Spaces:
Running
Running
Update app.py
Browse files
app.py
CHANGED
|
@@ -4,30 +4,38 @@ import os
|
|
| 4 |
import shutil
|
| 5 |
from faster_whisper import WhisperModel
|
| 6 |
|
| 7 |
-
# --- 1. Model Setup
|
| 8 |
model = None
|
| 9 |
|
| 10 |
def load_model():
|
| 11 |
global model
|
| 12 |
if model is None:
|
| 13 |
-
print("π₯ Loading Whisper Model
|
| 14 |
-
# 'base' model with int8 quantization for speed/accuracy balance
|
| 15 |
model = WhisperModel("base", device="cpu", compute_type="int8")
|
| 16 |
print("β
Model Loaded!")
|
| 17 |
return model
|
| 18 |
|
| 19 |
-
# --- 2.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 20 |
def download_audio_from_url(url):
|
| 21 |
output_path = "downloaded_audio"
|
| 22 |
-
# Cleanup old files
|
| 23 |
if os.path.exists(f"{output_path}.mp3"): os.remove(f"{output_path}.mp3")
|
| 24 |
|
| 25 |
-
|
|
|
|
|
|
|
| 26 |
|
| 27 |
ydl_opts = {
|
| 28 |
'format': 'bestaudio/best',
|
| 29 |
'outtmpl': output_path,
|
| 30 |
-
'ffmpeg_location': ffmpeg_dir,
|
| 31 |
'postprocessors': [{
|
| 32 |
'key': 'FFmpegExtractAudio',
|
| 33 |
'preferredcodec': 'mp3',
|
|
@@ -47,12 +55,11 @@ def download_audio_from_url(url):
|
|
| 47 |
ydl.download([url])
|
| 48 |
return f"{output_path}.mp3"
|
| 49 |
except Exception as e:
|
| 50 |
-
raise Exception(f"
|
| 51 |
|
| 52 |
-
# ---
|
| 53 |
def transcribe_media(url_input, file_input):
|
| 54 |
|
| 55 |
-
# Decide source: Priority given to File if both exist, else URL
|
| 56 |
audio_file_path = None
|
| 57 |
|
| 58 |
try:
|
|
@@ -67,15 +74,15 @@ def transcribe_media(url_input, file_input):
|
|
| 67 |
audio_file_path = download_audio_from_url(url_input)
|
| 68 |
|
| 69 |
else:
|
| 70 |
-
return "β οΈ Error:
|
| 71 |
|
| 72 |
-
# --- Transcribe ---
|
| 73 |
if not os.path.exists(audio_file_path):
|
| 74 |
-
return "β Error: File
|
| 75 |
|
|
|
|
| 76 |
current_model = load_model()
|
| 77 |
|
| 78 |
-
# Turbo Settings: beam_size=1 (Fast)
|
| 79 |
segments, _ = current_model.transcribe(
|
| 80 |
audio_file_path,
|
| 81 |
beam_size=1,
|
|
@@ -88,21 +95,19 @@ def transcribe_media(url_input, file_input):
|
|
| 88 |
except Exception as e:
|
| 89 |
return f"β Error: {str(e)}"
|
| 90 |
|
| 91 |
-
# ---
|
| 92 |
css = """
|
| 93 |
.container {max-width: 900px; margin: auto;}
|
| 94 |
.gr-button-primary {background: linear-gradient(90deg, #1CB5E0 0%, #000851 100%); border: none; color: white;}
|
| 95 |
-
.tab-nav {font-weight: bold; font-size: 1.1em;}
|
| 96 |
"""
|
| 97 |
|
| 98 |
with gr.Blocks(theme=gr.themes.Soft(), css=css) as demo:
|
| 99 |
|
| 100 |
with gr.Column(elem_classes="container"):
|
| 101 |
gr.Markdown("# π Turbo Transcriber (Link & Upload)")
|
| 102 |
-
gr.Markdown("Paste a TikTok link **OR** upload an Audio/Video file
|
| 103 |
|
| 104 |
with gr.Tabs():
|
| 105 |
-
|
| 106 |
# TAB 1: Link
|
| 107 |
with gr.TabItem("π Paste Link"):
|
| 108 |
url_in = gr.Textbox(label="TikTok / YouTube URL", placeholder="https://...")
|
|
@@ -110,24 +115,13 @@ with gr.Blocks(theme=gr.themes.Soft(), css=css) as demo:
|
|
| 110 |
|
| 111 |
# TAB 2: File Upload
|
| 112 |
with gr.TabItem("π Upload File"):
|
| 113 |
-
file_in = gr.Audio(label="Upload
|
| 114 |
btn_file = gr.Button("π Transcribe File", variant="primary")
|
| 115 |
|
| 116 |
-
# Output Area (Common for both)
|
| 117 |
transcript_out = gr.Code(label="Transcript Result", language="markdown", interactive=False, lines=15)
|
| 118 |
|
| 119 |
-
#
|
| 120 |
-
|
| 121 |
-
|
| 122 |
-
fn=transcribe_media,
|
| 123 |
-
inputs=[url_in, gr.State(None)], # Link diya, File None
|
| 124 |
-
outputs=transcript_out
|
| 125 |
-
)
|
| 126 |
-
|
| 127 |
-
btn_file.click(
|
| 128 |
-
fn=transcribe_media,
|
| 129 |
-
inputs=[gr.State(None), file_in], # Link None, File diya
|
| 130 |
-
outputs=transcript_out
|
| 131 |
-
)
|
| 132 |
|
| 133 |
demo.launch()
|
|
|
|
| 4 |
import shutil
|
| 5 |
from faster_whisper import WhisperModel
|
| 6 |
|
| 7 |
+
# --- 1. Model Setup ---
|
| 8 |
model = None
|
| 9 |
|
| 10 |
def load_model():
|
| 11 |
global model
|
| 12 |
if model is None:
|
| 13 |
+
print("π₯ Loading Whisper Model...")
|
|
|
|
| 14 |
model = WhisperModel("base", device="cpu", compute_type="int8")
|
| 15 |
print("β
Model Loaded!")
|
| 16 |
return model
|
| 17 |
|
| 18 |
+
# --- 2. Helper: Find FFmpeg ---
|
| 19 |
+
def get_ffmpeg_dir():
|
| 20 |
+
# System me ffmpeg kahan hai, ye pata lagao
|
| 21 |
+
path = shutil.which("ffmpeg")
|
| 22 |
+
if path:
|
| 23 |
+
return os.path.dirname(path) # Folder ka rasta return karo
|
| 24 |
+
return "/usr/bin" # Default fallback
|
| 25 |
+
|
| 26 |
+
# --- 3. Logic: Download Audio from URL ---
|
| 27 |
def download_audio_from_url(url):
|
| 28 |
output_path = "downloaded_audio"
|
|
|
|
| 29 |
if os.path.exists(f"{output_path}.mp3"): os.remove(f"{output_path}.mp3")
|
| 30 |
|
| 31 |
+
# Dynamic FFmpeg Path (Ye error fix karega)
|
| 32 |
+
ffmpeg_dir = get_ffmpeg_dir()
|
| 33 |
+
print(f"π§ FFmpeg found at: {ffmpeg_dir}")
|
| 34 |
|
| 35 |
ydl_opts = {
|
| 36 |
'format': 'bestaudio/best',
|
| 37 |
'outtmpl': output_path,
|
| 38 |
+
'ffmpeg_location': ffmpeg_dir, # <--- FIXED
|
| 39 |
'postprocessors': [{
|
| 40 |
'key': 'FFmpegExtractAudio',
|
| 41 |
'preferredcodec': 'mp3',
|
|
|
|
| 55 |
ydl.download([url])
|
| 56 |
return f"{output_path}.mp3"
|
| 57 |
except Exception as e:
|
| 58 |
+
raise Exception(f"Download Fail: {str(e)}")
|
| 59 |
|
| 60 |
+
# --- 4. Main Transcribe Function (Handles Both) ---
|
| 61 |
def transcribe_media(url_input, file_input):
|
| 62 |
|
|
|
|
| 63 |
audio_file_path = None
|
| 64 |
|
| 65 |
try:
|
|
|
|
| 74 |
audio_file_path = download_audio_from_url(url_input)
|
| 75 |
|
| 76 |
else:
|
| 77 |
+
return "β οΈ Error: Link daalein ya File upload karein."
|
| 78 |
|
|
|
|
| 79 |
if not os.path.exists(audio_file_path):
|
| 80 |
+
return "β Error: File nahi mili."
|
| 81 |
|
| 82 |
+
# --- Transcribe ---
|
| 83 |
current_model = load_model()
|
| 84 |
|
| 85 |
+
# Turbo Settings: beam_size=1 (Fast)
|
| 86 |
segments, _ = current_model.transcribe(
|
| 87 |
audio_file_path,
|
| 88 |
beam_size=1,
|
|
|
|
| 95 |
except Exception as e:
|
| 96 |
return f"β Error: {str(e)}"
|
| 97 |
|
| 98 |
+
# --- 5. UI ---
|
| 99 |
css = """
|
| 100 |
.container {max-width: 900px; margin: auto;}
|
| 101 |
.gr-button-primary {background: linear-gradient(90deg, #1CB5E0 0%, #000851 100%); border: none; color: white;}
|
|
|
|
| 102 |
"""
|
| 103 |
|
| 104 |
with gr.Blocks(theme=gr.themes.Soft(), css=css) as demo:
|
| 105 |
|
| 106 |
with gr.Column(elem_classes="container"):
|
| 107 |
gr.Markdown("# π Turbo Transcriber (Link & Upload)")
|
| 108 |
+
gr.Markdown("Paste a TikTok link **OR** upload an Audio/Video file.")
|
| 109 |
|
| 110 |
with gr.Tabs():
|
|
|
|
| 111 |
# TAB 1: Link
|
| 112 |
with gr.TabItem("π Paste Link"):
|
| 113 |
url_in = gr.Textbox(label="TikTok / YouTube URL", placeholder="https://...")
|
|
|
|
| 115 |
|
| 116 |
# TAB 2: File Upload
|
| 117 |
with gr.TabItem("π Upload File"):
|
| 118 |
+
file_in = gr.Audio(label="Upload File", type="filepath", sources=["upload", "microphone"])
|
| 119 |
btn_file = gr.Button("π Transcribe File", variant="primary")
|
| 120 |
|
|
|
|
| 121 |
transcript_out = gr.Code(label="Transcript Result", language="markdown", interactive=False, lines=15)
|
| 122 |
|
| 123 |
+
# Actions
|
| 124 |
+
btn_url.click(fn=transcribe_media, inputs=[url_in, gr.State(None)], outputs=transcript_out)
|
| 125 |
+
btn_file.click(fn=transcribe_media, inputs=[gr.State(None), file_in], outputs=transcript_out)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 126 |
|
| 127 |
demo.launch()
|