Your own pictures
An image slide usually shows a picture generated for the lesson. It can show one of yours instead: a screenshot, a chart, a diagram, a photo, a drawing.
Your picture carries no Generated image tag, and it doesn't count towards the 12 generated pictures an explainer can hold.
On Scrimba
Attach the picture to your prompt on scrimba.com/explain.
Explain shows it when you ask, or when the picture is what the lesson is about: the interface, chart, drawing or photo you're asking to have explained. Each one is shown at most once, alone on its slide, with the narration pointing into it.
Pictures it reads instead of shows
A screenshot of code, an error message or a page of text is material to read from, not a slide. The lesson teaches from what's in it: code comes back as a code slide, text as a card or a layout.
Ask for it directly if you want to see the screenshot itself.
Setting the style of the generated pictures
An attached picture can also steer the ones Explain generates. Attach a drawing of a character, or a picture in the style you want, and ask for the scenes to match. The explainer shows your picture once, early, then draws the rest from it.
From a coding agent
An agent connected over MCP, such as Claude Code, uploads pictures while it writes the lesson: a screenshot or diagram from the repository, a photo, or an image it generated with its own tools.
Each file is posted to https://scrimba.com/explain/uploads with the stream token in an X-Stream-Token header. The reply carries an attachment id, and the agent writes an image slide that names that id instead of a prompt. The authoring contract spells out the call, so there's nothing for you to set up.
From ChatGPT
A picture in a ChatGPT conversation works the same way, whether you uploaded it or ChatGPT generated it. Ask for it when you ask for the video:
Pictures uploaded from the ChatGPT mobile app can't be fetched, so the explainer is built without them. See ChatGPT / Codex.
Limits
Uploads from an agent or from ChatGPT are capped per explainer:
| Limit | |
|---|---|
| Pictures per explainer | 12 |
| File size | 8 MB each, 40 MB in total |
| Dimensions | 30 megapixels |
| Formats | PNG, JPEG, WebP, GIF |
A picture that's refused doesn't stop the explainer. The agent is told what was wrong and how many pictures it has left, and it builds the rest of the lesson without that one.