Video Subtitle Generation and Content Understanding with LLMs

Workflow

Video → speech-to-text (Whisper) → subtitle processing (LLM) → content understanding (LLM).

Subtitle Generation

Use Whisper to convert video audio to text, then use Qwen3.6 Plus for punctuation correction, segmentation, and terminology standardization to produce more readable subtitles.

Content Summarization

{"model":"qwen3.6-plus","messages":[{"role":"user","content":"Organize the following transcript into a timeline summary, marking key content at each timestamp:
{transcript}"}]}

Video Q&A

Use video subtitles as context. When users ask questions, retrieve relevant subtitle passages, then have the model generate answers — great for knowledge extraction and Q&A from long videos.