Update app.py

app.py

@@ -1,12 +1,23 @@
+import spaces
 import torch
+
 import gradio as gr
+import yt_dlp as youtube_dl
 from transformers import pipeline
-import openai
+from transformers.pipelines.audio_utils import ffmpeg_read
+
+import tempfile
 import os
+import openai
+
+# 設定 OpenAI API 金鑰（請替換為您自己的 API 金鑰）
+openai.api_key = "YOUR_OPENAI_API_KEY"
 
-# 使用 Whisper 模型进行语音转录
-MODEL_NAME = "openai/whisper-large-v2"  # 使用支持的模型名称
+MODEL_NAME = "openai/whisper-large-v3-turbo"
 BATCH_SIZE = 8
+FILE_LIMIT_MB = 1000
+YT_LENGTH_LIMIT_S = 3600  # 限制 YouTube 檔案為 1 小時
+
 device = 0 if torch.cuda.is_available() else "cpu"
 
 pipe = pipeline(
@@ -16,83 +27,146 @@ pipe = pipeline(
     device=device,
 )
 
-# 设置 OpenAI API 密钥
-openai.api_key = os.getenv('OPENAI_API_KEY')
-if openai.api_key is None:
-    raise ValueError("请设置 OpenAI API 密钥为环境变量 'OPENAI_API_KEY'。")
+def translate_text(input_text, target_language):
+    prompt = f"請將以下文字翻譯成{target_language}：\n\n{input_text}"
+    
+    response = openai.ChatCompletion.create(
+        model="gpt-4o",
+        messages=[
+            {"role": "user", "content": prompt}
+        ]
+    )
+    
+    translated_text = response['choices'][0]['message']['content'].strip()
+    return translated_text
+
+@spaces.GPU
+def transcribe(inputs, task, translate_option, target_language):
+    if inputs is None:
+        raise gr.Error("未提交音訊檔案！請在提交請求前上傳或錄製音訊檔案。")
+
+    result = pipe(inputs, batch_size=BATCH_SIZE, generate_kwargs={"task": task}, return_timestamps=True)
+    text = result["text"]
+
+    if translate_option == "是":
+        text = translate_text(text, target_language)
 
-# 定义语音转文字函数
-def transcribe(audio):
-    if audio is None:
-        raise gr.Error("请上传或录制音频文件。")
-    text = pipe(audio)["text"]
     return text
 
-# 定义翻译函数
-def translate_text(text, target_language):
-    if target_language == "None" or not target_language:
-        return "未选择翻译语言。", None
+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
 
-    prompt = f"请将以下文本翻译成 {target_language}：\n\n{text}"
+def download_yt_audio(yt_url, filename):
+    info_loader = youtube_dl.YoutubeDL()
+    
     try:
-        response = openai.ChatCompletion.create(
-            model="gpt-4",  # 使用 GPT-4 模型
-            messages=[{"role": "user", "content": prompt}],
-            max_tokens=1000,
-            n=1,
-            temperature=0.5,
-        )
-        translation = response.choices[0].message["content"].strip()
-        return translation
-    except Exception as e:
-        return f"翻译出错：{str(e)}"
-
-# 定义完整的处理流程函数
-def transcribe_and_translate(audio, target_language):
-    text = transcribe(audio)
-    if target_language != "None":
-        translation = translate_text(text, target_language)
-        return text, translation
-    else:
-        return text, "未选择翻译语言。"
-
-# 构建 Gradio 界面
-with gr.Blocks() as demo:
-    gr.Markdown("# 语音转文字并翻译应用")
-
-    with gr.Tabs():
-        with gr.TabItem("麦克风输入"):
-            audio_input = gr.Audio(source="microphone", type="filepath", label="录制音频")
-            language_dropdown = gr.Dropdown(
-                choices=["None", "English", "French", "German", "Spanish", "Chinese"],
-                value="None",
-                label="翻译成以下语言",
-            )
-            transcribe_button = gr.Button("开始转录和翻译")
-            original_text_output = gr.Textbox(label="转录文本")
-            translated_text_output = gr.Textbox(label="翻译文本")
-
-            transcribe_button.click(
-                fn=transcribe_and_translate,
-                inputs=[audio_input, language_dropdown],
-                outputs=[original_text_output, translated_text_output],
-            )
-
-        with gr.TabItem("上传音频文件"):
-            file_input = gr.Audio(type="filepath", label="上传音频文件")
-            language_dropdown_file = gr.Dropdown(
-                choices=["None", "English", "French", "German", "Spanish", "Chinese"],
-                value="None",
-                label="翻译成以下语言",
-            )
-            transcribe_button_file = gr.Button("开始转录和翻译")
-            original_text_output_file = gr.Textbox(label="转录文本")
-            translated_text_output_file = gr.Textbox(label="翻译文本")
-
-            transcribe_button_file.click(
-                fn=transcribe_and_translate,
-                inputs=[file_input, language_dropdown_file],
-                outputs=[original_text_output_file, translated_text_output_file],
-            )
-
-demo.launch()
+        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"最大 YouTube 長度為 {yt_length_limit_hms}，但獲得了長度為 {file_length_hms} 的影片。")
+    
+    ydl_opts = {"outtmpl": filename, "format": "worstvideo[ext=mp4]+bestaudio[ext=m4a]/best[ext=mp4]/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))
+
+@spaces.GPU
+def yt_transcribe(yt_url, task, translate_option, target_language, 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")
+        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}
+
+    result = pipe(inputs, batch_size=BATCH_SIZE, generate_kwargs={"task": task}, return_timestamps=True)
+    text = result["text"]
+
+    if translate_option == "是":
+        text = translate_text(text, target_language)
+
+    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="任務", value="transcribe"),
+        gr.Radio(["是", "否"], label="是否翻譯轉錄結果", value="否"),
+        gr.Dropdown(["英文", "日文", "法文", "德文", "西班牙文"], label="目標語言", value="英文")
+    ],
+    outputs="text",
+    title="清華大學多模態課程＆廖老師嫡傳弟子-第二組 「語音轉文字」model",
+    description=(
+        "只需點擊一下按鈕，即可轉錄長篇的麥克風或音訊輸入！演示使用了"
+        f"檢查點 [{MODEL_NAME}](https://huggingface.co/{MODEL_NAME}) 和 🤗 Transformers 來轉錄任意長度的音訊文件。"
+    ),
+    allow_flagging="never",
+)
+
+file_transcribe = gr.Interface(
+    fn=transcribe,
+    inputs=[
+        gr.Audio(sources="upload", type="filepath", label="音訊檔案"),
+        gr.Radio(["transcribe", "translate"], label="任務", value="transcribe"),
+        gr.Radio(["是", "否"], label="是否翻譯轉錄結果", value="否"),
+        gr.Dropdown(["英文", "日文", "法文", "德文", "西班牙文"], label="目標語言", value="英文")
+    ],
+    outputs="text",
+    title="Whisper Large V3: 轉錄音訊",
+    description=(
+        "只需點擊一下按鈕，即可轉錄長篇的麥克風或音訊輸入！演示使用了"
+        f"檢查點 [{MODEL_NAME}](https://huggingface.co/{MODEL_NAME}) 和 🤗 Transformers 來轉錄任意長度的音訊文件。"
+    ),
+    allow_flagging="never",
+)
+
+yt_transcribe = gr.Interface(
+    fn=yt_transcribe,
+    inputs=[
+        gr.Textbox(lines=1, placeholder="在此處貼上 YouTube 視頻的 URL", label="YouTube URL"),
+        gr.Radio(["transcribe", "translate"], label="任務", value="transcribe"),
+        gr.Radio(["是", "否"], label="是否翻譯轉錄結果", value="否"),
+        gr.Dropdown(["英文", "日文", "法文", "德文", "西班牙文"], label="目標語言", value="英文")
+    ],
+    outputs=["html", "text"],
+    title="Whisper Large V3: 轉錄 YouTube",
+    description=(
+        "只需點擊一下按鈕，即可轉錄長篇的 YouTube 視頻！演示使用了"
+        f"檢查點 [{MODEL_NAME}](https://huggingface.co/{MODEL_NAME}) 和 🤗 Transformers 來轉錄任意長度的視頻文件。"
+    ),
+    allow_flagging="never",
+)
+
+with demo:
+    gr.TabbedInterface([mf_transcribe, file_transcribe, yt_transcribe], ["麥克風", "音訊檔案", "YouTube"])
+
+demo.queue().launch(ssr_mode=False)