Generate viral short clips from long videos
Supports YouTube, TikTok, Instagram, Google Drive
Skip the download by reusing an existing job ID.
Leave blank for the server default, or enter the model name/alias already running on Kaggle.
The prompt and the reserved output must fit in the modelβs context window. You can change the context limit when starting the vLLM server.
Start with 1β2 clips. On a single 16 GB GPU, choose Whisper CPU or YouTube Subs to conserve memory. vLLM does not automatically fall back to cloud providers.