Spaces:
Runtime error
Runtime error
Create main.py
Browse files
main.py
ADDED
@@ -0,0 +1,57 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import time
|
2 |
+
from fastapi import FastAPI, File
|
3 |
+
from faster_whisper import WhisperModel
|
4 |
+
from utils import ffmpeg_read, stt
|
5 |
+
from sentence_transformers import SentenceTransformer, util
|
6 |
+
import torch
|
7 |
+
|
8 |
+
app = FastAPI()
|
9 |
+
|
10 |
+
whisper_models = ["tiny", "base", "small", "medium", "large-v1", "large-v2"]
|
11 |
+
audio_model = WhisperModel("base", compute_type="int8", device="cpu")
|
12 |
+
text_model = SentenceTransformer('all-MiniLM-L6-v2')
|
13 |
+
corpus_embeddings = torch.load('corpus_embeddings.pt')
|
14 |
+
|
15 |
+
|
16 |
+
def speech_to_text(upload_audio, model_type="whisper"):
|
17 |
+
"""
|
18 |
+
Transcribe audio using whisper model.
|
19 |
+
"""
|
20 |
+
audio_path = ffmpeg_read(upload_audio, sampling_rate=16000)
|
21 |
+
# Transcribe audio
|
22 |
+
if model_type == "whisper":
|
23 |
+
transcribe_options = dict(task="transcribe", language="ja", beam_size=5, best_of=5, vad_filter=True)
|
24 |
+
segments_raw, info = audio_model.transcribe(audio_path, **transcribe_options)
|
25 |
+
segments = [segment.text for segment in segments_raw]
|
26 |
+
return ' '.join(segments)
|
27 |
+
else:
|
28 |
+
text = stt(audio_path)
|
29 |
+
return text
|
30 |
+
|
31 |
+
|
32 |
+
@app.get("/")
|
33 |
+
def read_root():
|
34 |
+
return {"Message": "Application startup complete"}
|
35 |
+
|
36 |
+
|
37 |
+
@app.post("/voice_detect/")
|
38 |
+
async def voice_detect_api(
|
39 |
+
voice_input: bytes = File(None),
|
40 |
+
threshold: float = 0.8,
|
41 |
+
model_type: str = "whisper"
|
42 |
+
):
|
43 |
+
"""
|
44 |
+
API to detect voice from audio file.
|
45 |
+
"""
|
46 |
+
start = time.time()
|
47 |
+
text = speech_to_text(voice_input, model_type)
|
48 |
+
query_embedding = text_model.encode(text, convert_to_tensor=True)
|
49 |
+
hits = util.semantic_search(query_embedding, corpus_embeddings, top_k=1)[0]
|
50 |
+
if hits[0]['score'] > threshold:
|
51 |
+
similar = 1
|
52 |
+
else:
|
53 |
+
similar = 0
|
54 |
+
end = time.time()
|
55 |
+
return {"text": text,
|
56 |
+
"similar": similar,
|
57 |
+
"time_taken": end - start}
|