Spaces:
Sleeping
Sleeping
Add support for multiple input files and output files
Browse files- app.py +77 -29
- cli.py +0 -3
- src/download.py +9 -3
- src/source.py +70 -0
app.py
CHANGED
@@ -1,3 +1,4 @@
|
|
|
|
1 |
import math
|
2 |
from typing import Iterator
|
3 |
import argparse
|
@@ -6,9 +7,11 @@ from io import StringIO
|
|
6 |
import os
|
7 |
import pathlib
|
8 |
import tempfile
|
|
|
9 |
|
10 |
import torch
|
11 |
from src.modelCache import ModelCache
|
|
|
12 |
from src.vadParallel import ParallelContext, ParallelTranscription
|
13 |
|
14 |
# External programs
|
@@ -78,9 +81,9 @@ class WhisperTranscriber:
|
|
78 |
self.vad_cpu_cores = min(os.cpu_count(), MAX_AUTO_CPU_CORES)
|
79 |
print("[Auto parallel] Using GPU devices " + str(self.parallel_device_list) + " and " + str(self.vad_cpu_cores) + " CPU cores for VAD/transcription.")
|
80 |
|
81 |
-
def transcribe_webui(self, modelName, languageName, urlData,
|
82 |
try:
|
83 |
-
|
84 |
|
85 |
try:
|
86 |
selectedLanguage = languageName.lower() if len(languageName) > 0 else None
|
@@ -88,22 +91,84 @@ class WhisperTranscriber:
|
|
88 |
|
89 |
model = WhisperContainer(model_name=selectedModel, cache=self.model_cache)
|
90 |
|
91 |
-
#
|
92 |
-
|
|
|
|
|
|
|
93 |
|
94 |
# Write result
|
95 |
downloadDirectory = tempfile.mkdtemp()
|
96 |
-
|
97 |
-
|
98 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
99 |
|
100 |
return download, text, vtt
|
101 |
|
102 |
finally:
|
103 |
# Cleanup source
|
104 |
if self.deleteUploadedFiles:
|
105 |
-
|
106 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
107 |
|
108 |
except ExceededMaximumDuration as e:
|
109 |
return [], ("[ERROR]: Maximum remote video length is " + str(e.maxDuration) + "s, file was " + str(e.videoDuration) + "s"), "[ERROR]"
|
@@ -222,25 +287,8 @@ class WhisperTranscriber:
|
|
222 |
self.model_cache.clear()
|
223 |
self.vad_model = None
|
224 |
|
225 |
-
def __get_source(self, urlData,
|
226 |
-
|
227 |
-
# Download from YouTube
|
228 |
-
source = download_url(urlData, self.inputAudioMaxDuration)[0]
|
229 |
-
else:
|
230 |
-
# File input
|
231 |
-
source = uploadFile if uploadFile is not None else microphoneData
|
232 |
-
|
233 |
-
if self.inputAudioMaxDuration > 0:
|
234 |
-
# Calculate audio length
|
235 |
-
audioDuration = ffmpeg.probe(source)["format"]["duration"]
|
236 |
-
|
237 |
-
if float(audioDuration) > self.inputAudioMaxDuration:
|
238 |
-
raise ExceededMaximumDuration(videoDuration=audioDuration, maxDuration=self.inputAudioMaxDuration, message="Video is too long")
|
239 |
-
|
240 |
-
file_path = pathlib.Path(source)
|
241 |
-
sourceName = file_path.stem[:MAX_FILE_PREFIX_LENGTH] + file_path.suffix
|
242 |
-
|
243 |
-
return source, sourceName
|
244 |
|
245 |
def __get_max_line_width(self, language: str) -> int:
|
246 |
if (language and language.lower() in ["japanese", "ja", "chinese", "zh"]):
|
@@ -304,7 +352,7 @@ def create_ui(input_audio_max_duration, share=False, server_name: str = None, se
|
|
304 |
gr.Dropdown(choices=["tiny", "base", "small", "medium", "large"], value=default_model_name, label="Model"),
|
305 |
gr.Dropdown(choices=sorted(LANGUAGES), label="Language"),
|
306 |
gr.Text(label="URL (YouTube, etc.)"),
|
307 |
-
gr.
|
308 |
gr.Audio(source="microphone", type="filepath", label="Microphone Input"),
|
309 |
gr.Dropdown(choices=["transcribe", "translate"], label="Task"),
|
310 |
gr.Dropdown(choices=["none", "silero-vad", "silero-vad-skip-gaps", "silero-vad-expand-into-gaps", "periodic-vad"], value=default_vad, label="VAD"),
|
|
|
1 |
+
from datetime import datetime
|
2 |
import math
|
3 |
from typing import Iterator
|
4 |
import argparse
|
|
|
7 |
import os
|
8 |
import pathlib
|
9 |
import tempfile
|
10 |
+
import zipfile
|
11 |
|
12 |
import torch
|
13 |
from src.modelCache import ModelCache
|
14 |
+
from src.source import get_audio_source_collection
|
15 |
from src.vadParallel import ParallelContext, ParallelTranscription
|
16 |
|
17 |
# External programs
|
|
|
81 |
self.vad_cpu_cores = min(os.cpu_count(), MAX_AUTO_CPU_CORES)
|
82 |
print("[Auto parallel] Using GPU devices " + str(self.parallel_device_list) + " and " + str(self.vad_cpu_cores) + " CPU cores for VAD/transcription.")
|
83 |
|
84 |
+
def transcribe_webui(self, modelName, languageName, urlData, multipleFiles, microphoneData, task, vad, vadMergeWindow, vadMaxMergeSize, vadPadding, vadPromptWindow):
|
85 |
try:
|
86 |
+
sources = self.__get_source(urlData, multipleFiles, microphoneData)
|
87 |
|
88 |
try:
|
89 |
selectedLanguage = languageName.lower() if len(languageName) > 0 else None
|
|
|
91 |
|
92 |
model = WhisperContainer(model_name=selectedModel, cache=self.model_cache)
|
93 |
|
94 |
+
# Result
|
95 |
+
download = []
|
96 |
+
zip_file_lookup = {}
|
97 |
+
text = ""
|
98 |
+
vtt = ""
|
99 |
|
100 |
# Write result
|
101 |
downloadDirectory = tempfile.mkdtemp()
|
102 |
+
source_index = 0
|
103 |
+
|
104 |
+
# Execute whisper
|
105 |
+
for source in sources:
|
106 |
+
source_prefix = ""
|
107 |
+
|
108 |
+
if (len(sources) > 1):
|
109 |
+
# Prefix (minimum 2 digits)
|
110 |
+
source_index += 1
|
111 |
+
source_prefix = str(source_index).zfill(2) + "_"
|
112 |
+
print("Transcribing ", source.source_path)
|
113 |
+
|
114 |
+
# Transcribe
|
115 |
+
result = self.transcribe_file(model, source.source_path, selectedLanguage, task, vad, vadMergeWindow, vadMaxMergeSize, vadPadding, vadPromptWindow)
|
116 |
+
filePrefix = slugify(source_prefix + source.get_short_name(), allow_unicode=True)
|
117 |
+
|
118 |
+
source_download, source_text, source_vtt = self.write_result(result, filePrefix, downloadDirectory)
|
119 |
+
|
120 |
+
if len(sources) > 1:
|
121 |
+
# Add new line separators
|
122 |
+
if (len(source_text) > 0):
|
123 |
+
source_text += os.linesep + os.linesep
|
124 |
+
if (len(source_vtt) > 0):
|
125 |
+
source_vtt += os.linesep + os.linesep
|
126 |
+
|
127 |
+
# Append file name to source text too
|
128 |
+
source_text = source.get_full_name() + ":" + os.linesep + source_text
|
129 |
+
source_vtt = source.get_full_name() + ":" + os.linesep + source_vtt
|
130 |
+
|
131 |
+
# Add to result
|
132 |
+
download.extend(source_download)
|
133 |
+
text += source_text
|
134 |
+
vtt += source_vtt
|
135 |
+
|
136 |
+
if (len(sources) > 1):
|
137 |
+
# Zip files support at least 260 characters, but we'll play it safe and use 200
|
138 |
+
zipFilePrefix = slugify(source_prefix + source.get_short_name(max_length=200), allow_unicode=True)
|
139 |
+
|
140 |
+
# File names in ZIP file can be longer
|
141 |
+
for source_download_file in source_download:
|
142 |
+
# Get file postfix (after last -)
|
143 |
+
filePostfix = os.path.basename(source_download_file).split("-")[-1]
|
144 |
+
zip_file_name = zipFilePrefix + "-" + filePostfix
|
145 |
+
zip_file_lookup[source_download_file] = zip_file_name
|
146 |
+
|
147 |
+
# Create zip file from all sources
|
148 |
+
if len(sources) > 1:
|
149 |
+
downloadAllPath = os.path.join(downloadDirectory, "All_Output-" + datetime.now().strftime("%Y%m%d-%H%M%S") + ".zip")
|
150 |
+
|
151 |
+
with zipfile.ZipFile(downloadAllPath, 'w', zipfile.ZIP_DEFLATED) as zip:
|
152 |
+
for download_file in download:
|
153 |
+
# Get file name from lookup
|
154 |
+
zip_file_name = zip_file_lookup.get(download_file, os.path.basename(download_file))
|
155 |
+
zip.write(download_file, arcname=zip_file_name)
|
156 |
+
|
157 |
+
download.insert(0, downloadAllPath)
|
158 |
|
159 |
return download, text, vtt
|
160 |
|
161 |
finally:
|
162 |
# Cleanup source
|
163 |
if self.deleteUploadedFiles:
|
164 |
+
for source in sources:
|
165 |
+
print("Deleting source file " + source.source_path)
|
166 |
+
|
167 |
+
try:
|
168 |
+
os.remove(source.source_path)
|
169 |
+
except Exception as e:
|
170 |
+
# Ignore error - it's just a cleanup
|
171 |
+
print("Error deleting source file " + source.source_path + ": " + str(e))
|
172 |
|
173 |
except ExceededMaximumDuration as e:
|
174 |
return [], ("[ERROR]: Maximum remote video length is " + str(e.maxDuration) + "s, file was " + str(e.videoDuration) + "s"), "[ERROR]"
|
|
|
287 |
self.model_cache.clear()
|
288 |
self.vad_model = None
|
289 |
|
290 |
+
def __get_source(self, urlData, multipleFiles, microphoneData):
|
291 |
+
return get_audio_source_collection(urlData, multipleFiles, microphoneData, self.inputAudioMaxDuration)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
292 |
|
293 |
def __get_max_line_width(self, language: str) -> int:
|
294 |
if (language and language.lower() in ["japanese", "ja", "chinese", "zh"]):
|
|
|
352 |
gr.Dropdown(choices=["tiny", "base", "small", "medium", "large"], value=default_model_name, label="Model"),
|
353 |
gr.Dropdown(choices=sorted(LANGUAGES), label="Language"),
|
354 |
gr.Text(label="URL (YouTube, etc.)"),
|
355 |
+
gr.File(label="Upload Files", file_count="multiple"),
|
356 |
gr.Audio(source="microphone", type="filepath", label="Microphone Input"),
|
357 |
gr.Dropdown(choices=["transcribe", "translate"], label="Task"),
|
358 |
gr.Dropdown(choices=["none", "silero-vad", "silero-vad-skip-gaps", "silero-vad-expand-into-gaps", "periodic-vad"], value=default_vad, label="VAD"),
|
cli.py
CHANGED
@@ -5,8 +5,6 @@ from urllib.parse import urlparse
|
|
5 |
import warnings
|
6 |
import numpy as np
|
7 |
|
8 |
-
import whisper
|
9 |
-
|
10 |
import torch
|
11 |
from app import LANGUAGES, WhisperTranscriber
|
12 |
from src.download import download_url
|
@@ -14,7 +12,6 @@ from src.download import download_url
|
|
14 |
from src.utils import optional_float, optional_int, str2bool
|
15 |
from src.whisperContainer import WhisperContainer
|
16 |
|
17 |
-
|
18 |
def cli():
|
19 |
parser = argparse.ArgumentParser(formatter_class=argparse.ArgumentDefaultsHelpFormatter)
|
20 |
parser.add_argument("audio", nargs="+", type=str, help="audio file(s) to transcribe")
|
|
|
5 |
import warnings
|
6 |
import numpy as np
|
7 |
|
|
|
|
|
8 |
import torch
|
9 |
from app import LANGUAGES, WhisperTranscriber
|
10 |
from src.download import download_url
|
|
|
12 |
from src.utils import optional_float, optional_int, str2bool
|
13 |
from src.whisperContainer import WhisperContainer
|
14 |
|
|
|
15 |
def cli():
|
16 |
parser = argparse.ArgumentParser(formatter_class=argparse.ArgumentDefaultsHelpFormatter)
|
17 |
parser.add_argument("audio", nargs="+", type=str, help="audio file(s) to transcribe")
|
src/download.py
CHANGED
@@ -46,10 +46,16 @@ def _perform_download(url: str, maxDuration: int = None, outputTemplate: str = N
|
|
46 |
with YoutubeDL(ydl_opts) as ydl:
|
47 |
if maxDuration and maxDuration > 0:
|
48 |
info = ydl.extract_info(url, download=False)
|
49 |
-
|
50 |
|
51 |
-
|
52 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
53 |
|
54 |
ydl.add_post_processor(filename_collector)
|
55 |
ydl.download([url])
|
|
|
46 |
with YoutubeDL(ydl_opts) as ydl:
|
47 |
if maxDuration and maxDuration > 0:
|
48 |
info = ydl.extract_info(url, download=False)
|
49 |
+
entries = "entries" in info and info["entries"] or [info]
|
50 |
|
51 |
+
total_duration = 0
|
52 |
+
|
53 |
+
# Compute total duration
|
54 |
+
for entry in entries:
|
55 |
+
total_duration += float(entry["duration"])
|
56 |
+
|
57 |
+
if total_duration >= maxDuration:
|
58 |
+
raise ExceededMaximumDuration(videoDuration=total_duration, maxDuration=maxDuration, message="Video is too long")
|
59 |
|
60 |
ydl.add_post_processor(filename_collector)
|
61 |
ydl.download([url])
|
src/source.py
ADDED
@@ -0,0 +1,70 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# Gradio seems to truncate files without keeping the extension, so we need to truncate the file prefix ourself
|
2 |
+
import os
|
3 |
+
import pathlib
|
4 |
+
from typing import List
|
5 |
+
import zipfile
|
6 |
+
|
7 |
+
import ffmpeg
|
8 |
+
from more_itertools import unzip
|
9 |
+
|
10 |
+
from src.download import ExceededMaximumDuration, download_url
|
11 |
+
|
12 |
+
MAX_FILE_PREFIX_LENGTH = 17
|
13 |
+
|
14 |
+
class AudioSource:
|
15 |
+
def __init__(self, source_path, source_name = None):
|
16 |
+
self.source_path = source_path
|
17 |
+
self.source_name = source_name
|
18 |
+
|
19 |
+
# Load source name if not provided
|
20 |
+
if (self.source_name is None):
|
21 |
+
file_path = pathlib.Path(self.source_path)
|
22 |
+
self.source_name = file_path.name
|
23 |
+
|
24 |
+
def get_full_name(self):
|
25 |
+
return self.source_name
|
26 |
+
|
27 |
+
def get_short_name(self, max_length: int = MAX_FILE_PREFIX_LENGTH):
|
28 |
+
file_path = pathlib.Path(self.source_name)
|
29 |
+
short_name = file_path.stem[:max_length] + file_path.suffix
|
30 |
+
|
31 |
+
return short_name
|
32 |
+
|
33 |
+
def __str__(self) -> str:
|
34 |
+
return self.source_path
|
35 |
+
|
36 |
+
class AudioSourceCollection:
|
37 |
+
def __init__(self, sources: List[AudioSource]):
|
38 |
+
self.sources = sources
|
39 |
+
|
40 |
+
def __iter__(self):
|
41 |
+
return iter(self.sources)
|
42 |
+
|
43 |
+
def get_audio_source_collection(urlData: str, multipleFiles: List, microphoneData: str, input_audio_max_duration: float = -1) -> List[AudioSource]:
|
44 |
+
output: List[AudioSource] = []
|
45 |
+
|
46 |
+
if urlData:
|
47 |
+
# Download from YouTube. This could also be a playlist or a channel.
|
48 |
+
output.extend([ AudioSource(x) for x in download_url(urlData, input_audio_max_duration, playlistItems=None) ])
|
49 |
+
else:
|
50 |
+
# Add input files
|
51 |
+
if (multipleFiles is not None):
|
52 |
+
output.extend([ AudioSource(x.name) for x in multipleFiles ])
|
53 |
+
if (microphoneData is not None):
|
54 |
+
output.append(AudioSource(microphoneData))
|
55 |
+
|
56 |
+
total_duration = 0
|
57 |
+
|
58 |
+
# Calculate total audio length. We do this even if input_audio_max_duration
|
59 |
+
# is disabled to ensure that all the audio files are valid.
|
60 |
+
for source in output:
|
61 |
+
audioDuration = ffmpeg.probe(source.source_path)["format"]["duration"]
|
62 |
+
total_duration += float(audioDuration)
|
63 |
+
|
64 |
+
# Ensure the total duration of the audio is not too long
|
65 |
+
if input_audio_max_duration > 0:
|
66 |
+
if float(total_duration) > input_audio_max_duration:
|
67 |
+
raise ExceededMaximumDuration(videoDuration=total_duration, maxDuration=input_audio_max_duration, message="Video(s) is too long")
|
68 |
+
|
69 |
+
# Return a list of audio sources
|
70 |
+
return output
|