Qwen3VL Flux Text Styler + Simple Text Overlay Workflow
Smart workflow for stylizing white text to match the background of an image using Qwen3-VL + Flux.
The Qwen3VL Flux Text Styler node analyzes the image and automatically generates a detailed prompt that turns plain white text into volumetric, textured letters that perfectly match the lighting, materials and atmosphere of the scene.
Installation
Unpack the archive
Place the ComfyUI-qwen3vl-textstyler folder into your ComfyUI/custom_nodes directory
Restart ComfyUI
The archive also includes a simple ready-to-use workflow that uses two of my other nodes:
Simple Text Overlay → https://civarchive.com/models/2711044/simple-text-overlay-node
ZImageResolutionScaler (from Z-Image Resolution Picker) → https://civarchive.com/models/2344436/z-image-resolution-picker
How to use (important!)
Do not run the full workflow right away.
Load your image into the Load Image node
In the Simple Text Overlay node write your text, choose font, size, position, shadow etc.
To quickly check text placement — select only the Simple Text Overlay node (or the small group with it) and press the button that runs only selected nodes (the one highlighted with a red square on the screenshot). This is almost instant and barely loads your computer.
When the text looks perfect — press the main Run / Queue Prompt button to run the full workflow.
This way you avoid unnecessary load and can comfortably adjust the composition first.
What the Qwen3VL Flux Text Styler node does
The node builds a system prompt for Qwen3-VL. The model looks at the image with white text and outputs a ready Flux prompt that stylizes only the letters, leaving the background completely untouched.
Main settings:
• prefix — Mandatory start of the prompt (Transform the white text into... or Stylize white text into...)
• preserve_* — Keep background, colors, lighting, shadows, reflections, atmosphere, composition and original text layout unchanged
• force_volumetric — Prefer solid 3D volume. If the background clearly looks like paper/poster/sticker, the model is allowed to use a flatter treatment
• force_scene_interaction — Letters should interact with the scene (roots, sparks, smoke, cables etc. can wrap around or come out of the text)
• force_readable — Keep letter shapes clearly readable
• allow_particles_outside — Allow particles/effects to slightly go beyond letter boundaries
• extra_creativity — Bans boring generic descriptions and pushes for more artistic solutions
• protect_human_subjects — Do not affect faces, poses or clothing of people
• style_mode — Force a style: Tech/Industrial, Soft/Cinematic, Organic/Nature, Fantasy/Magic or Auto
• material_hint — Suggest the main material (metal, stone, ice, neon, wood, etc.)
• decoration_hint — Add decoration: vines, cracks, moss, cobwebs, rivets, glowing edges, etc.
• action_hint — Effect: burning, melting, frozen, dissolving, exploding, etc.
• scene_interaction_hint — Your own interaction description (example: “vines wrapping from the left side”)
• intensity — Strength of stylization (0.2 = subtle, 2.0 = maximum detail)
• compact_mode — Short or full system prompt
• negative_constraints — What is strictly forbidden (example: no chrome, no glowing edges, no cartoon style)
• extra_instruction — Any additional notes
• randomize + seed — Randomly pick material, decoration, action and style
Tips
• Always position the text first using partial execution of Simple Text Overlay • For clean results keep most preserve_* options enabled • If letters come out too flat or too heavy — adjust intensity and force_volumetric • negative_constraints is very useful when the model keeps adding unwanted effects
You can use the full workflow as is, or just the Qwen3VL Flux Text Styler node in your own pipelines.
Enjoy!