导入whisper库时触发TypeError：'NoneType'对象不可迭代

尝试用Python生成字幕，复制创作者代码并安装所有依赖后，运行时在import whisper行触发如下错误：

Exception has occurred:
TypeError argument of type 'NoneType' is not iterable
   File "D:\Python\Subtitle Generator\main.py", line 1, in <module>
     import whisper
TypeError: argument of type 'NoneType' is not iterable

移除该行则提示whisper缺失，完整代码如下：

import whisper
import os
import cv2
from moviepy.editor import ImageSequenceClip, AudioFileClip, VideoFileClip
from tqdm import tqdm

class VideoTranscriber:
    def __init__(self, model_path, video_path):
        self.model = whisper.load_model(model_path)
        self.video_path = video_path
        self.audio_path = ''
        self.text_array = []
        self.fps = 0
        self.char_width = 0

    def transcribe_video(self):
        print('Transcribing video')
        result = self.model.transcribe(self.audio_path)
        text = result["segments"][0]["text"]
        textsize = cv2.getTextSize(text, cv2.FONT_HERSHEY_SIMPLEX, 0.8, 2)[0]
        cap = cv2.VideoCapture(self.video_path)
        width = int(cap.get(cv2.CAP_PROP_FRAME_WIDTH))
        height = int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT))
        asp = 16/9
        ret, frame = cap.read()
        width = frame[:, int(int(width - 1 / asp * height) / 2):width - int((width - 1 / asp * height) / 2)].shape[1]
        width = width - (width * 0.1)
        self.fps = cap.get(cv2.CAP_PROP_FPS)
        self.char_width = int(textsize[0] / len(text))
        
        for j in tqdm(result["segments"]):
            lines = []
            text = j["text"]
            end = j["end"]
            start = j["start"]
            total_frames = int((end - start) * self.fps)
            start = start * self.fps
            total_chars = len(text)
            words = text.split(" ")
            i = 0
            
            while i < len(words):
                words[i] = words[i].strip()
                if words[i] == "":
                    i += 1
                    continue
                length_in_pixels = len(words[i]) * self.char_width
                remaining_pixels = width - length_in_pixels
                line = words[i] 
                
                while remaining_pixels > 0:
                    i += 1 
                    if i >= len(words):
                        break
                    length_in_pixels = len(words[i]) * self.char_width
                    remaining_pixels -= length_in_pixels
                    if remaining_pixels < 0:
                        continue
                    else:
                        line += " " + words[i]
                
                line_array = [line, int(start) + 15, int(len(line) / total_chars * total_frames) + int(start) + 15]
                start = int(len(line) / total_chars * total_frames) + int(start)
                lines.append(line_array)
                self.text_array.append(line_array)
        
        cap.release()
        print('Transcription complete')
    
    def extract_audio(self, output_audio_path='test_videos/audio.mp3'):
        print('Extracting audio')
        video = VideoFileClip(self.video_path)
        audio = video.audio 
        audio.write_audiofile(output_audio_path)
        self.audio_path = output_audio_path
        print('Audio extracted')
    
    def extract_frames(self, output_folder):
        print('Extracting frames')
        cap = cv2.VideoCapture(self.video_path)
        width = int(cap.get(cv2.CAP_PROP_FRAME_WIDTH))
        height = int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT))
        asp = width / height
        N_frames = 0
        
        while True:
            ret, frame = cap.read()
            if not ret:
                break
            
            frame = frame[:, int(int(width - 1 / asp * height) / 2):width - int((width - 1 / asp * height) / 2)]
            
            for i in self.text_array:
                if N_frames >= i[1] and N_frames <= i[2]:
                    text = i[0]
                    text_size, _ = cv2.getTextSize(text, cv2.FONT_HERSHEY_SIMPLEX, 0.8, 2)
                    text_x = int((frame.shape[1] - text_size[0]) / 2)
                    text_y = int(height/2)
                    cv2.putText(frame, text, (text_x, text_y), cv2.FONT_HERSHEY_SIMPLEX, 0.75, (0, 0, 255), 2)
                    break
            
            cv2.imwrite(os.path.join(output_folder, str(N_frames) + ".jpg"), frame)
            N_frames += 1
        
        cap.release()
        print('Frames extracted')

    def create_video(self, output_video_path):
        print('Creating video')
        image_folder = os.path.join(os.path.dirname(self.video_path), "frames")
        if not os.path.exists(image_folder):
            os.makedirs(image_folder)
        
        self.extract_frames(image_folder)
        
        print("Video saved at:", output_video_path)
        images = [img for img in os.listdir(image_folder) if img.endswith(".jpg")]
        images.sort(key=lambda x: int(x.split(".")[0]))
        
        frame = cv2.imread(os.path.join(image_folder, images[0]))
        height, width, layers = frame.shape
        
        clip = ImageSequenceClip([os.path.join(image_folder, image) for image in images], fps=self.fps)
        audio = AudioFileClip(self.audio_path)
        clip = clip.set_audio(audio)
        clip.write_videofile(output_video_path)

# Example usage
model_path = "base"
video_path = "Videos/video.mov"
output_video_path = "Videos/result.mov"

transcriber = VideoTranscriber(model_path, video_path)
transcriber.extract_audio()
transcriber.transcribe_video()
transcriber.create_video(output_video_path)

• 重新安装指定版本的whisper：卸载现有包后安装稳定版本，执行命令：

  pip uninstall openai-whisper -y
  pip install openai-whisper==20231117
  

• 确认Python环境一致性：检查运行代码的Python环境是否和安装依赖的环境一致，用where python（Windows）或which python（Linux/macOS）查看路径，再用pip list确认whisper已安装在该环境中。
• 搭建干净虚拟环境：创建全新虚拟环境，仅安装必要依赖，避免包冲突：

  pip install openai-whisper opencv-python moviepy tqdm
  

• 手动下载模型：若自动下载失败，手动下载对应模型（如base模型），放到whisper默认缓存路径（Windows：C:\Users\<用户名>\.cache\whisper；Linux/macOS：~/.cache/whisper），代码中指定本地模型路径。
• 调整代码初始化逻辑：将self.model = whisper.load_model(model_path)从__init__方法移到transcribe_video方法开头，避免导入阶段就加载模型引发潜在问题。

内容的提问来源于stack exchange，提问作者Flimzy