Update app.py

app.py

@@ -1,37 +1,50 @@
-import spaces
-import torch
-
 import gradio as gr
 import yt_dlp as youtube_dl
-from transformers import pipeline
-from transformers.pipelines.audio_utils import ffmpeg_read
-
+import whisperx
 import tempfile
 import os
+import torch
+import gc
 
-MODEL_NAME = "openai/whisper-large-v3-turbo"
-BATCH_SIZE = 8
-FILE_LIMIT_MB = 1000
-YT_LENGTH_LIMIT_S = 3600  # limit to 1 hour YouTube files
-
-device = 0 if torch.cuda.is_available() else "cpu"
-
-pipe = pipeline(
-    task="automatic-speech-recognition",
-    model=MODEL_NAME,
-    chunk_length_s=30,
-    device=device,
-)
+# WhisperX配置
+device = "cuda" if torch.cuda.is_available() else "cpu"
+batch_size = 4
+compute_type = "float32"
+MODEL_NAME = "large-v3"
+YT_LENGTH_LIMIT_S = 3600  # 1 hour YouTube files
 
+# 加载WhisperX模型
+model = whisperx.load_model(MODEL_NAME, device=device, compute_type=compute_type)
 
-@spaces.GPU
 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.")
+    
+    # 加载和转录音频
+    audio = whisperx.load_audio(inputs)
+    result = model.transcribe(audio, batch_size=batch_size)
+    print(result["segments"])  # 未对齐的文本片段
+
+    # 释放资源以节省GPU内存
+    gc.collect()
+    torch.cuda.empty_cache()
+    del model
+    
+    # 加载对齐模型
+    model_a, metadata = whisperx.load_align_model(language_code=result["language"], device=device)
+    result = whisperx.align(result["segments"], model_a, metadata, audio, device, return_char_alignments=False)
 
-    text = pipe(inputs, batch_size=BATCH_SIZE, generate_kwargs={"task": task}, return_timestamps=True)["text"]
-    return  text
-
+    # 说话人分离
+    diarize_model = whisperx.DiarizationPipeline(use_auth_token="your_huggingface_token", device=device)
+    result = whisperx.assign_word_speakers(diarize_model, result)
+    
+    # 格式化输出
+    transcript = ""
+    for segment in result['segments']:
+        speaker = segment.get('speaker', 'Unknown')
+        transcript += f"{speaker}: {segment['text']}\n"
+    
+    return transcript
 
 def _return_yt_html_embed(yt_url):
     video_id = yt_url.split("?v=")[-1]
@@ -49,22 +62,11 @@ def download_yt_audio(yt_url, filename):
     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]
+    file_length = info["duration"]
+    if file_length > YT_LENGTH_LIMIT_S:
+        raise gr.Error("YouTube video length exceeds the 1-hour limit.")
     
-    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 = {"outtmpl": filename, "format": "bestaudio[ext=m4a]"}
     
     with youtube_dl.YoutubeDL(ydl_opts) as ydl:
         try:
@@ -72,75 +74,27 @@ def download_yt_audio(yt_url, filename):
         except youtube_dl.utils.ExtractorError as err:
             raise gr.Error(str(err))
 
-@spaces.GPU
-def yt_transcribe(yt_url, task, max_filesize=75.0):
+def yt_transcribe(yt_url, task):
     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, "video.m4a")
         download_yt_audio(yt_url, filepath)
-        with open(filepath, "rb") as f:
-            inputs = f.read()
-
-    inputs = ffmpeg_read(inputs, pipe.feature_extractor.sampling_rate)
-    inputs = {"array": inputs, "sampling_rate": pipe.feature_extractor.sampling_rate}
-
-    text = pipe(inputs, batch_size=BATCH_SIZE, generate_kwargs={"task": task}, return_timestamps=True)["text"]
+        result = transcribe(filepath, task)
 
-    return html_embed_str, text
-
-
-demo = gr.Blocks(theme=gr.themes.Ocean())
-
-mf_transcribe = gr.Interface(
-    fn=transcribe,
-    inputs=[
-        gr.Audio(sources="microphone", type="filepath"),
-        gr.Radio(["transcribe", "translate"], label="Task", value="transcribe"),
-    ],
-    outputs="text",
-    title="Whisper Large V3 Turbo: Transcribe Audio",
-    description=(
-        "Transcribe long-form microphone or audio inputs with the click of a button! Demo uses the"
-        f" checkpoint [{MODEL_NAME}](https://huggingface.co/{MODEL_NAME}) and 🤗 Transformers to transcribe audio files"
-        " of arbitrary length."
-    ),
-    allow_flagging="never",
-)
-
-file_transcribe = gr.Interface(
-    fn=transcribe,
-    inputs=[
-        gr.Audio(sources="upload", type="filepath", label="Audio file"),
-        gr.Radio(["transcribe", "translate"], label="Task", value="transcribe"),
-    ],
-    outputs="text",
-    title="Whisper Large V3: Transcribe Audio",
-    description=(
-        "Transcribe long-form microphone or audio inputs with the click of a button! Demo uses the"
-        f" checkpoint [{MODEL_NAME}](https://huggingface.co/{MODEL_NAME}) and 🤗 Transformers to transcribe audio files"
-        " of arbitrary length."
-    ),
-    allow_flagging="never",
-)
+    return html_embed_str, result
 
-yt_transcribe = gr.Interface(
+# Gradio 界面设置
+demo = gr.Blocks()
+yt_transcribe_interface = gr.Interface(
     fn=yt_transcribe,
-    inputs=[
-        gr.Textbox(lines=1, placeholder="Paste the URL to a YouTube video here", label="YouTube URL"),
-        gr.Radio(["transcribe", "translate"], label="Task", value="transcribe")
-    ],
+    inputs=[gr.Textbox(lines=1, placeholder="Paste the URL to a YouTube video here", label="YouTube URL")],
     outputs=["html", "text"],
-    title="Whisper Large V3: Transcribe YouTube",
-    description=(
-        "Transcribe long-form YouTube videos with the click of a button! Demo uses the checkpoint"
-        f" [{MODEL_NAME}](https://huggingface.co/{MODEL_NAME}) and 🤗 Transformers to transcribe video files of"
-        " arbitrary length."
-    ),
-    allow_flagging="never",
+    title="WhisperX: Transcribe YouTube with Speaker Diarization",
+    description="Transcribe and diarize YouTube videos with WhisperX."
 )
 
 with demo:
-    gr.TabbedInterface([mf_transcribe, file_transcribe, yt_transcribe], ["Microphone", "Audio file", "YouTube"])
+    gr.TabbedInterface([yt_transcribe_interface], ["YouTube"])
 
-demo.queue().launch(ssr_mode=False)
+demo.launch()