Midjourney has opened testing of its first V8.2 image editing model to all users, moving the workflow from regeneration toward editing an image you already have.
Key facts
- Status: Open testing to all users since 2026-08-27
- Capabilities: Instruction editing, image-to-image with up to 4 references, inpainting, outpainting
- How to use it: Keep real photography for the product itself; let AI handle scene extension and aspect ratios. Altering product form is an ad-policy risk
What happened
The model supports instruction-based editing, image-to-image referencing up to four images at once, inpainting and outpainting, and is compatible with personalization, moodboards and srefs. It is available through the web interface or the `--edit` command on Discord, with updated interfaces on midjourney.com and alpha.midjourney.com.

What it means for independent sellers
For anyone producing store content, the parts that matter are the four-image referencing and inpainting. The long-standing problem with AI imagery in ecommerce is consistency: the same product looks subtly different across scenes, which confines AI output to mood shots and rules it out for product positions. Multi-reference plus inpainting makes "keep the product identical, change only the background and setting" practical, which is exactly what social creatives and banners need. A sane workflow: keep real photography for the primary product shot and let AI handle scene extension and aspect-ratio adaptation — extending one photo into 4:5 and 16:9 variants, for instance. One caution: altering the product's actual appearance with AI is a policy risk on most ad platforms. Extending a background is fine; changing the product's form is not.