ScribeShot
Product documentation
Analyze captured video screenshots with AI
A screenshot becomes much more useful when ScribeShot understands what is inside it. Vision analysis extracts a structured description of the frame so visual details can participate in retrieval and chat alongside the transcript.
Updated
Quick workflow
- 1
Capture a useful frame from the video.
- 2
Let ScribeShot analyze the image with your local vision model.
- 3
Keep the analysis with that screenshot inside the lesson folder.
- 4
Ask video-chat questions that require evidence from the captured visual.
What the analysis is for
The goal is not to generate a caption for display. The analysis gives ScribeShot searchable information about visible text, code, diagrams, labels, and the relationships inside the image. That lets retrieval find the right screenshot when a question refers to something that was shown rather than spoken.
Transcript plus visual context
Many tutorials have two parallel information channels: narration and the screen. The transcript covers narration. Screenshot analysis covers the screen. ScribeShot keeps both available so a question like “what did the architecture diagram show between the API and the worker?” can retrieve visual evidence rather than guessing from nearby speech.
Local on-device vision
Screenshot analysis runs locally on your Mac using supported on-device models like Qwen3-VL 8B and Gemma 4. Vision analysis extracts visible text, code, and diagrams without sending screenshots to external cloud providers.
See Run ScribeShot with local AI for the local model options.
Use analyzed screenshots in chat
Video chat searches both transcript material and the indexed screenshots. When a captured frame is relevant, ScribeShot can surface it inline with the response and connect the answer back to the lesson timestamp.
Common questions
Can ScribeShot read text that appears only inside a screenshot?
Yes. Vision analysis is specifically used so visible text and visual structure can become searchable context.
Does screenshot analysis require an AI API key?
No. Supported local models run on your Mac with zero cloud uploads.
Try visual context
Analyze a frame from a video you actually need
Capture a useful diagram, code block or slide and see how ScribeShot makes that visual available to retrieval and chat.