Model overview
Flux Kontext
Long-prompt semantic expert
Kontext understands up to 8K tokens of context, making it perfect for worlds, storyboards and chapter-based visual scripts.
- Context length
- 8192 tokens
- Recommended use cases
- Stories / worldbuilding
- Modes
- Text · Reference · Storyboard
Production ready
Flux Kontext Beta · 2024.11
Vendor
Flux AI
General notes
Every generation request is synced to your history so you can reuse prompts or iterate any time.
Recommended use cases
- Long memory
- Chapter output
- Storyboard exports
Generation console
Switch between Text-to-Image and Image-to-Image without leaving this page.
Key features
Key Features
Why teams pick Flux Kontext for their campaigns.
Detailed capabilities
Detailed Capabilities
Each capability ships with prompt samples and reference art so you can replicate results quickly.
Chapter-based descriptions yield unified color palettes.
Prompt example
Scene 1: Observatory above the desert. Scene 2: Underground archive with holographic scrolls. Character: historian Talan.
Output
Multi-scene illustrations with shared style.
Storyboard mode outputs cinematic 16:9 frames.
Prompt example
Shot 05 close-up of Mira receiving a message. Shot 06 wide shot of elevated train above flooded streets.
Output
Two frames with continuous visual language.
Keep character identity intact across different scenes.
Prompt example
Character dossier: Captain Ilya, silver hair, asymmetrical armor, calm demeanor. Render her briefing squad near icy cliffs.
Output
Consistent character renders from multiple angles.
How-to guide
Usage guide
1. Choose the model
Select Flux Kontext from the catalog and review the suggested storytelling ratios.
2. Enter prompts
Structure prompts with chapters or upload storyboards for reference-driven renders.
3. Click Create
Inspect the multi-frame output, tweak descriptive sections and regenerate as needed.
Social proof
Social media reactions
YouTube
“"Kontext finally lets me pour an entire chapter of lore into one prompt."”
Open thread“"Wrote a 900-word lore dump and it kept sigils consistent over 5 scenes."”
Open threadX
“"Writers can finally iterate visually without breaking continuity."”
Open threadOther models
Other models to explore
Jump to another image model in a single tap.
BFL's photography-grade model
High-resolution expert for Chinese prompts
FAQ
FAQ
Answers describe how this model behaves inside Nano Banana2.
CTA
Write the story once, render it all
Use long prompts and chapter tags so Kontext can auto-generate a cohesive visual script.