Quick summary: The January 2026 image-editing releases added text-guided changes inside chat, lowered the listed edit cost to 50 Moments, and corrected deletion behavior so removing an edited result could return to the earlier image instead of deleting the whole chain.
The feature changed editing from regeneration to modification
v4.8.0 introduced a dedicated editing model and a modal for changing an existing generated image. An edited image replaced the original result inside the message while remaining part of the conversation history. This made editing useful for localized changes such as clothing, objects, color, or background details.
The current workflow still begins with a generated image and a written instruction. It is not a manual mask or layer editor.
AI Video Gallery
A preview of selected portrait videos and AI character images.
The listed cost fell, but every attempt still matters
The release reduced image-editing cost from 100 to 50 Moments. Current Image Editing documentation also lists 50 Moments per edit. Because each refinement is a new paid action, users should request one coherent change and state what should remain unchanged.
Follow the operation-target-constraint format for image edits before generating several revisions.
The follow-up patch protected edit history
v4.8.1 fixed a cascade deletion bug and described a safer history behavior: deleting one edited image could revert to the previous result. This matters because a chain of edits should be treated as related outputs, not unrelated files.
The current documentation confirms repeated editing is possible, but users should still preserve a preferred result before experimenting further.
Editing is strongest for bounded changes
- Add or remove a specific item.
- Replace an object while keeping the pose.
- Change a color or material.
- Preserve lighting, expression, and composition explicitly.
Changing identity or rebuilding the whole scene is less predictable. A fresh generation is usually a clearer workflow when the requested person, pose, and background all need to change.
A valid edit can still be unwanted
A completed result that changed too much is different from a technical failure. The official guidance recommends naming the exact target and locking unchanged elements. When the service fails to create a result, preserve the source image, instruction, time, status, and Moment change rather than linking the problem to a broader generation article.
How the update fits the media workflow
Editing now sits between image generation and video creation: generate a usable base, refine a limited detail, then animate the preferred result. the source-image checks before creating a video cover the next stage. SecretOpedia’s Secrets AI image-editing update analysis tracks pricing or workflow changes.
There is also a current-state limitation that differs from the launch wording. The present Image Editing page describes the feature as SFW-only, even though the original v4.8.0 note referred to broader content support. Current documentation should control user expectations, and the live editor should be checked before spending Moments.
the Advanced image release analysis covers new generation with stronger identity adherence, while the Collection changes explain how edited outputs can be separated from Premium and Advanced source images. the Secrets AI review’s image-tool and Moments-value assessment provides the wider cost context.
Official sources checked July 12, 2026: Changelog, Image Editing, Image Generation, Moments & Payments.
