Update app.py

app.py

@@ -1,14 +1,28 @@
-# --- add to your existing ActualTTS app.py ---
+import os, asyncio, base64, json
+from fastapi import FastAPI, WebSocket, WebSocketDisconnect
 
-from fastapi import WebSocket, WebSocketDisconnect
-import asyncio
-import base64
+# Create app FIRST
+app = FastAPI(title="ActualTTS Streaming")
+
+PIPER = "/usr/local/bin/piper"
+DEFAULT_VOICE = "en_US-amy-medium"
+VOICE_DIR = "/home/user/voices"
+
+def voice_paths(voice: str):
+    onnx = os.path.join(VOICE_DIR, f"{voice}.onnx")
+    cfg = os.path.join(VOICE_DIR, f"{voice}.onnx.json")
+    return onnx, cfg
+
+def ensure_voice_available(voice: str):
+    onnx, cfg = voice_paths(voice)
+    if not os.path.exists(onnx) or not os.path.exists(cfg):
+        return f"Voice {voice} not found in {VOICE_DIR}"
+    return None
 
 @app.websocket("/ws/tts")
 async def ws_tts(ws: WebSocket):
     await ws.accept()
     try:
-        # First message must be JSON: {"text": "...", "voice": "en_US-amy-medium"}
         init = await ws.receive_json()
         text = (init.get("text") or "").strip()
         voice = (init.get("voice") or DEFAULT_VOICE).strip()
@@ -20,38 +34,32 @@ async def ws_tts(ws: WebSocket):
             return
 
         onnx, cfg = voice_paths(voice)
-
-        # Piper in raw mode (S16LE PCM) -> stdout
         cmd = [
             PIPER, "-m", str(onnx), "-c", str(cfg),
-            "--output-raw", "-",
-            # You can pass knobs too, e.g. "--length-scale","1.0"
+            "--output-raw", "-",   # raw PCM
         ]
 
-        # Start process; feed text via stdin
         proc = await asyncio.create_subprocess_exec(
             *cmd,
             stdin=asyncio.subprocess.PIPE,
             stdout=asyncio.subprocess.PIPE,
             stderr=asyncio.subprocess.PIPE,
         )
-        # Tell client what to expect
-        # Piper voices define sample rate in JSON; common: 22050 or 24000
-        # If you want to read it, parse cfg; for now send a safe default 22050
+
+        # Tell client expected audio format
         await ws.send_json({"type": "format", "encoding": "s16le", "sample_rate": 22050})
 
-        # Send text
+        # Send text to Piper
         proc.stdin.write(text.encode("utf-8"))
         await proc.stdin.drain()
         proc.stdin.close()
 
-        # Stream stdout in chunks
-        CHUNK = 4096  # bytes
+        # Stream audio chunks
+        CHUNK = 4096
         while True:
             chunk = await proc.stdout.read(CHUNK)
             if not chunk:
                 break
-            # Send as base64 so browser/Qt can decode
             await ws.send_json({"type": "audio", "data": base64.b64encode(chunk).decode("ascii")})
 
         rc = await proc.wait()