app.py

import warnings

# 過濾 SyntaxWarning 警告
warnings.filterwarnings("ignore", category=SyntaxWarning)

import gradio as gr
from moviepy.editor import VideoFileClip
import cv2
import base64
from openai import OpenAI
import os

def extract_frames(video_path, seconds_per_frame):
    base64Frames = []
    video = cv2.VideoCapture(video_path)
    total_frames = int(video.get(cv2.CAP_PROP_FRAME_COUNT))
    fps = video.get(cv2.CAP_PROP_FPS)
    frames_to_skip = int(fps * seconds_per_frame)
    curr_frame = 0

    while curr_frame < total_frames - 1:
        video.set(cv2.CAP_PROP_POS_FRAMES, curr_frame)
        success, frame = video.read()
        if not success:
            break
        _, buffer = cv2.imencode(".jpg", frame)
        base64Frames.append(base64.b64encode(buffer).decode("utf-8"))
        curr_frame += frames_to_skip
    video.release()
    return base64Frames

def extract_audio(video_path):
    base_video_path, _ = os.path.splitext(video_path)
    audio_path = f"{base_video_path}.mp3"
    clip = VideoFileClip(video_path)
    clip.audio.write_audiofile(audio_path, bitrate="32k")
    clip.audio.close()
    clip.close()
    return audio_path

def process_video(video_path, seconds_per_frame):
    base64Frames = extract_frames(video_path, seconds_per_frame)
    audio_path = extract_audio(video_path)
    return base64Frames, audio_path

def get_summary_prompt(summary_detail):
    if summary_detail == "簡短":
        return "生成一個簡短的摘要，不超過50字。"
    elif summary_detail == "中等":
        return "生成一個中等長度的摘要，概述影片的主要內容。"
    else:  # 詳細
        return "生成一個詳細的摘要，包括影片的主要內容和關鍵細節。"

def summarize_video(file_path, summary_detail):
    api_key = os.getenv("OPENAI_API_KEY")
    client = OpenAI(api_key=api_key)

    # 抽取幀和音頻（每0.5秒一幀）
    base64Frames, audio_path = process_video(file_path, seconds_per_frame=0.5)

    # 使用Whisper進行音頻轉錄
    transcription = client.audio.transcriptions.create(
        model="whisper-1", file=open(audio_path, "rb")
    )

    # 根據用戶選擇的詳細程度設置提示
    summary_prompt = get_summary_prompt(summary_detail)

    # 使用GPT-4o生成摘要
    response = client.chat.completions.create(
        model="gpt-4o",
        messages=[
            {
                "role": "system",
                "content": f"""您是一名優秀的摘要專家，請根據提供的影片和其轉錄內容生成Markdown格式的摘要。{summary_prompt}""",
            },
            {
                "role": "user",
                "content": [
                    "以下是從影片中提取的幀畫面",
                    *map(
                        lambda x: {
                            "type": "image_url",
                            "image_url": {
                                "url": f"data:image/jpg;base64,{x}",
                                "detail": "low",
                            },
                        },
                        base64Frames,
                    ),
                    {
                        "type": "text",
                        "text": f"這是影片的轉錄內容: {transcription.text}",
                    },
                ],
            },
        ],
        temperature=0,
    )

    return response.choices[0].message.content

demo = gr.Interface(
    fn=summarize_video,
    inputs=[
        gr.File(label="上傳影片 (mp4)"),
        gr.Radio(["簡短", "中等", "詳細"], label="摘要詳細程度", value="中等")
    ],
    outputs="markdown",
    title="影片摘要生成器",
    description="上傳影片，將生成影片的摘要。",
    flagging_dir=""  # 這會移除 Flag 功能
)

if __name__ == "__main__":
    demo.launch()