number }> = [];
@@ -173,16 +172,16 @@ function listInstalledVoices(): string[] {
     /* Kokoro not installed */
   }
   try {
-    // Piper RU files follow `ru_RU-<name>-<quality>.onnx` — report just `<name>`.
-    const piper = readdirSync(join(cache, "models", "piper-ru"));
-    for (const f of piper) {
-      if (!f.endsWith(".onnx")) continue;
-      const stem = f.replace(/\.onnx$/, "");
-      const name = stem.replace(/^ru_RU-/, "").split("-")[0];
-      if (name) voices.push(`ru-${name}`);
+    // Vosk-TTS Russian is a single multi-speaker model. Mirror the Rust-side
+    // gate (models::is_vosk_ru_cached) — checking model.onnx + bert/model.onnx
+    // avoids advertising voices that would fail to load on a partial install.
+    statSync(join(cache, "models", "vosk-ru", "model.onnx"));
+    statSync(join(cache, "models", "vosk-ru", "bert", "model.onnx"));
+    for (const id of ["f01", "f02", "f03", "m01", "m02"]) {
+      voices.push(`ru-vosk-${id}`);
     }
   } catch {
-    /* Piper not installed */
+    /* Vosk not installed */
   }
   return voices.sort();
 }