pi-voice 0.3.3 → 0.4.0

package/README.md

@@ -2,7 +2,9 @@
 
 Headless voice interface for the [Pi Coding Agent](https://github.com/badlogic/pi-mono). Hold a key, speak, and pi executes your instructions with voice feedback.
 
-https://github.com/user-attachments/assets/06a0c56f-76bf-48cb-9de3-f9ca48a7245d
+#### Demo using ElevenLabs provider (make sure unmuted)
+
+https://github.com/user-attachments/assets/76adb941-83cf-4394-b8d2-f6d73a1df8bc
 
 ## Installation
 
@@ -52,7 +54,7 @@ You can configure pi-voice in `.pi/pi-voice.json`:
 | Key | Description |
 | --- | --- |
 | `key` | Push-to-talk shortcut. Combine modifiers (`ctrl`, `shift`, `alt`/`opt`, `meta`/`cmd`) and a main key with `+`. Examples: `"ctrl+t"`, `"alt+space"`, `"ctrl+shift+r"`. Default: `"meta+shift+i"`. |
-| `provider` | Speech provider for STT & TTS. `"local"`, `"gemini"` (Vertex AI or Gemini API), or `"openai"`. Default: `"local"`. |
+| `provider` | Speech provider for STT & TTS. `"local"`, `"gemini"` (Vertex AI or Gemini API), `"openai"`, or `"elevenlabs"`. Default: `"local"`. |
 
 ### Environment variables
 
@@ -61,6 +63,7 @@ You can configure pi-voice in `.pi/pi-voice.json`:
 | `local` | None (model is auto-downloaded on first launch). Optional: `WHISPER_MODEL_PATH` (custom model path), `WHISPER_MODEL` (model name, default `medium-q5_0`), `SAY_VOICE` (macOS `say` voice name, e.g. `"Kyoko"`). |
 | `gemini` | **Vertex AI:** `GOOGLE_CLOUD_PROJECT`, `GOOGLE_CLOUD_LOCATION` (optional, default `us-central1`). **Gemini API:** `GEMINI_API_KEY` or `GOOGLE_API_KEY`. If `GOOGLE_CLOUD_PROJECT` is set, Vertex AI is used; set `GOOGLE_GENAI_USE_VERTEXAI=false` to force API key mode. |
 | `openai` | `OPENAI_API_KEY` |
+| `elevenlabs` | `ELEVENLABS_API_KEY`. Optional: `ELEVENLABS_VOICE_ID` (TTS voice, default `CwhRBWXzGAHq8TQ4Fs17`), `ELEVENLABS_TTS_MODEL` (default `eleven_flash_v2_5`). |
 
 #### Logging
 

package/out/cli/cli.js

@@ -18511,7 +18511,7 @@ var configFileSchema = exports_external.object({
       return false;
     }
   }, { message: "Invalid key binding" }).optional().default(DEFAULT_KEY_STRING),
-  provider: exports_external.enum(["local", "gemini", "openai"]).optional().default(DEFAULT_PROVIDER)
+  provider: exports_external.enum(["local", "gemini", "openai", "elevenlabs"]).optional().default(DEFAULT_PROVIDER)
 });
 
 class ConfigError extends Error {

package/out/main/index.js

@@ -7,6 +7,7 @@ import pino from "pino";
 import { readFileSync, existsSync, mkdirSync, createWriteStream, unlinkSync, writeFileSync } from "node:fs";
 import { z } from "zod";
 import OpenAI, { toFile } from "openai";
+import { ElevenLabsClient } from "@elevenlabs/elevenlabs-js";
 import { WhisperFullParams, WhisperSamplingStrategy, Whisper } from "@napi-rs/whisper";
 import { GoogleGenAI } from "@google/genai";
 import { Readable } from "node:stream";
@@ -259,7 +260,7 @@ const configFileSchema = z.object({
     },
     { message: "Invalid key binding" }
   ).optional().default(DEFAULT_KEY_STRING),
-  provider: z.enum(["local", "gemini", "openai"]).optional().default(DEFAULT_PROVIDER)
+  provider: z.enum(["local", "gemini", "openai", "elevenlabs"]).optional().default(DEFAULT_PROVIDER)
 });
 class ConfigError extends Error {
   constructor(configPath, details) {
@@ -470,6 +471,34 @@ async function transcribeOpenAI(audioBuffer) {
   });
   return transcription.text?.trim() ?? "";
 }
+let elevenlabsClient$1 = null;
+function getElevenLabsClient$1() {
+  if (elevenlabsClient$1) return elevenlabsClient$1;
+  const apiKey = process.env.ELEVENLABS_API_KEY;
+  if (!apiKey) {
+    throw new Error("ELEVENLABS_API_KEY environment variable is required");
+  }
+  elevenlabsClient$1 = new ElevenLabsClient({ apiKey });
+  return elevenlabsClient$1;
+}
+async function transcribeElevenLabs(audioBuffer) {
+  const client = getElevenLabsClient$1();
+  const result = await client.speechToText.convert({
+    file: {
+      data: audioBuffer,
+      filename: "recording.webm",
+      contentType: "audio/webm"
+    },
+    modelId: "scribe_v2"
+  });
+  if ("text" in result) {
+    return (result.text ?? "").trim();
+  }
+  if ("transcripts" in result && result.transcripts?.[0]) {
+    return (result.transcripts[0].text ?? "").trim();
+  }
+  return "";
+}
 async function transcribeLocal(samples) {
   const whisper = await getWhisperInstance();
   const params = new WhisperFullParams(WhisperSamplingStrategy.Greedy);
@@ -492,6 +521,9 @@ async function transcribe(audioData, provider = "local") {
     case "openai":
       text = await transcribeOpenAI(Buffer.from(audioData));
       break;
+    case "elevenlabs":
+      text = await transcribeElevenLabs(Buffer.from(audioData));
+      break;
     case "gemini":
     default:
       text = await transcribeGemini(Buffer.from(audioData));
@@ -513,7 +545,7 @@ function getOpenAIClient() {
 const TTS_SAMPLE_RATE = 24e3;
 const TTS_CHANNELS = 1;
 const TTS_BITS_PER_SAMPLE = 16;
-const OPENAI_PCM_CHUNK_SIZE = TTS_SAMPLE_RATE * (TTS_BITS_PER_SAMPLE / 8) * TTS_CHANNELS * 0.1;
+const PCM_CHUNK_SIZE = TTS_SAMPLE_RATE * (TTS_BITS_PER_SAMPLE / 8) * TTS_CHANNELS * 0.1;
 async function* synthesizeStreamGemini(text) {
   const client = getGeminiClient();
   const response = await client.models.generateContentStream({
@@ -583,7 +615,7 @@ async function* synthesizeStreamOpenAI(text) {
   let totalBytes = 0;
   let offset = 0;
   while (offset < fullBuffer.length) {
-    const end = Math.min(offset + OPENAI_PCM_CHUNK_SIZE, fullBuffer.length);
+    const end = Math.min(offset + PCM_CHUNK_SIZE, fullBuffer.length);
     const chunk = fullBuffer.subarray(offset, end);
     totalBytes += chunk.length;
     yield chunk;
@@ -594,6 +626,48 @@ async function* synthesizeStreamOpenAI(text) {
     "Streamed PCM audio"
   );
 }
+let elevenlabsClient = null;
+function getElevenLabsClient() {
+  if (elevenlabsClient) return elevenlabsClient;
+  const apiKey = process.env.ELEVENLABS_API_KEY;
+  if (!apiKey) {
+    throw new Error("ELEVENLABS_API_KEY environment variable is required");
+  }
+  elevenlabsClient = new ElevenLabsClient({ apiKey });
+  return elevenlabsClient;
+}
+const DEFAULT_ELEVENLABS_VOICE_ID = "CwhRBWXzGAHq8TQ4Fs17";
+async function* synthesizeStreamElevenLabs(text) {
+  const client = getElevenLabsClient();
+  const voiceId = process.env.ELEVENLABS_VOICE_ID ?? DEFAULT_ELEVENLABS_VOICE_ID;
+  const modelId = process.env.ELEVENLABS_TTS_MODEL ?? "eleven_flash_v2_5";
+  const audio = await client.textToSpeech.convert(voiceId, {
+    text,
+    modelId,
+    outputFormat: "pcm_24000"
+  });
+  const chunks = [];
+  const reader = audio.getReader();
+  while (true) {
+    const { done, value } = await reader.read();
+    if (done) break;
+    chunks.push(value);
+  }
+  const fullBuffer = Buffer.concat(chunks);
+  let totalBytes = 0;
+  let offset = 0;
+  while (offset < fullBuffer.length) {
+    const end = Math.min(offset + PCM_CHUNK_SIZE, fullBuffer.length);
+    const chunk = fullBuffer.subarray(offset, end);
+    totalBytes += chunk.length;
+    yield chunk;
+    offset = end;
+  }
+  logger.info(
+    { provider: "elevenlabs", totalBytes, text: text.substring(0, 50) },
+    "Streamed PCM audio"
+  );
+}
 function speakLocal(text) {
   return new Promise((resolve, reject) => {
     if (process.platform !== "darwin") {
@@ -632,6 +706,9 @@ async function* synthesizeStream(text, provider = "local") {
     case "openai":
       yield* synthesizeStreamOpenAI(text);
       break;
+    case "elevenlabs":
+      yield* synthesizeStreamElevenLabs(text);
+      break;
     case "gemini":
     default:
       yield* synthesizeStreamGemini(text);

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "pi-voice",
-  "version": "0.3.3",
+  "version": "0.4.0",
   "description": "Voice interface for pi coding agent",
   "author": "Yuku Kotani",
   "license": "MIT",
@@ -36,6 +36,7 @@
     "typescript": "^5"
   },
   "dependencies": {
+    "@elevenlabs/elevenlabs-js": "^2.35.0",
     "@google/genai": "^1.40.0",
     "@mariozechner/pi-coding-agent": "^0.52.7",
     "@napi-rs/whisper": "^0.0.4",