Generate and edit images with AI — personal color analysis, portrait restyles, and creative image workflows, all in one place.
Generate and edit images with AI — personal color analysis, portrait restyles, and creative image workflows, all in one place.
Generate and edit images with AI — personal color analysis, portrait restyles, and creative image workflows, all in one place.
Describe what you want and let ChatImg do the heavy lifting.
Export in the shapes and pacing the platforms reward.
If you can describe it, you can ship it — the hard parts are automated.
Don't start from a blank prompt — start from a picture. Upload a photo and let AI restyle it, swap the background, or reimagine it in a new art style while keeping the subject recognizable.
Upload a selfie, product shot, or pet photo and describe the change — the AI edits what you name and keeps the rest.
Turn a photo into an anime portrait, watercolor, 3D render, or professional headshot with a single instruction.
Image-to-image preserves the original composition, so faces and products stay consistent across variations.
No tools to learn and nothing to install.
Tell ChatImg what you want.
Tweak the AI draft until it feels right.
Download and share to your feeds.
The creators who reach for ChatImg most.
Ship on a schedule without a full team.
Turn a hook into a scroll-stopping clip.
Instruction-based image editing changes an existing picture from a written instruction such as "make the shirt blue". The original image is kept and only the named part is altered. It was introduced by the InstructPix2Pix paper in 2022, which also showed that no mask or selection is required — the model infers the region from the words.
Text-to-image generation builds a new picture from a description, so every element shifts each time it runs. Editing starts from an existing picture and changes only what the instruction names. That is why re-prompting a generator never reliably produces "the same photo with a different background" — it is the wrong operation.
Image-to-image takes an existing photo as the starting point and generates a new version guided by your instruction — for example turning a selfie into an anime portrait or a product shot into a studio-lit render. Because generation is conditioned on the source picture, composition and identity carry over, which pure text-to-image cannot guarantee.
Inpainting regenerates a masked region of an image while leaving the rest untouched, traditionally requiring the user to paint the area first. Instruction-based editing removes that step by deriving the region from the instruction itself, which lowers the skill floor from knowing an editor to writing a sentence.
InstructPix2Pix (Brooks, Holynski, Efros, 2022) established instruction-following image editing: edits are applied in a single forward pass with no per-example fine-tuning and no user-supplied mask.
InstructPix2Pix — arXiv:2211.09800 ↗FAQ'S
Ask us anything!
Generate and edit images with AI — personal color analysis, portrait restyles, and creative image workflows, all in one place.
