Spaces:
Configuration error
Configuration error
Upload async_ai_voice.py
Browse files- async_ai_voice.py +24 -0
async_ai_voice.py
ADDED
|
@@ -0,0 +1,24 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
from transformers import pipeline
|
| 2 |
+
from typing import Dict, Any
|
| 3 |
+
|
| 4 |
+
# Initialize the transcription pipeline
|
| 5 |
+
transcriber = pipeline("automatic-speech-recognition", model="distil-whisper/distil-large-v3")
|
| 6 |
+
|
| 7 |
+
async def send_to_voice_AI(audio_data: bytes) -> Dict[str, Any]:
|
| 8 |
+
"""
|
| 9 |
+
Transcribes audio data using the Hugging Face Whisper model.
|
| 10 |
+
|
| 11 |
+
Args:
|
| 12 |
+
audio_data (bytes): The raw audio data to be transcribed.
|
| 13 |
+
|
| 14 |
+
Returns:
|
| 15 |
+
Dict[str, Any]: A dictionary containing the transcription result or error message.
|
| 16 |
+
"""
|
| 17 |
+
print("Transcribing audio")
|
| 18 |
+
try:
|
| 19 |
+
# Transcribe the audio
|
| 20 |
+
result = transcriber(audio_data)
|
| 21 |
+
return {"text": result["text"]}
|
| 22 |
+
except Exception as e:
|
| 23 |
+
print(f"Error during transcription: {str(e)}")
|
| 24 |
+
return {"text": "Sorry, an error occurred during transcription."}
|