Codex ImageGen Adds a Lightbox + Canvas Workflow for Image Editing on Desktop

OpenAI is iterating on ImageGen inside Codex with a more direct editing loop: a lightbox to inspect images and a canvas surface for edits.

OpenAI Developers says “ImageGen in Codex just got a new lightbox and canvas,” framing it as a workflow improvement for exploring and refining visuals. (OpenAIDevs on X)

The post quotes a demo from David Kundel that frames the core interaction model: “A picture may be worth a thousand words, but editing one shouldn’t take a thousand-word prompt,” followed by examples of pointing at a character, erasing elements, and leaving comments for title placement. (dkundel on X)

What changed

• Lightbox viewer: a dedicated view for inspecting and iterating on generated images. (OpenAIDevs on X)
• Canvas-style editing: an editing surface meant for direct, localized adjustments (erase / annotate / place text) rather than rewriting prompts from scratch. (dkundel on X)

Why this matters (for builders)

If your Codex workflows include generating UI mockups, diagrams, product shots, or other visual assets, the practical win is shorter iteration cycles: you can request specific edits by pointing, erasing, or commenting, instead of repeatedly re-prompting the whole image.

What we are (and aren’t) claiming

We have not found a durable, official changelog entry describing the lightbox/canvas update as of this run. Accordingly, this post limits claims to what is explicitly stated in the approved-source posts above.

Sources

• OpenAI Developers post on X (Jul 30, 2026) — https://x.com/OpenAIDevs/status/2082944138635595782
• David Kundel demo post on X (quoted by OpenAI Developers, Jul 30, 2026) — https://x.com/dkundel/status/2082943803716194712