Yeah, you just said what was on my mind since I launched it. The code I wrote is for tutorial videos. Non-tutorial video responses are just gibberish. The representative videos on the top is a great idea. I'll look into it.
Can you tell me more about the video you entered? Did it have a transcript? How many hours long was it?
Sorry for that, I'm looking into it. The problem is for videos that have no transcript. Maybe it's because i'm feeding it the description of the video for now. I'll find some workaround for this. Thanks!
> The problem is for videos that have no transcript.
Whisper or other models can help with that too, but remember to preprocess to cut silence. The dataset tends to include ads in the captions, which results in hallucinated in from silence.
You could also add a transcript-evaluation step which checks whether this actually looks like a step-by-step video, but I'd consider skipping it for cost and efficiency. Trying to be helpful by evaluating whether the video is instructions or not is added complexity where bugs and strange behavior can creep in.
Feels like you might have to explicitly ask it not to put "drop a comment below" or "like and subscribe" into the instructions (or strip it from transcripts), since most YouTubers who take YouTube seriously are going to ask...
Consider passing the video and transcripts through SponsorBlock (removing sponsor, self-promo, interaction remember, intro and outro segments from the videos) before stepifying them, that might help
