Kira 2.1: Inpaint, Expand, Image Edit Updates

Edited

Inpaint: More Seamless, Natural Results

We’ve upgraded our Inpaint model to produce cleaner, more seamless edits when replacing or repairing areas of an image.

  • Smoother blending between edited and original regions

  • More natural texture and lighting continuity

  • Fewer visible patch boundaries and artifacts

  • Better realism in small-area retouching and object cleanup

Inpaint results now look more cohesive and visually integrated into the original photo.

inpaint

Expand: Major Model Upgrade

We’ve upgraded Expand to a new, more powerful model, delivering a huge leap forward in expansion quality and stability.

  • Smarter scene continuation with cleaner edges

  • Better background coherence and lighting consistency

  • Fewer distortions and artifacts around subjects

  • More natural, realistic extended frames

This upgrade makes Expand feel more accurate, more reliable, and far closer to real-world photography.

Image Edit: Improved Facial Consistency

Previously, our edit model struggled to preserve facial consistency across multiple edits, particularly on certain face types and some Asian facial features. In some cases, faces could subtly shift or lose identity during repeated edits.

After extensive experimentation and evaluation across multiple model families, we’ve now rolled out a new image editing model that significantly improves:

  • Facial identity stability across edits

  • Consistency in expressions and structure

  • Natural-looking details on diverse face types

  • Reliability in multi-step editing workflows

Edits now feel more stable, more human, and more predictable, especially in portrait-heavy use cases.

img edit

What This Means for Users

Together, these upgrades further make Kira 2.1:

  • More trustworthy for portraits

  • More powerful for creative expansion

  • More consistent across diverse users & face types

  • More seamless when editing or repairing image details

And we’re not stopping here, we’ll continue improving editing realism, diversity accuracy, and multi-edit stability in upcoming iterations.

Was this article helpful?

Sorry about that! Care to tell us more?

Thanks for the feedback!

There was an issue submitting your feedback
Please check your connection and try again.