Add wyoming-chatterbox.py

wyoming-chatterbox.py

@@ -0,0 +1,68 @@
+"""Wyoming protocol TTS wrapper for Chatterbox TTS server."""
+import argparse, asyncio, io, logging, wave
+import aiohttp
+from wyoming.audio import AudioChunk, AudioStop
+from wyoming.event import Event
+from wyoming.info import Attribution, Describe, Info, TtsProgram, TtsVoice
+from wyoming.server import AsyncEventHandler, AsyncServer
+from wyoming.tts import Synthesize
+
+_LOGGER = logging.getLogger(__name__)
+INFO = None
+
+class ChatterboxHandler(AsyncEventHandler):
+    def __init__(self, reader, writer, cli_args, *a, **kw):
+        super().__init__(reader, writer, *a, **kw)
+        self.cli_args = cli_args
+
+    async def handle_event(self, event: Event) -> bool:
+        if Describe.is_type(event.type):
+            await self.write_event(INFO.event())
+            return True
+        if not Synthesize.is_type(event.type):
+            return True
+        synth = Synthesize.from_event(event)
+        text = synth.text or ""
+        _LOGGER.info("Synthesizing: %s", text[:80])
+        try:
+            async with aiohttp.ClientSession() as session:
+                async with session.post(
+                    f"{self.cli_args.chatterbox_url}/v1/audio/speech",
+                    json={"model":"chatterbox","input":text,"voice":self.cli_args.voice,"response_format":"wav"},
+                    timeout=aiohttp.ClientTimeout(total=60)
+                ) as resp:
+                    wav_data = await resp.read()
+            with io.BytesIO(wav_data) as wav_io:
+                with wave.open(wav_io, "rb") as wf:
+                    audio = wf.readframes(wf.getnframes())
+                    await self.write_event(AudioChunk(rate=wf.getframerate(), width=wf.getsampwidth(), channels=wf.getnchannels(), audio=audio).event())
+                    await self.write_event(AudioStop().event())
+        except Exception:
+            _LOGGER.exception("TTS failed")
+        return True
+
+async def main():
+    global INFO
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--port", type=int, default=10201)
+    parser.add_argument("--chatterbox-url", default="http://10.2.1.104:8004")
+    parser.add_argument("--voice", default="deep.mp3")
+    args = parser.parse_args()
+
+    attr = Attribution(name="Chatterbox", url="https://github.com/resemble-ai/chatterbox")
+    INFO = Info(
+        tts=[TtsProgram(
+            name="chatterbox", description="Chatterbox TTS (Homer)",
+            attribution=attr, installed=True, version="1.0",
+            voices=[TtsVoice(name="deep", description="Deep - Lässige Assistentin",
+                             attribution=attr, version="1.0",
+                             languages=["de","en"], installed=True)],
+        )]
+    )
+    server = AsyncServer.from_uri(f"tcp://0.0.0.0:{args.port}")
+    _LOGGER.info("Wyoming Chatterbox on port %d, voice=%s", args.port, args.voice)
+    await server.run(lambda r,w: ChatterboxHandler(r, w, args))
+
+if __name__ == "__main__":
+    logging.basicConfig(level=logging.INFO)
+    asyncio.run(main())