How Flux prompts work
- Write Flux prompts as natural sentences, not comma-tag soup — describe the scene the way you’d describe a photograph.
- Negative prompts are largely unnecessary: Flux.1’s base pipeline mostly ignores them, so put your effort into a vivid, specific positive prompt.
- Flux renders legible in-image text well — put the exact words in quotes, e.g. a sign reading "RAMEN".
- Weighting and LoRAs are applied in your UI (ComfyUI nodes) rather than inline — keep the prompt itself clean and descriptive.
Example Flux prompt
A cozy cyberpunk ramen shop at night in the rain, warm neon signage reflecting on wet pavement, steam rising from bowls, a sign reading "RAMEN", cinematic, shallow depth of field
Build Flux prompts faster with Prompt Builder
- Snippet library for recurring scene elements, lighting and styles.
- AI assist turns a one-line idea into a full, structured Flux-style description.
- Save your best Flux scene recipes and share them as one-click workflows.
- A live token counter keeps long descriptive prompts in check.
Frequently asked questions
Do I need a negative prompt for Flux?▾
Usually not — Flux.1’s base pipeline ignores negative prompts. Spend your effort on a vivid, specific positive prompt instead.
How is prompting Flux different from SDXL?▾
Flux prefers natural-language sentences over comma-separated tags, renders in-image text reliably, and needs little to no negative prompt.
Can Flux put text in an image?▾
Yes — write the exact words in quotes in your prompt. Flux is far better at legible text than most Stable Diffusion models.
Keep exploring
- Prompt weighting & LoRA syntax guide
- Prompt glossary
- Other models: SDXL · Midjourney · Pony Diffusion