Spaces:
Sleeping
Sleeping
统一了代码风格
Browse files
api.py
CHANGED
@@ -1,8 +1,11 @@
|
|
1 |
# coding=utf-8
|
2 |
|
3 |
from io import BytesIO
|
4 |
-
from typing import Optional
|
|
|
|
|
5 |
|
|
|
6 |
from fastapi import FastAPI, File, UploadFile, HTTPException, Depends
|
7 |
from fastapi.middleware.cors import CORSMiddleware
|
8 |
from fastapi.security import HTTPBearer, HTTPAuthorizationCredentials
|
@@ -20,7 +23,7 @@ import gradio as gr
|
|
20 |
load_dotenv()
|
21 |
|
22 |
# 获取API Token
|
23 |
-
API_TOKEN = os.getenv("API_TOKEN")
|
24 |
if not API_TOKEN:
|
25 |
raise RuntimeError("API_TOKEN environment variable is not set")
|
26 |
|
@@ -52,7 +55,7 @@ model = AutoModel(
|
|
52 |
)
|
53 |
|
54 |
# 复用原有的格式化函数
|
55 |
-
|
56 |
"<|HAPPY|>": "😊",
|
57 |
"<|SAD|>": "😔",
|
58 |
"<|ANGRY|>": "😡",
|
@@ -62,7 +65,7 @@ emo_dict = {
|
|
62 |
"<|SURPRISED|>": "😮",
|
63 |
}
|
64 |
|
65 |
-
event_dict = {
|
66 |
"<|BGM|>": "🎼",
|
67 |
"<|Speech|>": "",
|
68 |
"<|Applause|>": "👏",
|
@@ -73,7 +76,7 @@ event_dict = {
|
|
73 |
"<|Cough|>": "🤧",
|
74 |
}
|
75 |
|
76 |
-
emoji_dict = {
|
77 |
"<|nospeech|><|Event_UNK|>": "❓",
|
78 |
"<|zh|>": "",
|
79 |
"<|en|>": "",
|
@@ -105,7 +108,7 @@ emoji_dict = {
|
|
105 |
"<|Event_UNK|>": "",
|
106 |
}
|
107 |
|
108 |
-
lang_dict = {
|
109 |
"<|zh|>": "<|lang|>",
|
110 |
"<|en|>": "<|lang|>",
|
111 |
"<|yue|>": "<|lang|>",
|
@@ -114,82 +117,105 @@ lang_dict = {
|
|
114 |
"<|nospeech|>": "<|lang|>",
|
115 |
}
|
116 |
|
117 |
-
emo_set = {"😊", "😔", "😡", "😰", "🤢", "😮"}
|
118 |
-
event_set = {"🎼", "👏", "😀", "😭", "🤧", "😷"}
|
119 |
|
120 |
|
121 |
-
def
|
122 |
-
|
123 |
-
|
124 |
-
|
|
|
125 |
|
126 |
|
127 |
-
def
|
128 |
-
|
129 |
-
|
130 |
-
|
131 |
-
|
132 |
-
|
133 |
-
|
134 |
-
|
135 |
-
|
136 |
-
for
|
137 |
-
if
|
138 |
-
|
139 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
140 |
|
|
|
141 |
for emoji in emo_set.union(event_set):
|
142 |
-
|
143 |
-
|
144 |
-
return
|
145 |
|
146 |
|
147 |
-
def
|
148 |
-
|
149 |
-
|
|
|
150 |
|
151 |
-
def get_event(
|
152 |
-
return
|
153 |
|
154 |
-
|
|
|
155 |
for lang in lang_dict:
|
156 |
-
|
157 |
-
|
158 |
-
|
159 |
-
|
160 |
-
|
161 |
-
|
|
|
|
|
|
|
|
|
162 |
continue
|
163 |
-
|
164 |
-
|
165 |
-
|
166 |
-
|
167 |
-
|
168 |
-
|
169 |
-
|
170 |
-
|
|
|
|
|
|
|
171 |
|
172 |
|
173 |
async def process_audio(audio_data: bytes, language: str = "auto") -> str:
|
174 |
-
"""
|
175 |
try:
|
176 |
-
#
|
177 |
audio_buffer = BytesIO(audio_data)
|
178 |
waveform, sample_rate = torchaudio.load(audio_buffer)
|
179 |
|
180 |
-
#
|
181 |
if waveform.shape[0] > 1:
|
182 |
waveform = waveform.mean(dim=0)
|
183 |
|
184 |
-
#
|
185 |
input_wav = waveform.numpy().astype(np.float32)
|
186 |
|
187 |
-
#
|
188 |
if sample_rate != 16000:
|
189 |
resampler = torchaudio.transforms.Resample(sample_rate, 16000)
|
190 |
input_wav = resampler(torch.from_numpy(input_wav)[None, :])[0, :].numpy()
|
191 |
|
192 |
-
#
|
193 |
text = model.generate(
|
194 |
input=input_wav,
|
195 |
cache={},
|
@@ -199,18 +225,18 @@ async def process_audio(audio_data: bytes, language: str = "auto") -> str:
|
|
199 |
merge_vad=True
|
200 |
)
|
201 |
|
202 |
-
#
|
203 |
result = text[0]["text"]
|
204 |
-
result =
|
205 |
|
206 |
return result
|
207 |
|
208 |
except Exception as e:
|
209 |
-
raise HTTPException(status_code=500, detail=f"
|
210 |
|
211 |
|
212 |
-
async def verify_token(credentials: HTTPAuthorizationCredentials = Depends(security)):
|
213 |
-
"""
|
214 |
if credentials.credentials != API_TOKEN:
|
215 |
raise HTTPException(
|
216 |
status_code=401,
|
@@ -225,49 +251,53 @@ async def transcribe_audio(
|
|
225 |
model: Optional[str] = "FunAudioLLM/SenseVoiceSmall",
|
226 |
language: Optional[str] = "auto",
|
227 |
token: HTTPAuthorizationCredentials = Depends(verify_token)
|
228 |
-
):
|
229 |
-
"""
|
230 |
|
231 |
Args:
|
232 |
-
file:
|
233 |
-
model:
|
234 |
-
language:
|
235 |
|
236 |
Returns:
|
237 |
-
{
|
238 |
-
"text": "
|
239 |
"error_code": 0,
|
240 |
"error_msg": "",
|
241 |
-
"process_time": 1.234 #
|
242 |
}
|
243 |
"""
|
244 |
start_time = time.time()
|
245 |
|
246 |
try:
|
|
|
247 |
if not file.filename.lower().endswith((".mp3", ".wav", ".flac", ".ogg", ".m4a")):
|
248 |
return {
|
249 |
"text": "",
|
250 |
"error_code": 400,
|
251 |
-
"error_msg": "
|
252 |
"process_time": time.time() - start_time
|
253 |
}
|
254 |
|
|
|
255 |
if model != "FunAudioLLM/SenseVoiceSmall":
|
256 |
return {
|
257 |
"text": "",
|
258 |
"error_code": 400,
|
259 |
-
"error_msg": "
|
260 |
"process_time": time.time() - start_time
|
261 |
}
|
262 |
|
|
|
263 |
if language not in ["auto", "zh", "en", "yue", "ja", "ko", "nospeech"]:
|
264 |
return {
|
265 |
"text": "",
|
266 |
"error_code": 400,
|
267 |
-
"error_msg": "
|
268 |
"process_time": time.time() - start_time
|
269 |
}
|
270 |
|
|
|
271 |
content = await file.read()
|
272 |
text = await process_audio(content, language)
|
273 |
|
@@ -287,33 +317,29 @@ async def transcribe_audio(
|
|
287 |
}
|
288 |
|
289 |
|
290 |
-
def transcribe_audio_gradio(audio, language="auto"):
|
291 |
-
"""Gradio
|
292 |
try:
|
293 |
if audio is None:
|
294 |
-
return "
|
295 |
|
296 |
-
#
|
297 |
-
|
298 |
-
|
299 |
-
|
300 |
-
print(fs, type(fs))
|
301 |
-
print(input_wav, type(input_wav))
|
302 |
-
print('------------------------------')
|
303 |
-
|
304 |
input_wav = input_wav.astype(np.float32) / np.iinfo(np.int16).max
|
305 |
|
306 |
-
#
|
307 |
if len(input_wav.shape) > 1:
|
308 |
input_wav = input_wav.mean(-1)
|
309 |
|
310 |
-
#
|
311 |
-
if
|
312 |
-
resampler = torchaudio.transforms.Resample(
|
313 |
-
|
314 |
-
input_wav = resampler(
|
315 |
|
316 |
-
#
|
317 |
text = model.generate(
|
318 |
input=input_wav,
|
319 |
cache={},
|
@@ -323,48 +349,53 @@ def transcribe_audio_gradio(audio, language="auto"):
|
|
323 |
merge_vad=True
|
324 |
)
|
325 |
|
326 |
-
#
|
327 |
result = text[0]["text"]
|
328 |
-
result =
|
329 |
|
330 |
return result
|
331 |
except Exception as e:
|
332 |
-
return f"
|
333 |
|
334 |
-
#
|
335 |
demo = gr.Interface(
|
336 |
fn=transcribe_audio_gradio,
|
337 |
inputs=[
|
338 |
-
gr.Audio(
|
|
|
|
|
|
|
|
|
339 |
gr.Dropdown(
|
340 |
choices=["auto", "zh", "en", "yue", "ja", "ko", "nospeech"],
|
341 |
value="auto",
|
342 |
-
label="
|
343 |
)
|
344 |
],
|
345 |
-
outputs=gr.Textbox(label="
|
346 |
-
title="SenseVoice
|
347 |
-
description="
|
348 |
examples=[
|
349 |
["examples/zh.mp3", "zh"],
|
350 |
["examples/en.mp3", "en"],
|
351 |
]
|
352 |
)
|
353 |
|
354 |
-
#
|
355 |
app = gr.mount_gradio_app(app, demo, path="/")
|
356 |
|
|
|
357 |
@app.get("/docs", include_in_schema=False)
|
358 |
async def custom_swagger_ui_html():
|
359 |
return HTMLResponse("""
|
360 |
<!DOCTYPE html>
|
361 |
<html>
|
362 |
<head>
|
363 |
-
<title>SenseVoice API
|
364 |
<meta http-equiv="refresh" content="0;url=/docs/" />
|
365 |
</head>
|
366 |
<body>
|
367 |
-
<p
|
368 |
</body>
|
369 |
</html>
|
370 |
""")
|
|
|
1 |
# coding=utf-8
|
2 |
|
3 |
from io import BytesIO
|
4 |
+
from typing import Optional, Dict, Any, List, Set, Union, Tuple
|
5 |
+
import os
|
6 |
+
import time
|
7 |
|
8 |
+
# Third-party imports
|
9 |
from fastapi import FastAPI, File, UploadFile, HTTPException, Depends
|
10 |
from fastapi.middleware.cors import CORSMiddleware
|
11 |
from fastapi.security import HTTPBearer, HTTPAuthorizationCredentials
|
|
|
23 |
load_dotenv()
|
24 |
|
25 |
# 获取API Token
|
26 |
+
API_TOKEN: str = os.getenv("API_TOKEN")
|
27 |
if not API_TOKEN:
|
28 |
raise RuntimeError("API_TOKEN environment variable is not set")
|
29 |
|
|
|
55 |
)
|
56 |
|
57 |
# 复用原有的格式化函数
|
58 |
+
emotion_dict: Dict[str, str] = {
|
59 |
"<|HAPPY|>": "😊",
|
60 |
"<|SAD|>": "😔",
|
61 |
"<|ANGRY|>": "😡",
|
|
|
65 |
"<|SURPRISED|>": "😮",
|
66 |
}
|
67 |
|
68 |
+
event_dict: Dict[str, str] = {
|
69 |
"<|BGM|>": "🎼",
|
70 |
"<|Speech|>": "",
|
71 |
"<|Applause|>": "👏",
|
|
|
76 |
"<|Cough|>": "🤧",
|
77 |
}
|
78 |
|
79 |
+
emoji_dict: Dict[str, str] = {
|
80 |
"<|nospeech|><|Event_UNK|>": "❓",
|
81 |
"<|zh|>": "",
|
82 |
"<|en|>": "",
|
|
|
108 |
"<|Event_UNK|>": "",
|
109 |
}
|
110 |
|
111 |
+
lang_dict: Dict[str, str] = {
|
112 |
"<|zh|>": "<|lang|>",
|
113 |
"<|en|>": "<|lang|>",
|
114 |
"<|yue|>": "<|lang|>",
|
|
|
117 |
"<|nospeech|>": "<|lang|>",
|
118 |
}
|
119 |
|
120 |
+
emo_set: Set[str] = {"😊", "😔", "😡", "😰", "🤢", "😮"}
|
121 |
+
event_set: Set[str] = {"🎼", "👏", "😀", "😭", "🤧", "😷"}
|
122 |
|
123 |
|
124 |
+
def format_text_basic(text: str) -> str:
|
125 |
+
"""Replace special tokens with corresponding emojis"""
|
126 |
+
for token in emoji_dict:
|
127 |
+
text = text.replace(token, emoji_dict[token])
|
128 |
+
return text
|
129 |
|
130 |
|
131 |
+
def format_text_with_emotion(text: str) -> str:
|
132 |
+
"""Format text with emotion and event markers"""
|
133 |
+
token_count: Dict[str, int] = {}
|
134 |
+
original_text = text
|
135 |
+
for token in emoji_dict:
|
136 |
+
token_count[token] = text.count(token)
|
137 |
+
|
138 |
+
# Determine dominant emotion
|
139 |
+
dominant_emotion = "<|NEUTRAL|>"
|
140 |
+
for emotion in emotion_dict:
|
141 |
+
if token_count[emotion] > token_count[dominant_emotion]:
|
142 |
+
dominant_emotion = emotion
|
143 |
+
|
144 |
+
# Add event markers
|
145 |
+
text = original_text
|
146 |
+
for event in event_dict:
|
147 |
+
if token_count[event] > 0:
|
148 |
+
text = event_dict[event] + text
|
149 |
+
|
150 |
+
# Replace all tokens with their emoji equivalents
|
151 |
+
for token in emoji_dict:
|
152 |
+
text = text.replace(token, emoji_dict[token])
|
153 |
+
|
154 |
+
# Add dominant emotion
|
155 |
+
text = text + emotion_dict[dominant_emotion]
|
156 |
|
157 |
+
# Clean up emoji spacing
|
158 |
for emoji in emo_set.union(event_set):
|
159 |
+
text = text.replace(" " + emoji, emoji)
|
160 |
+
text = text.replace(emoji + " ", emoji)
|
161 |
+
return text.strip()
|
162 |
|
163 |
|
164 |
+
def format_text_advanced(text: str) -> str:
|
165 |
+
"""Advanced text formatting with multilingual and complex token handling"""
|
166 |
+
def get_emotion(text: str) -> Optional[str]:
|
167 |
+
return text[-1] if text[-1] in emo_set else None
|
168 |
|
169 |
+
def get_event(text: str) -> Optional[str]:
|
170 |
+
return text[0] if text[0] in event_set else None
|
171 |
|
172 |
+
# Handle special cases
|
173 |
+
text = text.replace("<|nospeech|><|Event_UNK|>", "❓")
|
174 |
for lang in lang_dict:
|
175 |
+
text = text.replace(lang, "<|lang|>")
|
176 |
+
|
177 |
+
# Process text segments
|
178 |
+
text_segments: List[str] = [format_text_with_emotion(segment).strip() for segment in text.split("<|lang|>")]
|
179 |
+
formatted_text = " " + text_segments[0]
|
180 |
+
current_event = get_event(formatted_text)
|
181 |
+
|
182 |
+
# Merge segments
|
183 |
+
for i in range(1, len(text_segments)):
|
184 |
+
if not text_segments[i]:
|
185 |
continue
|
186 |
+
|
187 |
+
if get_event(text_segments[i]) == current_event and get_event(text_segments[i]) is not None:
|
188 |
+
text_segments[i] = text_segments[i][1:]
|
189 |
+
current_event = get_event(text_segments[i])
|
190 |
+
|
191 |
+
if get_emotion(text_segments[i]) is not None and get_emotion(text_segments[i]) == get_emotion(formatted_text):
|
192 |
+
formatted_text = formatted_text[:-1]
|
193 |
+
formatted_text += text_segments[i].strip()
|
194 |
+
|
195 |
+
formatted_text = formatted_text.replace("The.", " ")
|
196 |
+
return formatted_text.strip()
|
197 |
|
198 |
|
199 |
async def process_audio(audio_data: bytes, language: str = "auto") -> str:
|
200 |
+
"""Process audio data and return transcription result"""
|
201 |
try:
|
202 |
+
# Convert bytes to numpy array
|
203 |
audio_buffer = BytesIO(audio_data)
|
204 |
waveform, sample_rate = torchaudio.load(audio_buffer)
|
205 |
|
206 |
+
# Convert to mono channel
|
207 |
if waveform.shape[0] > 1:
|
208 |
waveform = waveform.mean(dim=0)
|
209 |
|
210 |
+
# Convert to numpy array and normalize
|
211 |
input_wav = waveform.numpy().astype(np.float32)
|
212 |
|
213 |
+
# Resample to 16kHz if needed
|
214 |
if sample_rate != 16000:
|
215 |
resampler = torchaudio.transforms.Resample(sample_rate, 16000)
|
216 |
input_wav = resampler(torch.from_numpy(input_wav)[None, :])[0, :].numpy()
|
217 |
|
218 |
+
# Model inference
|
219 |
text = model.generate(
|
220 |
input=input_wav,
|
221 |
cache={},
|
|
|
225 |
merge_vad=True
|
226 |
)
|
227 |
|
228 |
+
# Format result
|
229 |
result = text[0]["text"]
|
230 |
+
result = format_text_advanced(result)
|
231 |
|
232 |
return result
|
233 |
|
234 |
except Exception as e:
|
235 |
+
raise HTTPException(status_code=500, detail=f"Audio processing failed: {str(e)}")
|
236 |
|
237 |
|
238 |
+
async def verify_token(credentials: HTTPAuthorizationCredentials = Depends(security)) -> HTTPAuthorizationCredentials:
|
239 |
+
"""Verify Bearer Token authentication"""
|
240 |
if credentials.credentials != API_TOKEN:
|
241 |
raise HTTPException(
|
242 |
status_code=401,
|
|
|
251 |
model: Optional[str] = "FunAudioLLM/SenseVoiceSmall",
|
252 |
language: Optional[str] = "auto",
|
253 |
token: HTTPAuthorizationCredentials = Depends(verify_token)
|
254 |
+
) -> Dict[str, Union[str, int, float]]:
|
255 |
+
"""Audio transcription endpoint
|
256 |
|
257 |
Args:
|
258 |
+
file: Audio file (supports common audio formats)
|
259 |
+
model: Model name, currently only supports FunAudioLLM/SenseVoiceSmall
|
260 |
+
language: Language code, supports auto/zh/en/yue/ja/ko/nospeech
|
261 |
|
262 |
Returns:
|
263 |
+
Dict[str, Union[str, int, float]]: {
|
264 |
+
"text": "Transcription result",
|
265 |
"error_code": 0,
|
266 |
"error_msg": "",
|
267 |
+
"process_time": 1.234 # Processing time in seconds
|
268 |
}
|
269 |
"""
|
270 |
start_time = time.time()
|
271 |
|
272 |
try:
|
273 |
+
# Validate file format
|
274 |
if not file.filename.lower().endswith((".mp3", ".wav", ".flac", ".ogg", ".m4a")):
|
275 |
return {
|
276 |
"text": "",
|
277 |
"error_code": 400,
|
278 |
+
"error_msg": "Unsupported audio format",
|
279 |
"process_time": time.time() - start_time
|
280 |
}
|
281 |
|
282 |
+
# Validate model
|
283 |
if model != "FunAudioLLM/SenseVoiceSmall":
|
284 |
return {
|
285 |
"text": "",
|
286 |
"error_code": 400,
|
287 |
+
"error_msg": "Unsupported model",
|
288 |
"process_time": time.time() - start_time
|
289 |
}
|
290 |
|
291 |
+
# Validate language
|
292 |
if language not in ["auto", "zh", "en", "yue", "ja", "ko", "nospeech"]:
|
293 |
return {
|
294 |
"text": "",
|
295 |
"error_code": 400,
|
296 |
+
"error_msg": "Unsupported language",
|
297 |
"process_time": time.time() - start_time
|
298 |
}
|
299 |
|
300 |
+
# Process audio
|
301 |
content = await file.read()
|
302 |
text = await process_audio(content, language)
|
303 |
|
|
|
317 |
}
|
318 |
|
319 |
|
320 |
+
def transcribe_audio_gradio(audio: Optional[Tuple[int, np.ndarray]], language: str = "auto") -> str:
|
321 |
+
"""Gradio interface for audio transcription"""
|
322 |
try:
|
323 |
if audio is None:
|
324 |
+
return "Please upload an audio file"
|
325 |
|
326 |
+
# Extract audio data
|
327 |
+
sample_rate, input_wav = audio
|
328 |
+
|
329 |
+
# Normalize audio
|
|
|
|
|
|
|
|
|
330 |
input_wav = input_wav.astype(np.float32) / np.iinfo(np.int16).max
|
331 |
|
332 |
+
# Convert to mono
|
333 |
if len(input_wav.shape) > 1:
|
334 |
input_wav = input_wav.mean(-1)
|
335 |
|
336 |
+
# Resample to 16kHz if needed
|
337 |
+
if sample_rate != 16000:
|
338 |
+
resampler = torchaudio.transforms.Resample(sample_rate, 16000)
|
339 |
+
input_wav_tensor = torch.from_numpy(input_wav).to(torch.float32)
|
340 |
+
input_wav = resampler(input_wav_tensor[None, :])[0, :].numpy()
|
341 |
|
342 |
+
# Model inference
|
343 |
text = model.generate(
|
344 |
input=input_wav,
|
345 |
cache={},
|
|
|
349 |
merge_vad=True
|
350 |
)
|
351 |
|
352 |
+
# Format result
|
353 |
result = text[0]["text"]
|
354 |
+
result = format_text_advanced(result)
|
355 |
|
356 |
return result
|
357 |
except Exception as e:
|
358 |
+
return f"Processing failed: {str(e)}"
|
359 |
|
360 |
+
# Create Gradio interface with localized labels
|
361 |
demo = gr.Interface(
|
362 |
fn=transcribe_audio_gradio,
|
363 |
inputs=[
|
364 |
+
gr.Audio(
|
365 |
+
sources=["upload", "microphone"],
|
366 |
+
type="numpy",
|
367 |
+
label="Upload audio or record from microphone"
|
368 |
+
),
|
369 |
gr.Dropdown(
|
370 |
choices=["auto", "zh", "en", "yue", "ja", "ko", "nospeech"],
|
371 |
value="auto",
|
372 |
+
label="Select Language"
|
373 |
)
|
374 |
],
|
375 |
+
outputs=gr.Textbox(label="Recognition Result"),
|
376 |
+
title="SenseVoice Speech Recognition",
|
377 |
+
description="Multi-language speech transcription service supporting Chinese, English, Cantonese, Japanese, and Korean",
|
378 |
examples=[
|
379 |
["examples/zh.mp3", "zh"],
|
380 |
["examples/en.mp3", "en"],
|
381 |
]
|
382 |
)
|
383 |
|
384 |
+
# Mount Gradio app to FastAPI
|
385 |
app = gr.mount_gradio_app(app, demo, path="/")
|
386 |
|
387 |
+
# Custom Swagger UI redirect
|
388 |
@app.get("/docs", include_in_schema=False)
|
389 |
async def custom_swagger_ui_html():
|
390 |
return HTMLResponse("""
|
391 |
<!DOCTYPE html>
|
392 |
<html>
|
393 |
<head>
|
394 |
+
<title>SenseVoice API Documentation</title>
|
395 |
<meta http-equiv="refresh" content="0;url=/docs/" />
|
396 |
</head>
|
397 |
<body>
|
398 |
+
<p>Redirecting to API documentation...</p>
|
399 |
</body>
|
400 |
</html>
|
401 |
""")
|