Spaces:
Sleeping
Sleeping
Let max line width depend on the language
Browse files
app.py
CHANGED
@@ -1,3 +1,4 @@
|
|
|
|
1 |
from typing import Iterator
|
2 |
|
3 |
from io import StringIO
|
@@ -15,15 +16,15 @@ from download import downloadUrl
|
|
15 |
|
16 |
from utils import slugify, write_srt, write_vtt
|
17 |
|
18 |
-
#import os
|
19 |
-
#os.system("pip install git+https://github.com/openai/whisper.git")
|
20 |
-
|
21 |
# Limitations (set to -1 to disable)
|
22 |
DEFAULT_INPUT_AUDIO_MAX_DURATION = 600 # seconds
|
23 |
|
24 |
# Whether or not to automatically delete all uploaded files, to save disk space
|
25 |
DELETE_UPLOADED_FILES = True
|
26 |
|
|
|
|
|
|
|
27 |
LANGUAGES = [
|
28 |
"English", "Chinese", "German", "Spanish", "Russian", "Korean",
|
29 |
"French", "Japanese", "Portuguese", "Turkish", "Polish", "Catalan",
|
@@ -74,8 +75,13 @@ class UI:
|
|
74 |
result = model.transcribe(source, language=selectedLanguage, task=task)
|
75 |
|
76 |
text = result["text"]
|
77 |
-
|
78 |
-
|
|
|
|
|
|
|
|
|
|
|
79 |
|
80 |
# Files that can be downloaded
|
81 |
downloadDirectory = tempfile.mkdtemp()
|
@@ -95,6 +101,15 @@ class UI:
|
|
95 |
os.remove(source)
|
96 |
|
97 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
98 |
def getSource(urlData, uploadFile, microphoneData):
|
99 |
if urlData:
|
100 |
# Download from YouTube
|
@@ -104,7 +119,7 @@ def getSource(urlData, uploadFile, microphoneData):
|
|
104 |
source = uploadFile if uploadFile is not None else microphoneData
|
105 |
|
106 |
file_path = pathlib.Path(source)
|
107 |
-
sourceName = file_path.stem[:
|
108 |
|
109 |
return source, sourceName
|
110 |
|
@@ -115,13 +130,13 @@ def createFile(text: str, directory: str, fileName: str) -> str:
|
|
115 |
|
116 |
return file.name
|
117 |
|
118 |
-
def getSubs(segments: Iterator[dict], format: str) -> str:
|
119 |
segmentStream = StringIO()
|
120 |
|
121 |
if format == 'vtt':
|
122 |
-
write_vtt(segments, file=segmentStream)
|
123 |
elif format == 'srt':
|
124 |
-
write_srt(segments, file=segmentStream)
|
125 |
else:
|
126 |
raise Exception("Unknown format " + format)
|
127 |
|
|
|
1 |
+
import re
|
2 |
from typing import Iterator
|
3 |
|
4 |
from io import StringIO
|
|
|
16 |
|
17 |
from utils import slugify, write_srt, write_vtt
|
18 |
|
|
|
|
|
|
|
19 |
# Limitations (set to -1 to disable)
|
20 |
DEFAULT_INPUT_AUDIO_MAX_DURATION = 600 # seconds
|
21 |
|
22 |
# Whether or not to automatically delete all uploaded files, to save disk space
|
23 |
DELETE_UPLOADED_FILES = True
|
24 |
|
25 |
+
# Gradio seems to truncate files without keeping the extension, so we need to truncate the file prefix ourself
|
26 |
+
MAX_FILE_PREFIX_LENGTH = 17
|
27 |
+
|
28 |
LANGUAGES = [
|
29 |
"English", "Chinese", "German", "Spanish", "Russian", "Korean",
|
30 |
"French", "Japanese", "Portuguese", "Turkish", "Polish", "Catalan",
|
|
|
75 |
result = model.transcribe(source, language=selectedLanguage, task=task)
|
76 |
|
77 |
text = result["text"]
|
78 |
+
|
79 |
+
language = result["language"]
|
80 |
+
languageMaxLineWidth = getMaxLineWidth(language)
|
81 |
+
|
82 |
+
print("Max line width " + str(languageMaxLineWidth))
|
83 |
+
vtt = getSubs(result["segments"], "vtt", languageMaxLineWidth)
|
84 |
+
srt = getSubs(result["segments"], "srt", languageMaxLineWidth)
|
85 |
|
86 |
# Files that can be downloaded
|
87 |
downloadDirectory = tempfile.mkdtemp()
|
|
|
101 |
os.remove(source)
|
102 |
|
103 |
|
104 |
+
def getMaxLineWidth(language: str) -> int:
|
105 |
+
if (language == "ja" or language == "zh"):
|
106 |
+
# Chinese characters and kana are wider, so limit line length to 40 characters
|
107 |
+
return 40
|
108 |
+
else:
|
109 |
+
# TODO: Add more languages
|
110 |
+
# 80 latin characters should fit on a 1080p/720p screen
|
111 |
+
return 80
|
112 |
+
|
113 |
def getSource(urlData, uploadFile, microphoneData):
|
114 |
if urlData:
|
115 |
# Download from YouTube
|
|
|
119 |
source = uploadFile if uploadFile is not None else microphoneData
|
120 |
|
121 |
file_path = pathlib.Path(source)
|
122 |
+
sourceName = file_path.stem[:MAX_FILE_PREFIX_LENGTH] + file_path.suffix
|
123 |
|
124 |
return source, sourceName
|
125 |
|
|
|
130 |
|
131 |
return file.name
|
132 |
|
133 |
+
def getSubs(segments: Iterator[dict], format: str, maxLineWidth: int) -> str:
|
134 |
segmentStream = StringIO()
|
135 |
|
136 |
if format == 'vtt':
|
137 |
+
write_vtt(segments, file=segmentStream, maxLineWidth=maxLineWidth)
|
138 |
elif format == 'srt':
|
139 |
+
write_srt(segments, file=segmentStream, maxLineWidth=maxLineWidth)
|
140 |
else:
|
141 |
raise Exception("Unknown format " + format)
|
142 |
|
utils.py
CHANGED
@@ -53,10 +53,10 @@ def write_txt(transcript: Iterator[dict], file: TextIO):
|
|
53 |
print(segment['text'].strip(), file=file, flush=True)
|
54 |
|
55 |
|
56 |
-
def write_vtt(transcript: Iterator[dict], file: TextIO):
|
57 |
print("WEBVTT\n", file=file)
|
58 |
for segment in transcript:
|
59 |
-
text = processText(segment['text']).replace('-->', '->')
|
60 |
|
61 |
print(
|
62 |
f"{format_timestamp(segment['start'])} --> {format_timestamp(segment['end'])}\n"
|
@@ -66,7 +66,7 @@ def write_vtt(transcript: Iterator[dict], file: TextIO):
|
|
66 |
)
|
67 |
|
68 |
|
69 |
-
def write_srt(transcript: Iterator[dict], file: TextIO):
|
70 |
"""
|
71 |
Write a transcript to a file in SRT format.
|
72 |
Example usage:
|
@@ -79,7 +79,7 @@ def write_srt(transcript: Iterator[dict], file: TextIO):
|
|
79 |
write_srt(result["segments"], file=srt)
|
80 |
"""
|
81 |
for i, segment in enumerate(transcript, start=1):
|
82 |
-
text = processText(segment['text'].strip()).replace('-->', '->')
|
83 |
|
84 |
# write srt lines
|
85 |
print(
|
@@ -91,8 +91,11 @@ def write_srt(transcript: Iterator[dict], file: TextIO):
|
|
91 |
flush=True,
|
92 |
)
|
93 |
|
94 |
-
def processText(text: str):
|
95 |
-
|
|
|
|
|
|
|
96 |
return '\n'.join(lines)
|
97 |
|
98 |
def slugify(value, allow_unicode=False):
|
|
|
53 |
print(segment['text'].strip(), file=file, flush=True)
|
54 |
|
55 |
|
56 |
+
def write_vtt(transcript: Iterator[dict], file: TextIO, maxLineWidth=None):
|
57 |
print("WEBVTT\n", file=file)
|
58 |
for segment in transcript:
|
59 |
+
text = processText(segment['text'], maxLineWidth).replace('-->', '->')
|
60 |
|
61 |
print(
|
62 |
f"{format_timestamp(segment['start'])} --> {format_timestamp(segment['end'])}\n"
|
|
|
66 |
)
|
67 |
|
68 |
|
69 |
+
def write_srt(transcript: Iterator[dict], file: TextIO, maxLineWidth=None):
|
70 |
"""
|
71 |
Write a transcript to a file in SRT format.
|
72 |
Example usage:
|
|
|
79 |
write_srt(result["segments"], file=srt)
|
80 |
"""
|
81 |
for i, segment in enumerate(transcript, start=1):
|
82 |
+
text = processText(segment['text'].strip(), maxLineWidth).replace('-->', '->')
|
83 |
|
84 |
# write srt lines
|
85 |
print(
|
|
|
91 |
flush=True,
|
92 |
)
|
93 |
|
94 |
+
def processText(text: str, maxLineWidth=None):
|
95 |
+
if (maxLineWidth is None or maxLineWidth < 0):
|
96 |
+
return text
|
97 |
+
|
98 |
+
lines = textwrap.wrap(text, width=maxLineWidth, tabsize=4)
|
99 |
return '\n'.join(lines)
|
100 |
|
101 |
def slugify(value, allow_unicode=False):
|