Update app.py

app.py

@@ -1,21 +1,24 @@
 import torch
-
 import gradio as gr
 import yt_dlp as youtube_dl
 from transformers import pipeline
 from transformers.pipelines.audio_utils import ffmpeg_read
+from urllib.parse import urlparse, parse_qs
 
 import tempfile
 import time
 import os
 
+# Constants
 MODEL_NAME = "dataprizma/whisper-large-v3-turbo"
 BATCH_SIZE = 8
 FILE_LIMIT_MB = 1000
-YT_LENGTH_LIMIT_S = 3600  # limit to 1 hour YouTube files
+YT_LENGTH_LIMIT_S = 3600  # 1 hour limit
 
+# Device selection
 device = 0 if torch.cuda.is_available() else "cpu"
 
+# Load Whisper pipeline
 pipe = pipeline(
     task="automatic-speech-recognition",
     model=MODEL_NAME,
@@ -23,89 +26,84 @@ pipe = pipeline(
     device=device,
 )
 
+# Extract YouTube Video ID
+def _extract_yt_video_id(yt_url):
+    parsed_url = urlparse(yt_url)
+    return parse_qs(parsed_url.query).get("v", [""])[0]
+
+# Embed YouTube Video in HTML
+def _return_yt_html_embed(yt_url):
+    video_id = _extract_yt_video_id(yt_url)
+    if not video_id:
+        raise gr.Error("Invalid YouTube URL. Please check and try again.")
+    return f'<center> <iframe width="500" height="320" src="https://www.youtube.com/embed/{video_id}"></iframe> </center>'
 
+# Transcription function
 def transcribe(inputs, task):
     if inputs is None:
         raise gr.Error("No audio file submitted! Please upload or record an audio file before submitting your request.")
 
-    text = pipe(inputs, batch_size=BATCH_SIZE, generate_kwargs={"task": task}, return_timestamps=True)["text"]
-    return  text
-
-
-def _return_yt_html_embed(yt_url):
-    video_id = yt_url.split("?v=")[-1]
-    HTML_str = (
-        f'<center> <iframe width="500" height="320" src="https://www.youtube.com/embed/{video_id}"> </iframe>'
-        " </center>"
-    )
-    return HTML_str
+    text = pipe(
+        {"input_features": inputs}, 
+        batch_size=BATCH_SIZE, 
+        generate_kwargs={"task": task, "forced_decoder_ids": None}, 
+        return_timestamps=True
+    )["text"]
+    
+    return text
 
+# Download YouTube audio
 def download_yt_audio(yt_url, filename):
-    info_loader = youtube_dl.YoutubeDL()
-    
-    try:
-        info = info_loader.extract_info(yt_url, download=False)
-    except youtube_dl.utils.DownloadError as err:
-        raise gr.Error(str(err))
-    
-    file_length = info["duration_string"]
-    file_h_m_s = file_length.split(":")
-    file_h_m_s = [int(sub_length) for sub_length in file_h_m_s]
-    
-    if len(file_h_m_s) == 1:
-        file_h_m_s.insert(0, 0)
-    if len(file_h_m_s) == 2:
-        file_h_m_s.insert(0, 0)
-    file_length_s = file_h_m_s[0] * 3600 + file_h_m_s[1] * 60 + file_h_m_s[2]
-    
-    if file_length_s > YT_LENGTH_LIMIT_S:
-        yt_length_limit_hms = time.strftime("%HH:%MM:%SS", time.gmtime(YT_LENGTH_LIMIT_S))
-        file_length_hms = time.strftime("%HH:%MM:%SS", time.gmtime(file_length_s))
-        raise gr.Error(f"Maximum YouTube length is {yt_length_limit_hms}, got {file_length_hms} YouTube video.")
-    
-    ydl_opts = {"outtmpl": filename, "format": "worstvideo[ext=mp4]+bestaudio[ext=m4a]/best[ext=mp4]/best"}
+    ydl_opts = {
+        "format": "bestaudio/best",
+        "outtmpl": filename,
+        "postprocessors": [
+            {"key": "FFmpegExtractAudio", "preferredcodec": "mp3", "preferredquality": "192"}
+        ],
+    }
     
     with youtube_dl.YoutubeDL(ydl_opts) as ydl:
         try:
+            info = ydl.extract_info(yt_url, download=False)
+            file_length_s = info.get("duration", 0)  # Duration in seconds
+            if file_length_s > YT_LENGTH_LIMIT_S:
+                raise gr.Error(f"Maximum YouTube length is 1 hour. Your video is {file_length_s // 3600}h {file_length_s % 3600 // 60}m {file_length_s % 60}s.")
             ydl.download([yt_url])
-        except youtube_dl.utils.ExtractorError as err:
+        except youtube_dl.utils.DownloadError as err:
             raise gr.Error(str(err))
 
-
+# YouTube transcription function
 def yt_transcribe(yt_url, task, max_filesize=75.0):
     html_embed_str = _return_yt_html_embed(yt_url)
 
     with tempfile.TemporaryDirectory() as tmpdirname:
-        filepath = os.path.join(tmpdirname, "video.mp4")
+        filepath = os.path.join(tmpdirname, "audio.mp3")
         download_yt_audio(yt_url, filepath)
+
+        if os.path.getsize(filepath) > max_filesize * 1024 * 1024:
+            raise gr.Error(f"File too large! Max allowed size is {max_filesize}MB.")
+
         with open(filepath, "rb") as f:
-            inputs = f.read()
+            inputs = ffmpeg_read(f.read(), pipe.feature_extractor.sampling_rate)
 
-    inputs = ffmpeg_read(inputs, pipe.feature_extractor.sampling_rate)
-    inputs = {"array": inputs, "sampling_rate": pipe.feature_extractor.sampling_rate}
+    inputs = {
+        "array": inputs,
+        "sampling_rate": pipe.feature_extractor.sampling_rate,
+        "attention_mask": torch.ones(len(inputs), dtype=torch.long),
+    }
 
-    text = pipe(inputs, batch_size=BATCH_SIZE, generate_kwargs={"task": task}, return_timestamps=True)["text"]
+    text = pipe(
+        {"input_features": inputs}, 
+        batch_size=BATCH_SIZE, 
+        generate_kwargs={"task": task, "forced_decoder_ids": None}, 
+        return_timestamps=True
+    )["text"]
 
     return html_embed_str, text
 
-
+# Gradio UI
 demo = gr.Blocks()
 
-mf_transcribe = gr.Interface(
-    fn=transcribe,
-    inputs=[
-        gr.Audio(type="filepath"),
-        gr.Radio(["transcribe", "translate"], label="Task"),
-    ],
-    outputs="text",
-    theme="huggingface",
-    title="Whisper Large V3: Transcribe Audio",
-    description=(
-        "Whisper Large V3 fine-tuned for Uzbek language by Dataprizma"
-    ),
-    allow_flagging="never",
-)
-
 file_transcribe = gr.Interface(
     fn=transcribe,
     inputs=[
@@ -113,30 +111,24 @@ file_transcribe = gr.Interface(
         gr.Radio(["transcribe", "translate"], label="Task"),
     ],
     outputs="text",
-    theme="huggingface",
     title="Whisper Large V3: Transcribe Audio",
-    description=(
-        "Whisper Large V3 fine-tuned for Uzbek language by Dataprizma"
-    ),
-    allow_flagging="never",
+    description="Whisper Large V3 fine-tuned for Uzbek language by Dataprizma",
+    flagging_mode="never",
 )
 
 yt_transcribe = gr.Interface(
     fn=yt_transcribe,
     inputs=[
-        gr.Textbox(lines=1, placeholder="Paste the URL to a YouTube video here", label="YouTube URL"),
+        gr.Textbox(lines=1, placeholder="Paste YouTube URL here", label="YouTube URL"),
         gr.Radio(["transcribe", "translate"], label="Task")
     ],
     outputs=["html", "text"],
-    theme="huggingface",
     title="Whisper Large V3: Transcribe YouTube",
-    description=(
-        "Whisper Large V3 fine-tuned for Uzbek language by Dataprizma"
-    ),
-    allow_flagging="never",
+    description="Whisper Large V3 fine-tuned for Uzbek language by Dataprizma",
+    flagging_mode="never",
 )
 
 with demo:
-    gr.TabbedInterface([mf_transcribe, file_transcribe, yt_transcribe], ["Microphone", "Audio file", "YouTube"])
+    gr.TabbedInterface([file_transcribe, yt_transcribe], ["Audio file", "YouTube"])
 
 demo.launch()