from flask import Flask, request, jsonify, Response from faster_whisper import WhisperModel import torch import io import time import datetime from threading import Semaphore import os from werkzeug.utils import secure_filename import tempfile from moviepy.editor import VideoFileClip import logging import torchaudio import ffmpeg # 导入 ffmpeg-python # 配置日志 logging.basicConfig(level=logging.INFO, format='%(asctime)s - %(levelname)s - %(message)s') app = Flask(__name__) # 配置 MAX_CONCURRENT_REQUESTS = 2 MAX_FILE_DURATION = 60 * 30 TEMPORARY_FOLDER = tempfile.gettempdir() ALLOWED_AUDIO_EXTENSIONS = {'mp3', 'wav', 'ogg', 'm4a', 'flac', 'aac', 'wma', 'opus', 'aiff'} ALLOWED_VIDEO_EXTENSIONS = {'mp4', 'avi', 'mov', 'mkv', 'webm', 'flv', 'wmv', 'mpeg', 'mpg', '3gp'} ALLOWED_EXTENSIONS = ALLOWED_AUDIO_EXTENSIONS.union(ALLOWED_VIDEO_EXTENSIONS) API_KEY = os.environ.get("API_KEY") MODEL_NAME = os.environ.get("WHISPER_MODEL", "guillaumekln/faster-whisper-large-v2") # 设备检查 device = "cuda" if torch.cuda.is_available() else "cpu" compute_type = "float16" if device == "cuda" else "int8" logging.info(f"使用设备: {device}，计算类型: {compute_type}") # Faster Whisper 设置 beamsize = 2 try: wmodel = WhisperModel( MODEL_NAME, device=device, compute_type=compute_type, download_root="./model_cache" ) logging.info(f"模型 {MODEL_NAME} 加载成功.") except Exception as e: logging.error(f"加载模型 {MODEL_NAME} 失败: {e}") wmodel = None # 并发控制 request_semaphore = Semaphore(MAX_CONCURRENT_REQUESTS) active_requests = 0 def validate_api_key(request): api_key = request.headers.get('X-API-Key') if api_key == API_KEY: return True else: return False def allowed_file(filename): return '.' in filename and \ filename.rsplit('.', 1)[1].lower() in ALLOWED_EXTENSIONS def cleanup_temp_files(*file_paths): for file_path in file_paths: try: if file_path and os.path.exists(file_path): os.remove(file_path) logging.info(f"删除临时文件: {file_path}") except Exception as e: logging.error(f"删除临时文件 {file_path} 出错: {str(e)}") def extract_audio_from_video(video_path, output_audio_path): try: # 使用 ffmpeg-python 调用 FFmpeg ffmpeg.input(video_path).output(output_audio_path, acodec='pcm_s16le').run(capture_stdout=True, capture_stderr=True) # or use with more options: # ffmpeg.input(video_path).output(output_audio_path, acodec='pcm_s16le', ar=44100, ac=2).run(capture_stdout=True, capture_stderr=True) # 检查视频时长 video = VideoFileClip(video_path) # moviepy if video.duration > MAX_FILE_DURATION: video.close() raise ValueError(f"视频时长超过 {MAX_FILE_DURATION} 秒") video.close() return output_audio_path except Exception as e: logging.exception("提取视频中的音频出错") raise Exception(f"提取视频中的音频出错: {str(e)}") @app.route("/health", methods=["GET"]) def health_check(): return jsonify({ 'status': 'API 正在运行', 'timestamp': datetime.datetime.now().isoformat(), 'device': device, 'compute_type': compute_type, 'active_requests': active_requests, 'max_duration_supported': MAX_FILE_DURATION, 'supported_formats': list(ALLOWED_EXTENSIONS), 'model': MODEL_NAME }) @app.route("/status/busy", methods=["GET"]) def server_busy(): is_busy = active_requests >= MAX_CONCURRENT_REQUESTS return jsonify({ 'is_busy': is_busy, 'active_requests': active_requests, 'max_capacity': MAX_CONCURRENT_REQUESTS }) @app.route("/whisper_transcribe", methods=["POST"]) def transcribe(): global active_requests if not validate_api_key(request): return jsonify({'error': '无效的 API 密钥'}), 401 if not request_semaphore.acquire(blocking=False): return jsonify({'error': '服务器繁忙'}), 503 active_requests += 1 start_time = time.time() temp_file_path = None temp_audio_path = None try: if wmodel is None: return jsonify({'error': '模型加载失败。请检查服务器日志。'}), 500 if 'file' not in request.files: return jsonify({'error': '未提供文件'}), 400 file = request.files['file'] if not (file and allowed_file(file.filename)): return jsonify({'error': f'无效的文件格式。支持：{", ".join(ALLOWED_EXTENSIONS)}'}), 400 # 保存上传的文件到临时位置 temp_file_path = os.path.join(TEMPORARY_FOLDER, secure_filename(file.filename)) file.save(temp_file_path) # 检查是否是视频文件，如果是，则提取音频 file_extension = file.filename.rsplit('.', 1)[1].lower() is_video = file_extension in ALLOWED_VIDEO_EXTENSIONS if is_video: temp_audio_path = os.path.join(TEMPORARY_FOLDER, f"temp_audio_{int(time.time())}.wav") extract_audio_from_video(temp_file_path, temp_audio_path) transcription_file = temp_audio_path else: transcription_file = temp_file_path # 检查音频文件时长 try: # 使用 torchaudio.load 加载音频，并指定格式 waveform, sample_rate = torchaudio.load(transcription_file, format=file_extension) duration = waveform.size(1) / sample_rate if duration > MAX_FILE_DURATION: raise ValueError(f"音频时长超过 {MAX_FILE_DURATION} 秒") except Exception as load_err: logging.exception(f"使用 torchaudio.load 加载音频文件出错: {transcription_file}") try: # 尝试使用 soundfile 后端加载 (禁用 sox_io) torchaudio.set_audio_backend("soundfile") # 强制使用 soundfile 后端 waveform, sample_rate = torchaudio.load(transcription_file) # 不要指定文件扩展名 duration = waveform.size(1) / sample_rate if duration > MAX_FILE_DURATION: raise ValueError(f"音频时长超过 {MAX_FILE_DURATION} 秒") except Exception as soundfile_err: logging.exception(f"使用 soundfile 后端加载音频文件出错: {transcription_file}") return jsonify({'error': f'使用两个后端加载音频文件都出错: {str(soundfile_err)}'}), 400 finally: torchaudio.set_audio_backend("default") # 恢复默认音频后端 # 转录音频文件 segments, _ = wmodel.transcribe( transcription_file, beam_size=beamsize, vad_filter=True, without_timestamps=True, compression_ratio_threshold=2.4, word_timestamps=False ) full_text = " ".join(segment.text for segment in segments) return jsonify({ 'transcription': full_text, 'file_type': 'video' if is_video else 'audio' }), 200 except Exception as e: logging.exception("转录过程中发生异常") return jsonify({'error': str(e)}), 500 finally: cleanup_temp_files(temp_file_path, temp_audio_path) active_requests -= 1 request_semaphore.release() print(f"处理时间：{time.time() - start_time:.2f}s (活动请求：{active_requests})") if __name__ == "__main__": # 创建临时文件夹（如果不存在） if not os.path.exists(TEMPORARY_FOLDER): os.makedirs(TEMPORARY_FOLDER) logging.info(f"创建临时文件夹: {TEMPORARY_FOLDER}") app.run(host="0.0.0.0", port=7860, threaded=True)