π€ Leon Recraft Image API
Recraft's style presets, without the website
- image
- image_url
- seed
Recraft is the model you reach for when the job isn't "photorealistic" but "looks like it was designed" - crisp illustrations, pixel art, hand-drawn line work, the kind of stylistic control most text-to-image models can't hold. Its flagship recraft-v3 lives behind an API, and this node brings it into ComfyUI through the pack's HyprLab base: type a prompt, pick a style from the dropdown, and get back an image with that exact illustration flavor baked in. If you've ever fought a local model for an hour trying to get consistent flat-vector output, you'll get why a node like this exists.
How it works
Same shared machinery as the rest of the pack's API nodes - HyprLabImageGenerationNodeBase builds an OpenAI-style payload, POSTs to https://api.hyprlab.io/v1/images/generations with a Bearer token, retries with exponential backoff (up to five attempts), and decodes the result into an RGBA tensor. What makes this node different is what goes in the payload: a style field and a size field, both as enums, both doing the heavy creative lifting.
Inputs that matter
style- the reason to use this node. 23 options, defaultrealistic_image, but the interesting half is thedigital_illustrationfamily:pixel_art,hand_drawn,grain,infantile_sketch,2d_art_poster,handmade_3d,hand_drawn_outline,engraving_color,2d_art_poster_2, and more. Pick the illustration style here instead of describing it in the prompt; that's what the model is actually good at.size- 15 options from1024x1024up to2048x2048-class dimensions (1024x2048, 2048x1024, etc.). Recraft sizes are preset, not arbitrary - pick the closest to what you need.prompt- capped at 10,000 characters per the tooltip, and it's the usual "what's in the scene" description. Keep the style words out; thestyleenum owns that job.
Outputs: image, image_url, seed - the pack's standard triple.
Installing
ComfyUI Manager (search Leon) or clone:
cd ComfyUI/custom_nodes/
git clone https://github.com/l3ony2k/comfyui-leon-nodes comfyui-leon-nodes
pip install -r requirements.txt
Restart and paste a HyprLab key. The dependencies are light and there are no model files - this node is a remote render, so your GPU stays idle while it works.
Common issues
The standard API-node failures apply - swap the default YOUR_API_KEY_HERE key, watch the console for the logged request when a 4xx comes back. Style-wise, the two traps are overlapping: putting "pixel art" in the prompt and selecting pixel_art in the enum usually fights the model, and the default output format here is webp, not png - if a downstream node chokes, flip output_format to png. And because the style presets are the point, experiment there first before touching the prompt; two minutes of dropdown-browsing teaches you more about recraft-v3 than an hour of prompt edits.
Inputs (9)
| Name | Type | Default | Description |
|---|---|---|---|
| prompt | STRING | cat | Main text that influences the image generation (max 10,000 characters) |
| model | COMBO | recraft-v3 | Recraft AI model to use for generation |
| output_format | COMBO | webp | Format of the output image |
| seed | INT | 00β18446744073709550000 | Random seed for reproducible results |
| api_url | STRING | https://api.hyprlab.io/v1/images/generations | API URL |
| api_key | STRING | YOUR_API_KEY_HERE | Your HyprLab API key |
| response_format | COMBO | url | How the response data should be formatted |
| sizeopt | COMBO | 1024x1024 | Dimensions of the generated image |
| styleopt | COMBO | realistic_image | Artistic style or filter applied to the generated image |
Outputs (3)
| Name | Type | Description |
|---|---|---|
| image | IMAGE | β |
| image_url | STRING | β |
| seed | INT | β |