Midjourney releases V8.2 Edit update: text-based image editing

Midjourney has opened public beta testing for its first image editing model for V8.2, which edits images based on text instructions, accepts up to 4 reference images, can repaint selected areas, and extend the canvas beyond the original boundaries.

Compiled by Sergey KostenchukPublished 2026-08-28Updated 2026-09-18

2026-08-28 Multimodal OpenAI

🎨 Midjourney V8.2 Edit: editing images with text

Midjourney has opened public beta testing for the Edit model for V8.2 — the "Edit Model for V8" announcement was released on August 27, 2026. It edits images based on text instructions, performs inpainting and outpainting, and accepts up to 4 reference images instead of omni-reference. Launch — in the Editor tab or with the --edit command in Discord.

🌍 By integrating text-based edits into the main generation pipeline, Midjourney responds to ChatGPT/GPT-Image and Google Gemini (Nano Banana): competition is shifting from one-off generation to an environment for sequential edits.

👤 A subscription from $10/month (200 minutes of Fast GPU) is required, there is no separate fee — GPU time is deducted. Edits are made with ordinary words, without Photoshop or manual masks. This is an early beta: complex cases are unstable, and users are asked to report failures with an emoji in the lightbox or via jobid.

Source 1: https://updates.midjourney.com/edit-model-for-v8/ Source 2: https://news.aibase.com/news/30690

Sources