| """Transcription via Whisper on Modal.""" |
|
|
| import base64 |
| import os |
| from pathlib import Path |
|
|
| import httpx |
|
|
|
|
| WHISPER_URL = os.environ.get( |
| "MODAL_WHISPER_URL", |
| "https://anubhavbharadwaaj--slugvoice-whisper-whisperserver-api.modal.run", |
| ) |
| _HERE = Path(__file__).resolve().parent |
| _SAMPLE_AUDIO = _HERE / "sample_session.wav" |
|
|
|
|
| def _resolve_audio_path(audio_path: str) -> Path: |
| """Return a stable audio path, tolerating evicted Gradio example temp files.""" |
| path = Path(audio_path) |
| if path.exists(): |
| return path |
|
|
| name = path.name.lower() |
| if name.startswith("sample_session") and _SAMPLE_AUDIO.exists(): |
| return _SAMPLE_AUDIO |
|
|
| raise FileNotFoundError( |
| f"Audio file is no longer available: {audio_path}. " |
| "Please re-upload the file and try again." |
| ) |
|
|
|
|
| def transcribe_audio(audio_path: str) -> str: |
| """Transcribe audio using Whisper on Modal.""" |
| resolved_path = _resolve_audio_path(audio_path) |
| with open(resolved_path, "rb") as f: |
| audio_b64 = base64.b64encode(f.read()).decode() |
|
|
| resp = httpx.post( |
| WHISPER_URL, |
| json={"audio": audio_b64}, |
| timeout=120, |
| ) |
| resp.raise_for_status() |
| return resp.json()["text"] |
|
|
