feat(voice/clone): support MP3 and other audio formats for reference voices (#20)

## Summary
- Adds support for MP3, FLAC, OGG, and other audio formats when
uploading reference voices for cloning
- Replace WAV-only `audio_read()` with a `soundfile`-based
implementation
- Update web UI file picker to accept multiple audio formats

## Changes
- `pocket_tts/data/audio.py`: Use `soundfile` library instead of
Python's `wave` module
- `pocket_tts/main.py`: Preserve original file extension in temp files
for format detection
- `pocket_tts/main.py`: Add missing `truncate=True` to CLI generate
command
- `pocket_tts/static/index.html`: Update file input to accept
`.wav,.mp3,.flac,.ogg,.m4a,audio/*`
- `pyproject.toml`: Add `soundfile>=0.12.0` dependency

## Tested
- [x] Upload a WAV file as reference voice and generate speech
- [x] Upload an MP3 file as reference voice and generate speech
- [x] Upload a FLAC file as reference voice and generate speech
- [x] Verify CLI `--voice` option works with MP3 files

---------

Co-authored-by: Gabriel de Marmiesse <gabrieldemarmiesse@gmail.com>
Co-authored-by: Václav Volhejn <8401624+vvolhejn@users.noreply.github.com>

pocket_tts/data/audio.py

@@ -21,17 +21,35 @@ FIRST_CHUNK_LENGTH_SECONDS = float(os.environ.get("FIRST_CHUNK_LENGTH_SECONDS",
 
 
 def audio_read(filepath: str | Path) -> tuple[torch.Tensor, int]:
-    """Read audio using Python's wave module."""
-    with wave.open(str(filepath), "rb") as wav_file:
-        sample_rate = wav_file.getframerate()
+    """Read audio file. WAV uses built-in wave module; other formats require soundfile."""
+    filepath = Path(filepath)
 
-        # Read all audio data as 16-bit signed integers
-        raw_data = wav_file.readframes(-1)
-        samples = np.frombuffer(raw_data, dtype=np.int16).astype(np.float32) / 32768.0
+    if filepath.suffix.lower() == ".wav":
+        # Use built-in wave module for WAV files
+        with wave.open(str(filepath), "rb") as wav_file:
+            sample_rate = wav_file.getframerate()
+            n_channels = wav_file.getnchannels()
+            raw_data = wav_file.readframes(-1)
+            samples = np.frombuffer(raw_data, dtype=np.int16).astype(np.float32) / 32768.0
+            if n_channels > 1:
+                samples = samples.reshape(-1, n_channels).mean(axis=1)
+            return torch.from_numpy(samples).unsqueeze(0), sample_rate
 
-        # Return as mono tensor (channels, samples)
-        wav = torch.from_numpy(samples.reshape(1, -1))
-        return wav, sample_rate
+    # For non-WAV formats, use soundfile (optional dependency)
+    try:
+        import soundfile as sf
+    except ImportError as e:
+        raise ImportError(
+            "soundfile is required to read non-WAV audio files. "
+            "Install with: `pip install soundfile` or `uvx --with soundfile`"
+        ) from e
+
+    data, sample_rate = sf.read(str(filepath), dtype="float32")
+    if data.ndim == 1:
+        wav = torch.from_numpy(data).unsqueeze(0)
+    else:
+        wav = torch.from_numpy(data.mean(axis=1)).unsqueeze(0)
+    return wav, sample_rate
 
 
 class StreamingWAVWriter:

pocket_tts/main.py

@@ -142,11 +142,12 @@ def text_to_speech(
             raise HTTPException(
                 status_code=400, detail="voice_url must start with http://, https://, or hf://"
             )
-        model_state = tts_model._cached_get_state_for_audio_prompt(voice_url, truncate=True)
+        model_state = tts_model._cached_get_state_for_audio_prompt(voice_url)
         logging.warning("Using voice from URL: %s", voice_url)
     elif voice_wav is not None:
-        # Use uploaded voice file
-        with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as temp_file:
+        # Use uploaded voice file - preserve extension for format detection
+        suffix = Path(voice_wav.filename).suffix if voice_wav.filename else ".wav"
+        with tempfile.NamedTemporaryFile(delete=False, suffix=suffix) as temp_file:
             content = voice_wav.file.read()
             temp_file.write(content)
             temp_file.flush()

pocket_tts/static/index.html

@@ -51,16 +51,16 @@
 
         <div class="bg-white p-4 rounded-lg shadow">
             <label for="voice-wav-input" class="block text-sm font-medium text-gray-700 mb-2">
-                Or upload a WAV file for voice cloning:
+                Or upload an audio file for voice cloning:
             </label>
             <input
                 type="file"
                 id="voice-wav-input"
                 class="w-full border rounded p-2"
-                accept=".wav,audio/wav"
+                accept=".wav,.mp3,.flac,.ogg,.m4a,audio/*"
             />
             <p class="text-xs text-gray-500 mt-1">
-                Upload a WAV file to use as voice reference. Takes precedence over voice URL.
+                Upload an audio file (WAV, MP3, FLAC, etc.) to use as voice reference. Takes precedence over voice URL.
             </p>
         </div>