"Gemini's speech recognition always produces inaccurate timestamps.

[abhijeet12s]

"Gemini's speech recognition always produces inaccurate timestamps. Instead of relying on it, can we use cv2 image detection to detect scene changes in the video? After detecting these scene changes, we can cut the video into clips based on these scenes. These video clips can then be provided to Gemini for transcription. Finally, we will generate an SRT file and insert the transcribed audio text into it."