Spaces:
Running
Running
Update app.py
Browse files
app.py
CHANGED
|
@@ -7,57 +7,159 @@ from faster_whisper import WhisperModel
|
|
| 7 |
from indic_transliteration import sanscript
|
| 8 |
from indic_transliteration.sanscript import transliterate
|
| 9 |
|
| 10 |
-
#
|
| 11 |
-
#
|
| 12 |
-
#
|
| 13 |
-
|
| 14 |
|
| 15 |
def load_model():
|
| 16 |
-
global
|
| 17 |
-
if
|
| 18 |
-
|
| 19 |
-
return
|
| 20 |
-
|
| 21 |
-
#
|
| 22 |
-
#
|
| 23 |
-
#
|
| 24 |
def get_ffmpeg():
|
| 25 |
return shutil.which("ffmpeg") or "/usr/bin/ffmpeg"
|
| 26 |
|
| 27 |
-
|
| 28 |
-
|
| 29 |
-
|
| 30 |
-
|
| 31 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 32 |
|
| 33 |
-
|
| 34 |
-
os.remove(video_path)
|
| 35 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 36 |
ydl_opts = {
|
| 37 |
-
"format": "best",
|
| 38 |
-
"outtmpl":
|
| 39 |
-
"quiet": True,
|
| 40 |
"nocheckcertificate": True,
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 41 |
}
|
| 42 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 43 |
with yt_dlp.YoutubeDL(ydl_opts) as ydl:
|
| 44 |
ydl.download([url])
|
| 45 |
|
| 46 |
-
|
|
|
|
|
|
|
|
|
|
| 47 |
|
| 48 |
-
#
|
| 49 |
-
#
|
| 50 |
-
#
|
| 51 |
-
def
|
| 52 |
audio_path = "extracted_audio.wav"
|
| 53 |
-
|
| 54 |
if os.path.exists(audio_path):
|
| 55 |
os.remove(audio_path)
|
| 56 |
|
| 57 |
subprocess.run(
|
| 58 |
[
|
| 59 |
-
get_ffmpeg(),
|
| 60 |
-
"-y",
|
| 61 |
"-i", video_path,
|
| 62 |
"-vn",
|
| 63 |
"-ac", "1",
|
|
@@ -67,137 +169,415 @@ def extract_audio_safe(video_path):
|
|
| 67 |
stdout=subprocess.DEVNULL,
|
| 68 |
stderr=subprocess.DEVNULL
|
| 69 |
)
|
| 70 |
-
|
| 71 |
return audio_path
|
| 72 |
|
| 73 |
-
#
|
| 74 |
-
#
|
| 75 |
-
#
|
| 76 |
def normalize_script(text, lang):
|
| 77 |
if lang == "hi":
|
| 78 |
try:
|
| 79 |
return transliterate(text, sanscript.ARABIC, sanscript.DEVANAGARI)
|
| 80 |
-
except:
|
| 81 |
return text
|
| 82 |
return text
|
| 83 |
|
| 84 |
-
#
|
| 85 |
-
#
|
| 86 |
-
#
|
| 87 |
-
def transcribe(url, file, lang_choice):
|
| 88 |
try:
|
| 89 |
-
#
|
| 90 |
if file:
|
| 91 |
ext = os.path.splitext(file)[1].lower()
|
| 92 |
-
if ext in [".mp3", ".wav", ".m4a"]
|
| 93 |
-
audio = file
|
| 94 |
-
else:
|
| 95 |
-
audio = extract_audio_safe(file)
|
| 96 |
|
| 97 |
-
#
|
| 98 |
-
elif url:
|
| 99 |
-
|
| 100 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 101 |
|
| 102 |
else:
|
| 103 |
return "β οΈ Please paste a URL or upload a file."
|
| 104 |
|
| 105 |
-
# Safety check
|
| 106 |
-
if not os.path.exists(audio) or os.path.getsize(audio) <
|
| 107 |
-
return "β Audio extraction failed.
|
| 108 |
|
| 109 |
-
|
|
|
|
| 110 |
language = None if lang_choice == "Auto Detect" else lang_choice
|
| 111 |
|
| 112 |
-
segments, info =
|
| 113 |
audio,
|
| 114 |
beam_size=1,
|
| 115 |
vad_filter=True,
|
| 116 |
language=language
|
| 117 |
)
|
| 118 |
|
| 119 |
-
raw_text
|
| 120 |
final_text = normalize_script(raw_text, info.language)
|
| 121 |
|
| 122 |
-
return f"π Detected Language: {info.language}\n\n{final_text
|
| 123 |
|
| 124 |
except Exception as e:
|
| 125 |
-
|
| 126 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 127 |
return f"β Error: {str(e)}"
|
| 128 |
|
| 129 |
-
#
|
| 130 |
-
#
|
| 131 |
-
#
|
| 132 |
-
|
| 133 |
-
|
| 134 |
-
|
| 135 |
-
|
| 136 |
-
|
| 137 |
-
|
| 138 |
-
|
| 139 |
-
|
| 140 |
-
|
| 141 |
-
|
| 142 |
-
}
|
| 143 |
-
.
|
| 144 |
-
|
| 145 |
-
|
| 146 |
-
|
| 147 |
-
|
| 148 |
-
}
|
| 149 |
-
|
| 150 |
-
|
| 151 |
-
color:
|
| 152 |
-
}
|
| 153 |
-
|
| 154 |
-
|
| 155 |
-
|
| 156 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 157 |
"""
|
| 158 |
|
| 159 |
-
|
| 160 |
-
|
| 161 |
-
|
| 162 |
-
|
| 163 |
-
|
| 164 |
-
|
| 165 |
-
|
| 166 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 167 |
|
| 168 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 169 |
with gr.TabItem("π Paste Link"):
|
| 170 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
| 171 |
btn_url = gr.Button("π§ Transcribe Link", variant="primary")
|
| 172 |
|
|
|
|
| 173 |
with gr.TabItem("π Upload File"):
|
| 174 |
-
|
| 175 |
label="Upload Video / Audio",
|
| 176 |
-
file_types=[".mp4", ".mkv", ".mov", ".webm", ".avi",
|
|
|
|
| 177 |
)
|
| 178 |
btn_file = gr.Button("π Transcribe File", variant="primary")
|
| 179 |
|
|
|
|
|
|
|
|
|
|
| 180 |
lang = gr.Dropdown(
|
| 181 |
label="π Transcript Language",
|
| 182 |
choices=[
|
| 183 |
-
"Auto Detect",
|
| 184 |
-
"
|
| 185 |
-
"ur",
|
| 186 |
-
"en",
|
| 187 |
-
"ar",
|
| 188 |
-
"fr",
|
| 189 |
-
"de",
|
| 190 |
-
"es",
|
| 191 |
-
"ru",
|
| 192 |
-
"ja",
|
| 193 |
-
"zh"
|
| 194 |
],
|
| 195 |
value="Auto Detect"
|
| 196 |
)
|
| 197 |
|
| 198 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 199 |
|
| 200 |
-
|
| 201 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 202 |
|
| 203 |
-
demo.launch()
|
|
|
|
| 7 |
from indic_transliteration import sanscript
|
| 8 |
from indic_transliteration.sanscript import transliterate
|
| 9 |
|
| 10 |
+
# βββββββββββββββββββββββββββββββββββββββββββββββ
|
| 11 |
+
# WHISPER MODEL (lazy load)
|
| 12 |
+
# βββββββββββββββββββββββββββββββββββββββββββββββ
|
| 13 |
+
_model = None
|
| 14 |
|
| 15 |
def load_model():
|
| 16 |
+
global _model
|
| 17 |
+
if _model is None:
|
| 18 |
+
_model = WhisperModel("base", device="cpu", compute_type="int8")
|
| 19 |
+
return _model
|
| 20 |
+
|
| 21 |
+
# βββββββββββββββββββββββββββββββββββββββββββββββ
|
| 22 |
+
# UTILS
|
| 23 |
+
# βββββββββββββββββββββββββββββββββββββββββββββββ
|
| 24 |
def get_ffmpeg():
|
| 25 |
return shutil.which("ffmpeg") or "/usr/bin/ffmpeg"
|
| 26 |
|
| 27 |
+
def cleanup_old_files():
|
| 28 |
+
for f in ["downloaded_video.mp4", "downloaded_video.m4a",
|
| 29 |
+
"downloaded_video.webm", "downloaded_video.mp3",
|
| 30 |
+
"downloaded_video.wav", "downloaded_video.ogg",
|
| 31 |
+
"extracted_audio.wav", "temp_cookies.txt"]:
|
| 32 |
+
if os.path.exists(f):
|
| 33 |
+
try:
|
| 34 |
+
os.remove(f)
|
| 35 |
+
except Exception:
|
| 36 |
+
pass
|
| 37 |
+
|
| 38 |
+
def find_downloaded_file():
|
| 39 |
+
"""Find whatever yt-dlp actually downloaded."""
|
| 40 |
+
for ext in ["mp4", "m4a", "webm", "mp3", "wav", "ogg", "mkv", "flv"]:
|
| 41 |
+
path = f"downloaded_video.{ext}"
|
| 42 |
+
if os.path.exists(path):
|
| 43 |
+
return path
|
| 44 |
+
return None
|
| 45 |
+
|
| 46 |
+
# βββββββββββββββββββββββββββββββββββββββββββββββ
|
| 47 |
+
# COOKIES HELPER
|
| 48 |
+
# βββββββββββββββββββββββββββββββββββββββββββββββ
|
| 49 |
+
def save_cookies(cookies_str):
|
| 50 |
+
"""
|
| 51 |
+
Accept cookies in two formats:
|
| 52 |
+
1) Netscape format (starts with '# Netscape HTTP Cookie File')
|
| 53 |
+
2) key=value; key2=value2 (browser copy-paste format)
|
| 54 |
+
Returns path to cookie file, or None.
|
| 55 |
+
"""
|
| 56 |
+
if not cookies_str or not cookies_str.strip():
|
| 57 |
+
return None
|
| 58 |
+
|
| 59 |
+
cookies_file = "temp_cookies.txt"
|
| 60 |
+
raw = cookies_str.strip()
|
| 61 |
+
|
| 62 |
+
if raw.startswith("# Netscape"):
|
| 63 |
+
with open(cookies_file, "w") as f:
|
| 64 |
+
f.write(raw)
|
| 65 |
+
else:
|
| 66 |
+
# Convert simple "key=value; key2=value2" into Netscape format
|
| 67 |
+
lines = ["# Netscape HTTP Cookie File"]
|
| 68 |
+
for pair in raw.split(";"):
|
| 69 |
+
pair = pair.strip()
|
| 70 |
+
if "=" in pair:
|
| 71 |
+
k, v = pair.split("=", 1)
|
| 72 |
+
lines.append(f".instagram.com\tTRUE\t/\tFALSE\t9999999999\t{k.strip()}\t{v.strip()}")
|
| 73 |
+
with open(cookies_file, "w") as f:
|
| 74 |
+
f.write("\n".join(lines))
|
| 75 |
|
| 76 |
+
return cookies_file
|
|
|
|
| 77 |
|
| 78 |
+
# βββββββββββββββββββββββββββββββββββββββββββββββ
|
| 79 |
+
# DOWNLOAD (YouTube / Instagram / Others)
|
| 80 |
+
# βββββββββββββββββββββββββββββββββββββββββββββββ
|
| 81 |
+
def download_video(url, cookies_str=""):
|
| 82 |
+
cleanup_old_files()
|
| 83 |
+
|
| 84 |
+
is_youtube = any(x in url for x in ["youtube.com", "youtu.be"])
|
| 85 |
+
is_instagram = "instagram.com" in url
|
| 86 |
+
is_tiktok = "tiktok.com" in url
|
| 87 |
+
|
| 88 |
+
cookies_file = save_cookies(cookies_str)
|
| 89 |
+
|
| 90 |
+
# ββ Base options ββ
|
| 91 |
ydl_opts = {
|
| 92 |
+
"format" : "bestaudio/best",
|
| 93 |
+
"outtmpl" : "downloaded_video.%(ext)s",
|
| 94 |
+
"quiet" : True,
|
| 95 |
"nocheckcertificate": True,
|
| 96 |
+
"retries" : 5,
|
| 97 |
+
"fragment_retries" : 5,
|
| 98 |
+
"ignoreerrors" : False,
|
| 99 |
+
"http_headers" : {
|
| 100 |
+
"User-Agent": (
|
| 101 |
+
"Mozilla/5.0 (Linux; Android 12; Pixel 6) "
|
| 102 |
+
"AppleWebKit/537.36 (KHTML, like Gecko) "
|
| 103 |
+
"Chrome/112.0.0.0 Mobile Safari/537.36"
|
| 104 |
+
),
|
| 105 |
+
"Accept-Language": "en-US,en;q=0.9",
|
| 106 |
+
},
|
| 107 |
}
|
| 108 |
|
| 109 |
+
if cookies_file:
|
| 110 |
+
ydl_opts["cookiefile"] = cookies_file
|
| 111 |
+
|
| 112 |
+
# ββ YouTube-specific bypass (android player avoids bot detection) ββ
|
| 113 |
+
if is_youtube:
|
| 114 |
+
ydl_opts["extractor_args"] = {
|
| 115 |
+
"youtube": {
|
| 116 |
+
"player_client": ["android", "ios", "web"],
|
| 117 |
+
"player_skip" : ["webpage"],
|
| 118 |
+
}
|
| 119 |
+
}
|
| 120 |
+
ydl_opts["http_headers"]["User-Agent"] = (
|
| 121 |
+
"com.google.android.youtube/17.36.4 "
|
| 122 |
+
"(Linux; U; Android 12; GB) gzip"
|
| 123 |
+
)
|
| 124 |
+
|
| 125 |
+
# ββ Instagram-specific headers ββ
|
| 126 |
+
if is_instagram:
|
| 127 |
+
ydl_opts["http_headers"] = {
|
| 128 |
+
"User-Agent" : "Instagram 219.0.0.12.117 Android (31/12; 420dpi; 1080x2400; samsung; SM-G991B; o1s; exynos2100; en_US; 346877738)",
|
| 129 |
+
"Accept" : "*/*",
|
| 130 |
+
"Accept-Language" : "en-US",
|
| 131 |
+
"X-IG-App-ID" : "936619743392459",
|
| 132 |
+
}
|
| 133 |
+
if not cookies_file:
|
| 134 |
+
raise ValueError(
|
| 135 |
+
"instagram_no_cookies"
|
| 136 |
+
)
|
| 137 |
+
|
| 138 |
+
# ββ TikTok ββ
|
| 139 |
+
if is_tiktok:
|
| 140 |
+
ydl_opts["http_headers"]["User-Agent"] = (
|
| 141 |
+
"TikTok 26.2.0 rv:262018 (iPhone; iOS 14.4.2; en_US) Cronet"
|
| 142 |
+
)
|
| 143 |
+
|
| 144 |
with yt_dlp.YoutubeDL(ydl_opts) as ydl:
|
| 145 |
ydl.download([url])
|
| 146 |
|
| 147 |
+
path = find_downloaded_file()
|
| 148 |
+
if not path:
|
| 149 |
+
raise FileNotFoundError("Download completed but file not found.")
|
| 150 |
+
return path
|
| 151 |
|
| 152 |
+
# βββββββββββββββββββββββββββββββββββββββββββββββ
|
| 153 |
+
# AUDIO EXTRACTION
|
| 154 |
+
# βββββββββββββββββββββββββββββββββββββββββββββββ
|
| 155 |
+
def extract_audio(video_path):
|
| 156 |
audio_path = "extracted_audio.wav"
|
|
|
|
| 157 |
if os.path.exists(audio_path):
|
| 158 |
os.remove(audio_path)
|
| 159 |
|
| 160 |
subprocess.run(
|
| 161 |
[
|
| 162 |
+
get_ffmpeg(), "-y",
|
|
|
|
| 163 |
"-i", video_path,
|
| 164 |
"-vn",
|
| 165 |
"-ac", "1",
|
|
|
|
| 169 |
stdout=subprocess.DEVNULL,
|
| 170 |
stderr=subprocess.DEVNULL
|
| 171 |
)
|
|
|
|
| 172 |
return audio_path
|
| 173 |
|
| 174 |
+
# βββββββββββββββββββββββββββββββββββββββββββββββ
|
| 175 |
+
# SCRIPT NORMALIZER
|
| 176 |
+
# βββββββββββββββββββββββββββββββββββββββββββββββ
|
| 177 |
def normalize_script(text, lang):
|
| 178 |
if lang == "hi":
|
| 179 |
try:
|
| 180 |
return transliterate(text, sanscript.ARABIC, sanscript.DEVANAGARI)
|
| 181 |
+
except Exception:
|
| 182 |
return text
|
| 183 |
return text
|
| 184 |
|
| 185 |
+
# βββββββββββββββββββββββββββββββββββββββββββββββ
|
| 186 |
+
# MAIN TRANSCRIBE
|
| 187 |
+
# βββββββββββββββββββββββββββββββββββββββββββββββ
|
| 188 |
+
def transcribe(url, file, lang_choice, cookies_str=""):
|
| 189 |
try:
|
| 190 |
+
# ββ FILE MODE ββ
|
| 191 |
if file:
|
| 192 |
ext = os.path.splitext(file)[1].lower()
|
| 193 |
+
audio = file if ext in [".mp3", ".wav", ".m4a"] else extract_audio(file)
|
|
|
|
|
|
|
|
|
|
| 194 |
|
| 195 |
+
# ββ URL MODE ββ
|
| 196 |
+
elif url and url.strip():
|
| 197 |
+
try:
|
| 198 |
+
video = download_video(url.strip(), cookies_str)
|
| 199 |
+
except ValueError as ve:
|
| 200 |
+
if "instagram_no_cookies" in str(ve):
|
| 201 |
+
return (
|
| 202 |
+
"β Instagram requires login cookies to download.\n\n"
|
| 203 |
+
"π How to get cookies:\n"
|
| 204 |
+
"1. Open Instagram in Chrome/Firefox\n"
|
| 205 |
+
"2. Install 'Cookie Editor' browser extension\n"
|
| 206 |
+
"3. Click Export β Copy (Netscape format)\n"
|
| 207 |
+
"4. Paste in the π Cookies box below and try again.\n\n"
|
| 208 |
+
"Or simply download the video and upload the file directly."
|
| 209 |
+
)
|
| 210 |
+
raise
|
| 211 |
+
except Exception as e:
|
| 212 |
+
err = str(e).lower()
|
| 213 |
+
if "sign in" in err or "login" in err or "private" in err:
|
| 214 |
+
return (
|
| 215 |
+
"β This content requires login.\n\n"
|
| 216 |
+
"Paste your browser cookies in the π Cookies box and retry,\n"
|
| 217 |
+
"or download the video and upload it directly."
|
| 218 |
+
)
|
| 219 |
+
if "http error 429" in err or "too many" in err:
|
| 220 |
+
return "β Rate limited by the platform. Wait a few minutes and try again."
|
| 221 |
+
if "youtube" in err and ("bot" in err or "sign in" in err):
|
| 222 |
+
return (
|
| 223 |
+
"β YouTube bot detection triggered.\n\n"
|
| 224 |
+
"Paste your YouTube cookies in the π Cookies box and retry."
|
| 225 |
+
)
|
| 226 |
+
raise
|
| 227 |
+
|
| 228 |
+
audio = extract_audio(video)
|
| 229 |
|
| 230 |
else:
|
| 231 |
return "β οΈ Please paste a URL or upload a file."
|
| 232 |
|
| 233 |
+
# ββ Safety check ββ
|
| 234 |
+
if not os.path.exists(audio) or os.path.getsize(audio) < 5000:
|
| 235 |
+
return "β Audio extraction failed. File may be too short or corrupted."
|
| 236 |
|
| 237 |
+
# ββ Transcribe ββ
|
| 238 |
+
m = load_model()
|
| 239 |
language = None if lang_choice == "Auto Detect" else lang_choice
|
| 240 |
|
| 241 |
+
segments, info = m.transcribe(
|
| 242 |
audio,
|
| 243 |
beam_size=1,
|
| 244 |
vad_filter=True,
|
| 245 |
language=language
|
| 246 |
)
|
| 247 |
|
| 248 |
+
raw_text = " ".join(s.text for s in segments).strip()
|
| 249 |
final_text = normalize_script(raw_text, info.language)
|
| 250 |
|
| 251 |
+
return f"π Detected Language: {info.language.upper()}\n\n{final_text}"
|
| 252 |
|
| 253 |
except Exception as e:
|
| 254 |
+
err = str(e).lower()
|
| 255 |
+
if "instagram" in err:
|
| 256 |
+
return (
|
| 257 |
+
"β Instagram download failed.\n"
|
| 258 |
+
"Upload the video file directly, or paste cookies and retry."
|
| 259 |
+
)
|
| 260 |
+
if "ffmpeg" in err:
|
| 261 |
+
return "β FFmpeg error during audio extraction."
|
| 262 |
return f"β Error: {str(e)}"
|
| 263 |
|
| 264 |
+
# βββββββββββββββββββββββββββββββββββββββββββββββ
|
| 265 |
+
# PREMIUM UI
|
| 266 |
+
# βββββββββββββββββββββββββββββββββββββββββββββββ
|
| 267 |
+
|
| 268 |
+
CSS = """
|
| 269 |
+
/* ββ Imports ββ */
|
| 270 |
+
@import url('https://fonts.googleapis.com/css2?family=Orbitron:wght@700;900&family=Sora:wght@300;400;600;700&display=swap');
|
| 271 |
+
|
| 272 |
+
/* ββ Base ββ */
|
| 273 |
+
body, .gradio-container {
|
| 274 |
+
background: radial-gradient(ellipse at 20% 0%, #0d1b2a 0%, #0a0f1e 50%, #020408 100%) !important;
|
| 275 |
+
font-family: 'Sora', sans-serif !important;
|
| 276 |
+
}
|
| 277 |
+
footer { display: none !important; }
|
| 278 |
+
.gap { gap: 16px; }
|
| 279 |
+
|
| 280 |
+
/* ββ RGB Animations ββ */
|
| 281 |
+
@keyframes rgb-glow {
|
| 282 |
+
0% { color:#ff0080; text-shadow:0 0 14px #ff0080,0 0 28px #ff0080; }
|
| 283 |
+
16% { color:#ff8c00; text-shadow:0 0 14px #ff8c00,0 0 28px #ff8c00; }
|
| 284 |
+
33% { color:#ffe600; text-shadow:0 0 14px #ffe600,0 0 28px #ffe600; }
|
| 285 |
+
50% { color:#00ff88; text-shadow:0 0 14px #00ff88,0 0 28px #00ff88; }
|
| 286 |
+
66% { color:#00c8ff; text-shadow:0 0 14px #00c8ff,0 0 28px #00c8ff; }
|
| 287 |
+
83% { color:#a855f7; text-shadow:0 0 14px #a855f7,0 0 28px #a855f7; }
|
| 288 |
+
100% { color:#ff0080; text-shadow:0 0 14px #ff0080,0 0 28px #ff0080; }
|
| 289 |
+
}
|
| 290 |
+
@keyframes rgb-border {
|
| 291 |
+
0% { border-color:#ff0080; box-shadow:0 0 12px #ff0080,inset 0 0 8px rgba(255,0,128,.08); }
|
| 292 |
+
16% { border-color:#ff8c00; box-shadow:0 0 12px #ff8c00,inset 0 0 8px rgba(255,140,0,.08); }
|
| 293 |
+
33% { border-color:#ffe600; box-shadow:0 0 12px #ffe600,inset 0 0 8px rgba(255,230,0,.08); }
|
| 294 |
+
50% { border-color:#00ff88; box-shadow:0 0 12px #00ff88,inset 0 0 8px rgba(0,255,136,.08); }
|
| 295 |
+
66% { border-color:#00c8ff; box-shadow:0 0 12px #00c8ff,inset 0 0 8px rgba(0,200,255,.08); }
|
| 296 |
+
83% { border-color:#a855f7; box-shadow:0 0 12px #a855f7,inset 0 0 8px rgba(168,85,247,.08); }
|
| 297 |
+
100% { border-color:#ff0080; box-shadow:0 0 12px #ff0080,inset 0 0 8px rgba(255,0,128,.08); }
|
| 298 |
+
}
|
| 299 |
+
@keyframes gradient-bg {
|
| 300 |
+
0% { background-position:0% 50%; }
|
| 301 |
+
50% { background-position:100% 50%; }
|
| 302 |
+
100% { background-position:0% 50%; }
|
| 303 |
+
}
|
| 304 |
+
@keyframes scan {
|
| 305 |
+
0% { background-position:0 0; }
|
| 306 |
+
100% { background-position:0 100px; }
|
| 307 |
+
}
|
| 308 |
+
@keyframes pulse-badge {
|
| 309 |
+
0%,100% { transform:scale(1); box-shadow:0 0 18px rgba(0,200,255,.4); }
|
| 310 |
+
50% { transform:scale(1.03);box-shadow:0 0 36px rgba(168,85,247,.6); }
|
| 311 |
+
}
|
| 312 |
+
|
| 313 |
+
/* ββ Header ββ */
|
| 314 |
+
.hero-wrap {
|
| 315 |
+
text-align: center;
|
| 316 |
+
padding: 32px 20px 20px;
|
| 317 |
+
position: relative;
|
| 318 |
+
}
|
| 319 |
+
.hero-title {
|
| 320 |
+
font-family: 'Orbitron', sans-serif;
|
| 321 |
+
font-size: clamp(1.6em, 4vw, 2.6em);
|
| 322 |
+
font-weight: 900;
|
| 323 |
+
letter-spacing: 2px;
|
| 324 |
+
animation: rgb-glow 3s linear infinite;
|
| 325 |
+
margin: 0 0 8px;
|
| 326 |
+
}
|
| 327 |
+
.hero-sub {
|
| 328 |
+
font-family: 'Sora', sans-serif;
|
| 329 |
+
font-size: .95em;
|
| 330 |
+
color: #94a3b8;
|
| 331 |
+
margin: 0 0 18px;
|
| 332 |
+
letter-spacing: .5px;
|
| 333 |
+
}
|
| 334 |
+
.made-badge {
|
| 335 |
+
display: inline-block;
|
| 336 |
+
font-family: 'Orbitron', sans-serif;
|
| 337 |
+
font-size: .72em;
|
| 338 |
+
font-weight: 700;
|
| 339 |
+
letter-spacing: 4px;
|
| 340 |
+
padding: 7px 26px;
|
| 341 |
+
border: 2px solid #00c8ff;
|
| 342 |
+
border-radius: 50px;
|
| 343 |
+
animation: rgb-glow 3s linear infinite, rgb-border 3s linear infinite, pulse-badge 3s ease-in-out infinite;
|
| 344 |
+
text-transform: uppercase;
|
| 345 |
+
background: rgba(0,0,0,.15);
|
| 346 |
+
}
|
| 347 |
+
|
| 348 |
+
/* ββ Platform badges ββ */
|
| 349 |
+
.platforms {
|
| 350 |
+
display: flex;
|
| 351 |
+
justify-content: center;
|
| 352 |
+
gap: 10px;
|
| 353 |
+
flex-wrap: wrap;
|
| 354 |
+
margin: 14px 0 0;
|
| 355 |
+
}
|
| 356 |
+
.plat-tag {
|
| 357 |
+
font-size: .72em;
|
| 358 |
+
padding: 4px 14px;
|
| 359 |
+
border-radius: 50px;
|
| 360 |
+
border: 1px solid rgba(255,255,255,.15);
|
| 361 |
+
color: #cbd5e1;
|
| 362 |
+
background: rgba(255,255,255,.05);
|
| 363 |
+
letter-spacing: .5px;
|
| 364 |
+
}
|
| 365 |
+
.plat-ok { border-color: rgba(0,255,136,.35); color:#00ff88; }
|
| 366 |
+
.plat-warn{ border-color: rgba(255,190,0,.35); color:#fbbf24; }
|
| 367 |
+
|
| 368 |
+
/* ββ Glass card ββ */
|
| 369 |
+
.glass-card {
|
| 370 |
+
background: rgba(255,255,255,.04) !important;
|
| 371 |
+
backdrop-filter: blur(24px) !important;
|
| 372 |
+
border: 1px solid rgba(255,255,255,.08) !important;
|
| 373 |
+
border-radius: 20px !important;
|
| 374 |
+
box-shadow: 0 24px 60px rgba(0,0,0,.5) !important;
|
| 375 |
+
}
|
| 376 |
+
|
| 377 |
+
/* ββ Divider ββ */
|
| 378 |
+
.rgb-line {
|
| 379 |
+
height: 2px;
|
| 380 |
+
margin: 28px 0;
|
| 381 |
+
background: linear-gradient(90deg,transparent,#00c8ff,#a855f7,#ff0080,transparent);
|
| 382 |
+
background-size: 200% 200%;
|
| 383 |
+
animation: gradient-bg 4s ease infinite;
|
| 384 |
+
border-radius: 2px;
|
| 385 |
+
}
|
| 386 |
+
|
| 387 |
+
/* ββ Tabs ββ */
|
| 388 |
+
.tabs .tab-nav button {
|
| 389 |
+
font-family: 'Sora', sans-serif !important;
|
| 390 |
+
font-weight: 600 !important;
|
| 391 |
+
font-size: .9em !important;
|
| 392 |
+
color: #64748b !important;
|
| 393 |
+
border-radius: 10px 10px 0 0 !important;
|
| 394 |
+
padding: 10px 20px !important;
|
| 395 |
+
transition: all .25s !important;
|
| 396 |
+
}
|
| 397 |
+
.tabs .tab-nav button.selected {
|
| 398 |
+
color: #00c8ff !important;
|
| 399 |
+
border-bottom: 2px solid #00c8ff !important;
|
| 400 |
+
background: rgba(0,200,255,.07) !important;
|
| 401 |
+
}
|
| 402 |
+
|
| 403 |
+
/* ββ Inputs ββ */
|
| 404 |
+
input, textarea, .gr-input, .gr-textarea, .codemirror-wrapper {
|
| 405 |
+
background: rgba(255,255,255,.06) !important;
|
| 406 |
+
border: 1px solid rgba(255,255,255,.1) !important;
|
| 407 |
+
color: #e2e8f0 !important;
|
| 408 |
+
border-radius: 12px !important;
|
| 409 |
+
font-family: 'Sora', sans-serif !important;
|
| 410 |
+
}
|
| 411 |
+
input:focus, textarea:focus {
|
| 412 |
+
border-color: #00c8ff !important;
|
| 413 |
+
box-shadow: 0 0 12px rgba(0,200,255,.2) !important;
|
| 414 |
+
outline: none !important;
|
| 415 |
+
}
|
| 416 |
+
label, .label-wrap span {
|
| 417 |
+
color: #94a3b8 !important;
|
| 418 |
+
font-family: 'Sora', sans-serif !important;
|
| 419 |
+
font-size: .85em !important;
|
| 420 |
+
letter-spacing: .3px !important;
|
| 421 |
+
}
|
| 422 |
+
|
| 423 |
+
/* ββ Primary button ββ */
|
| 424 |
+
button.primary, .gr-button-primary {
|
| 425 |
+
background: linear-gradient(135deg,#0ea5e9,#2563eb) !important;
|
| 426 |
+
border: none !important;
|
| 427 |
+
color: #fff !important;
|
| 428 |
+
font-family: 'Sora', sans-serif !important;
|
| 429 |
+
font-weight: 700 !important;
|
| 430 |
+
border-radius: 12px !important;
|
| 431 |
+
padding: 12px 20px !important;
|
| 432 |
+
font-size: 1em !important;
|
| 433 |
+
box-shadow: 0 4px 18px rgba(14,165,233,.35) !important;
|
| 434 |
+
transition: all .3s ease !important;
|
| 435 |
+
}
|
| 436 |
+
button.primary:hover {
|
| 437 |
+
background: linear-gradient(135deg,#38bdf8,#1d4ed8) !important;
|
| 438 |
+
box-shadow: 0 6px 28px rgba(14,165,233,.55) !important;
|
| 439 |
+
transform: translateY(-1px) !important;
|
| 440 |
+
}
|
| 441 |
+
|
| 442 |
+
/* ββ Output code block ββ */
|
| 443 |
+
.codemirror-wrapper {
|
| 444 |
+
background: rgba(0,0,0,.4) !important;
|
| 445 |
+
border: 1px solid rgba(0,200,255,.2) !important;
|
| 446 |
+
}
|
| 447 |
+
.CodeMirror {
|
| 448 |
+
background: transparent !important;
|
| 449 |
+
color: #a5f3fc !important;
|
| 450 |
+
font-family: 'Sora', sans-serif !important;
|
| 451 |
+
}
|
| 452 |
+
|
| 453 |
+
/* ββ Accordion (cookies) ββ */
|
| 454 |
+
.accordion-header button {
|
| 455 |
+
background: rgba(255,255,255,.04) !important;
|
| 456 |
+
border: 1px solid rgba(255,200,0,.2) !important;
|
| 457 |
+
color: #fbbf24 !important;
|
| 458 |
+
border-radius: 12px !important;
|
| 459 |
+
font-family: 'Sora', sans-serif !important;
|
| 460 |
+
font-weight: 600 !important;
|
| 461 |
+
}
|
| 462 |
+
|
| 463 |
+
/* ββ Warning note ββ */
|
| 464 |
+
.note-warn {
|
| 465 |
+
background: rgba(251,191,36,.07);
|
| 466 |
+
border: 1px solid rgba(251,191,36,.25);
|
| 467 |
+
border-radius: 12px;
|
| 468 |
+
padding: 12px 18px;
|
| 469 |
+
color: #fbbf24;
|
| 470 |
+
font-size: .82em;
|
| 471 |
+
line-height: 1.6;
|
| 472 |
+
margin-top: 6px;
|
| 473 |
+
}
|
| 474 |
+
|
| 475 |
+
/* ββ Dropdown ββ */
|
| 476 |
+
.gr-dropdown select, select {
|
| 477 |
+
background: rgba(255,255,255,.06) !important;
|
| 478 |
+
color: #e2e8f0 !important;
|
| 479 |
+
border: 1px solid rgba(255,255,255,.1) !important;
|
| 480 |
+
border-radius: 10px !important;
|
| 481 |
+
}
|
| 482 |
+
|
| 483 |
+
/* ββ Scrollbar ββ */
|
| 484 |
+
::-webkit-scrollbar { width:6px; }
|
| 485 |
+
::-webkit-scrollbar-track { background:transparent; }
|
| 486 |
+
::-webkit-scrollbar-thumb { background:#1e40af; border-radius:4px; }
|
| 487 |
"""
|
| 488 |
|
| 489 |
+
HEADER_HTML = """
|
| 490 |
+
<div class="hero-wrap">
|
| 491 |
+
<div class="hero-title">ποΈ Universal Transcript Tool</div>
|
| 492 |
+
<div class="hero-sub">Speech to text β any platform, any language</div>
|
| 493 |
+
<div class="made-badge">β¦ Made by Deepu β¦</div>
|
| 494 |
+
<div class="platforms">
|
| 495 |
+
<span class="plat-tag plat-ok">β YouTube</span>
|
| 496 |
+
<span class="plat-tag plat-ok">β TikTok</span>
|
| 497 |
+
<span class="plat-tag plat-ok">β Facebook</span>
|
| 498 |
+
<span class="plat-tag plat-ok">β Twitter / X</span>
|
| 499 |
+
<span class="plat-tag plat-warn">β Instagram (needs cookies)</span>
|
| 500 |
+
<span class="plat-tag plat-ok">β File Upload</span>
|
| 501 |
+
</div>
|
| 502 |
+
</div>
|
| 503 |
+
"""
|
| 504 |
+
|
| 505 |
+
COOKIES_HELP = """
|
| 506 |
+
<div class="note-warn">
|
| 507 |
+
<strong>π When do you need cookies?</strong><br>
|
| 508 |
+
Instagram always Β· Private YouTube Β· Age-restricted content<br><br>
|
| 509 |
+
<strong>How to get cookies (2 steps):</strong><br>
|
| 510 |
+
1. Install <em>Cookie Editor</em> extension in Chrome / Firefox<br>
|
| 511 |
+
2. Open the site β click extension β <strong>Export β Netscape format</strong> β paste here
|
| 512 |
+
</div>
|
| 513 |
+
"""
|
| 514 |
+
|
| 515 |
+
# βββββββββββββββββββββββββββββββββββββββββββββββ
|
| 516 |
+
# BUILD UI
|
| 517 |
+
# βββββββββββββββββββββββββββββββββββββββββββββββ
|
| 518 |
+
with gr.Blocks(css=CSS, theme=gr.themes.Base()) as demo:
|
| 519 |
|
| 520 |
+
gr.HTML(HEADER_HTML)
|
| 521 |
+
gr.HTML('<div class="rgb-line"></div>')
|
| 522 |
+
|
| 523 |
+
with gr.Column(elem_classes="glass-card", scale=1):
|
| 524 |
+
|
| 525 |
+
with gr.Tabs(elem_classes="tabs"):
|
| 526 |
+
|
| 527 |
+
# ββ Tab 1: URL ββ
|
| 528 |
with gr.TabItem("π Paste Link"):
|
| 529 |
+
url_input = gr.Textbox(
|
| 530 |
+
label="Video URL",
|
| 531 |
+
placeholder="https://youtube.com/watch?v=... or TikTok / Facebook / Twitter link",
|
| 532 |
+
lines=1
|
| 533 |
+
)
|
| 534 |
btn_url = gr.Button("π§ Transcribe Link", variant="primary")
|
| 535 |
|
| 536 |
+
# ββ Tab 2: Upload ββ
|
| 537 |
with gr.TabItem("π Upload File"):
|
| 538 |
+
file_input = gr.File(
|
| 539 |
label="Upload Video / Audio",
|
| 540 |
+
file_types=[".mp4", ".mkv", ".mov", ".webm", ".avi",
|
| 541 |
+
".mp3", ".wav", ".m4a"]
|
| 542 |
)
|
| 543 |
btn_file = gr.Button("π Transcribe File", variant="primary")
|
| 544 |
|
| 545 |
+
gr.HTML('<div class="rgb-line"></div>')
|
| 546 |
+
|
| 547 |
+
# ββ Language ββ
|
| 548 |
lang = gr.Dropdown(
|
| 549 |
label="π Transcript Language",
|
| 550 |
choices=[
|
| 551 |
+
"Auto Detect", "hi", "ur", "en", "ar",
|
| 552 |
+
"fr", "de", "es", "ru", "ja", "zh"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 553 |
],
|
| 554 |
value="Auto Detect"
|
| 555 |
)
|
| 556 |
|
| 557 |
+
# ββ Cookies (Accordion) ββ
|
| 558 |
+
with gr.Accordion("π Cookies β Instagram / Private Content (click to expand)", open=False):
|
| 559 |
+
gr.HTML(COOKIES_HELP)
|
| 560 |
+
cookies_input = gr.Textbox(
|
| 561 |
+
label="Paste Netscape Cookies Here",
|
| 562 |
+
placeholder="# Netscape HTTP Cookie File\n.instagram.com TRUE / FALSE 9999999999 sessionid XXXXX",
|
| 563 |
+
lines=6
|
| 564 |
+
)
|
| 565 |
|
| 566 |
+
gr.HTML('<div class="rgb-line"></div>')
|
| 567 |
+
|
| 568 |
+
# ββ Output ββ
|
| 569 |
+
output = gr.Code(label="π Transcript Output", lines=14, language=None)
|
| 570 |
+
|
| 571 |
+
# ββ Events ββ
|
| 572 |
+
btn_url.click(
|
| 573 |
+
fn=transcribe,
|
| 574 |
+
inputs=[url_input, gr.State(None), lang, cookies_input],
|
| 575 |
+
outputs=output
|
| 576 |
+
)
|
| 577 |
+
btn_file.click(
|
| 578 |
+
fn=transcribe,
|
| 579 |
+
inputs=[gr.State(None), file_input, lang, cookies_input],
|
| 580 |
+
outputs=output
|
| 581 |
+
)
|
| 582 |
|
| 583 |
+
demo.launch()
|