Black Forest Labs' instruction-following image edit model — surgical, consistent, and finally better than Photoshop Generative Fill for production work.
Flux Kontext is Black Forest Labs' instruction-follow edit model, built on top of the Flux.1 image generation family. The pitch: edit an existing image with pure text instructions ("change the jacket to red, keep everything else identical") — and have the edit land on-target with character and scene consistency intact.
Who it's for: Ecommerce studios, ad agencies, and content teams doing high-volume image retouches. Hobbyists should still default to Photoshop Generative Fill or Midjourney Edit for one-offs.
"Change the sky to sunset, keep everything else". Kontext actually delivers — unlike Stable Diffusion Inpainting which hallucinated half the image. Edits stay true to the source.
Feed a character reference and Kontext keeps face, build, and outfit across edits. Massive for storyboarding, ad campaigns, and IP-consistent content.
Stack instructions across rounds without quality degradation. Round 1: change the background. Round 2: change the lighting. Round 3: add a product. Each step preserves prior state.
REST + Python SDK with sub-5s latency. ~$0.05 per edit on Kontext Pro, $0.08 on Kontext Max for the highest quality. Batch API available.
Flux Kontext is the first AI image-edit model we trust for production ecommerce and ad work. If you're still doing clipping paths and mask-based inpainting in 2026, this will save you a day's work per week. Not a replacement for a designer — a force multiplier for one.
The base image generation model Kontext edits on top of.
Generation flow that pairs well as pre-input to Kontext edits.
Photoshop's generative fill — Kontext's main competitor.
Vector + brand-safe generation that complements Kontext.