Update app.py

app.py

@@ -1,60 +1,137 @@
+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 tempfile
 import os
 import time
-import json
-import gradio as gr
-import torch
-import torchaudio
-import numpy as np
-from denoiser.demucs import Demucs
-from pydub import AudioSegment
-from transformers import AutoModelForSequenceClassification, AutoTokenizer
-
-# 设置 Hugging Face Hub 的 Access Token
-auth_token = os.getenv("HF_TOKEN")
-
-# 加载私有模型
-model_id = "DeepLearning101/Speech-Quality-Inspection_Meta-Denoiser"
-model = AutoModelForSequenceClassification.from_pretrained(model_id, token=auth_token)
-tokenizer = AutoTokenizer.from_pretrained(model_id, token=auth_token)
-
-def transcribe(file_upload, microphone):
-    file = microphone if microphone is not None else file_upload
-    demucs_model = Demucs(hidden=64)
-    state_dict = torch.load("path_to_model_checkpoint", map_location='cpu')  # 请确保提供正确的模型文件路径
-    demucs_model.load_state_dict(state_dict)
-    x, sr = torchaudio.load(file)
-    out = demucs_model(x[None])[0]
-    out = out / max(out.abs().max().item(), 1)
-    torchaudio.save('enhanced.wav', out, sr)
-    enhanced = AudioSegment.from_wav('enhanced.wav')  # 只有去完噪的需要降bitrate再做语音识别
-    enhanced.export('enhanced.wav', format="wav", bitrate="256k")
-
-    # 假设模型是用于文本分类
-    inputs = tokenizer("enhanced.wav", return_tensors="pt")
-    outputs = model(**inputs)
-    predictions = torch.nn.functional.softmax(outputs.logits, dim=-1)
-
-    return "enhanced.wav", predictions
-
-demo = gr.Interface(
+
+MODEL_NAME = "openai/whisper-large-v3"
+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,
+)
+
+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
+
+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"]
+    file_length_s = int(file_length)
+    
+    if file_length_s > YT_LENGTH_LIMIT_S:
+        yt_length_limit_hms = time.strftime("%H:%M:%S", time.gmtime(YT_LENGTH_LIMIT_S))
+        file_length_hms = time.strftime("%H:%M:%S", 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": "bestaudio/best"}
+    
+    with youtube_dl.YoutubeDL(ydl_opts) as ydl:
+        try:
+            ydl.download([yt_url])
+        except youtube_dl.utils.ExtractorError as err:
+            raise gr.Error(str(err))
+
+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, "audio.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"]
+
+    return html_embed_str, text
+
+mf_transcribe = gr.Interface(
     fn=transcribe,
     inputs=[
-        gr.Audio(type="filepath", label="语音质检麦克风实时录音"),
-        gr.Audio(type="filepath", label="语音质检原始音档"),
+        gr.Audio(type="filepath"),
+        gr.Radio(["transcribe", "translate"], label="Task", value="transcribe"),
     ],
-    outputs=[
-        gr.Audio(type="filepath", label="Output"),
-        gr.Textbox(label="Model Predictions")
+    outputs="text",
+    layout="horizontal",
+    theme="huggingface",
+    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",
+)
+
+file_transcribe = gr.Interface(
+    fn=transcribe,
+    inputs=[
+        gr.Audio(type="filepath", label="Audio file"),
+        gr.Radio(["transcribe", "translate"], label="Task", value="transcribe"),
     ],
-    title="<p style='text-align: center'><a href='https://www.twman.org/AI' target='_blank'>语音质检噪音去除 (语音增强)：Meta Denoiser</a>",
-    description="为了提升语音识别的效果，可以在识别前先进行噪音去除",
+    outputs="text",
+    layout="horizontal",
+    theme="huggingface",
+    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",
-    examples=[
-        ["exampleAudio/15s_2020-03-27_sep1.wav"],
-        ["exampleAudio/13s_2020-03-27_sep2.wav"],
-        ["exampleAudio/30s_2020-04-23_sep1.wav"],
-        ["exampleAudio/15s_2020-04-23_sep2.wav"],
+)
+
+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.Radio(["transcribe", "translate"], label="Task", value="transcribe")
     ],
+    outputs=["html", "text"],
+    layout="horizontal",
+    theme="huggingface",
+    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",
 )
 
-demo.launch(debug=True)
+with gr.Blocks() as demo:
+    gr.TabbedInterface([mf_transcribe, file_transcribe, yt_transcribe], ["Microphone", "Audio file", "YouTube"])
+
+demo.launch(enable_queue=True)