Spaces:
Sleeping
Sleeping
Hugo Rodrigues
commited on
Commit
·
6db451f
1
Parent(s):
357cae7
transcribe-audio
Browse files
main.py
CHANGED
@@ -1,10 +1,14 @@
|
|
|
|
1 |
import time
|
2 |
from scipy.io.wavfile import write
|
|
|
|
|
3 |
|
4 |
|
5 |
# from typing import Union
|
6 |
# from pydantic import BaseModel
|
7 |
-
|
|
|
8 |
from fastapi.middleware.cors import CORSMiddleware
|
9 |
from fastapi.responses import FileResponse
|
10 |
|
@@ -93,3 +97,23 @@ async def audio(inputs, src_lang="eng", tgt_lang="por", speaker_id=5):
|
|
93 |
audio_array_from_text)
|
94 |
|
95 |
return FileResponse(f"/tmp/output{start_time}.wav", media_type="audio/mpeg")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
|
2 |
import time
|
3 |
from scipy.io.wavfile import write
|
4 |
+
import torchaudio
|
5 |
+
import numpy as np
|
6 |
|
7 |
|
8 |
# from typing import Union
|
9 |
# from pydantic import BaseModel
|
10 |
+
|
11 |
+
from fastapi import FastAPI, UploadFile
|
12 |
from fastapi.middleware.cors import CORSMiddleware
|
13 |
from fastapi.responses import FileResponse
|
14 |
|
|
|
97 |
audio_array_from_text)
|
98 |
|
99 |
return FileResponse(f"/tmp/output{start_time}.wav", media_type="audio/mpeg")
|
100 |
+
|
101 |
+
|
102 |
+
@app.post("/transcribe-audio")
|
103 |
+
async def transcribe_audio(soundFile: UploadFile, tgt_lang='eng'):
|
104 |
+
start_time = time.time()
|
105 |
+
# process input
|
106 |
+
inputFile = soundFile.file.read()
|
107 |
+
|
108 |
+
audio_data = np.frombuffer(inputFile, dtype=np.int16)
|
109 |
+
|
110 |
+
audio_inputs = processor(
|
111 |
+
audios=audio_data, return_tensors="pt").to(device)
|
112 |
+
|
113 |
+
audio_array_from_audio = model.generate(
|
114 |
+
**audio_inputs, tgt_lang=tgt_lang)[0].cpu().numpy().squeeze()
|
115 |
+
|
116 |
+
write(f"/tmp/output{start_time}.wav", model.config.sampling_rate,
|
117 |
+
audio_array_from_audio)
|
118 |
+
|
119 |
+
return FileResponse(f"/tmp/output{start_time}.wav", media_type="audio/wav")
|