Flux Kontext + Custom Models ( Mix models )
Kontext + Custom Models: A Powerful Duo for Visual Creation
Image editing is evolving quickly, and Kontext opens new possibilities by combining contextual image understanding with the power of Custom Models. This synergy enables more accurate, creative, and tailored results.
What is Kontext?
Kontext is a visual editing engine that understands the image as a whole:
- Identify subjects, objects, and backgrounds.
- Modify elements with simple text instructions.
- Build coherent scenes and refine them through iterations.
With Kontext, you don’t need to rewrite everything: the system interprets the image and applies your changes smoothly.
How Custom Models help Kontext
Custom Models enhance Kontext by providing:
- A specific visual style (realism, fashion, cartoon, cinema, etc.).
- Dedicated universes (avatar, products, architecture, storytelling).
- Visual consistency across a series of images.
- Precision, especially on text labels on products
👉 In short: Kontext provides the intelligence, Custom Models add the personality.
Use the Mix model panel to try it!
How to use Kontext + custom models
With your custom Product Model.
1 - Import your image
- My base image
2 - Select your Custom Model
In you Custom Model tab, select your trained model.
3 - Write your modifications
Here I want to remove the hands.
"remove the hand"
Result:
4 - Continuity in your modifications
Hover the image card then select "Reuse Image", it will update Kontext with your new image
Then write your modifications
"closeup on the top of the product. don't change anything else. maintain background and consistency."
5 - Add Multiple Models with Mix Models !
Let's add a Model in the Models Mixer
Select "Create models mix"
Add the models you want to add ( my custom Gold Liquid model )
And let's change the background:
"change the background by a golden liquid background. don't change the product. maintain product consistency. text is "Balm Amour" "Nourishing lip treatment"
Image models doesn't know everything, it's like chatgpt, they know a little bit of everything but are not experts, for example without my Gold custom model :
"change the background by a golden liquid background. don't change the product. maintain product consistency."
That's why create custom model helps you to create precise image.
Advanced Editing Guide
Kontext understands the scene (subject, background, light, perspective) and applies your text instructions contextually.
Custom Models inject style, character identity, brand universe, or consistent cinematic look. Together: precision + consistency.
1) Basics (Quick but Crucial)
Recommended prompt structure
[Main action] + [who/what] + [visual details] + [framing/camera] + [coherence constraints]
Useful constraints
- “keep background,” “same lighting,” “same time of day”
- “same frame,” “same focal length,” “same viewpoint”
- “same character” / “same outfit” / “lock background”
When to load a Custom Model?
- When you want a specific style (fashion, vintage, cinema, illustration) or a recurring character/product.
- Use only one Custom Model at a time to maximize consistency. Add its name in the prompt if needed (e.g., “in style [ModelName]”).
2) Detailed Use Cases
Generate product images with high conformity and quality!
Goal: modify outfit without breaking pose, light, or setting.
Base image | Model | Kontext |
---|---|---|
| | |
Prompt type
Change the subject’s clothes to [specific outfit]; keep pose, expression, and lighting; same frame and background.
Model:
Leather Dress by Givenchy
Replace by Product
Goal: replace product and fill in scene naturally.
Generate Interior Design with Flux | Floor Lamp Model | Kontext |
---|---|---|
| | |
Prompt:
insert a small floor lamp near the sofa
Model:
NV Gallery Floor Lamp
🎥 Update Scene
Goal: Change day time or style with custom model
pt:
Change to night time, clear american night. switch on the light
Try different mix models
Models | Infrared Black | Orange Flat | Psychadelic Shift |
---|---|---|---|
Results | | | |
🤝 Add Character
Goal: Add a Custom Avatar
Image | Avatar Custom | Result |
---|---|---|
| | |
Prompt:
Add [second character: look/outfit]; interaction [handshake/hug/object sharing]; credible postures; consistent shadows between subjects.
"don't change the woman on the sofa. add a woman on the right, seated." - change format to 16:9
Best Practices for Custom Models
- Only one Custom Model at a time to avoid conflicts.
- Describe content before style.
- Adjust style if too strong: “subtle style, prioritize realism.”
Prompt type
[Action] … in style [CustomModelName]; respect visual identity; coherent with scene; realistic material textures.
4) Iteration & Control
- One changement after one
- Lock background: “keep exact background, maintain background.”
- Lock product: "same product, text is "LOGO" "
- Lock subject: “same face, body type, hairstyle.”
- Light continuity: “same time, same shadows.”
- Camera: specify position + focal length (“waist shot, 50mm”).
- Color: “consistent grading, moderate contrast.”
6) Mini Quality Checklist
- ✅ Clean face/hands
- ✅ Stable horizon
- ✅ Uniform colors & grain
- ✅ Logical shadows
- ✅ Credible scale
Updated on: 05/09/2025
Thank you!