FLUX.1 Kontext

Overview

AI image tool set called FLUX.1 Kontext is by Black Forest Labs. With it, you can use it to make or change pictures using text or other images. 

It mixes inputs. You can use both text and images to get the results you want. That means you can just type what you're thinking or show a sample and it’ll work with both.

You can edit small details. Want to fix just one part of the picture? FLUX.1 Kontext lets you do that without touching the rest.

Characters stay the same. This is big. Your characters don’t suddenly look like strangers in each new image. Their look stays solid through every edit. This is the most important thing: FLUX.1 Kontext precisely preserves identity (of e.g. a reference character or object) across multiple scenes and environments. Where ChatGPT will slightly change the face as you introduce modifications, this thing keeps identities truly consistent.

It’s fast. We’re talking up to 8x quicker than others out there. You get results in like 3 to 5 seconds for 1MP images. That’s almost instant.

You can build as you go. Do a little now then tweak it later. It keeps things clean so the quality stays sharp each time you go back and add more. 

Flux Kontext can be used for image restoration by uploading a photo and entering a detailed prompt like: "Restore and colorize this image. Remove any scratches or imperfections." It performs quite well in preserving character identity during restoration and colorization.

Different versions for different needs:

Kontext [Pro]. Best for when you're doing edits step by step. It supports both text-to-image and sharp touch-ups.

Kontext [Max]. This one’s still in the works but already shows strong prompt results clean fonts and high render quality without slowing down.

Kontext [Dev]. Open weight 12B model you can tinker with if you’re into custom setups. But local free install won't give you a commercial licence, only for personal use. API providers have license, so if you use the model using the BFL API - you're paying the license cost and the images are ok to use  commercially. This model also provides scripts for inference and LoRA tuning utilizing the Diffusers library. An official demonstration is available on Hugging Face Spaces, accompanied by two community demonstrations focused on multi-context editing and LoRA exploration. The team is closely observing usage patterns and may impose bans on developers who consistently breach usage policies via the FLUX API.

How to access FLUX.1 Kontext

You can use the FLUX.1 Kontext models through the BFL API or with tools like Freepik, Krea AI, Lightricks, OpenArt and LeonardoAI. There's also the 👉 BFL Playground a browser demo you can try right now for free (you get 200 free credits upon signing in with email) to test what the models can do.

Multi-image Kontext

For combining 2 images together, there's an experimental version currently available on Replicate at $0.08 per image and on Freepik (see my example below)

ComfyUI has also shared a method to use multiple images with FLUX.1 Kontext by utilizing their Image Stitch node. This approach can help with tasks like object editing, character addition, and image blending.