🤖 Leon Pruna API

Pruna - the Munich-based AI startup you've probably seen mentioned in model-optimization circles - has an image model in the mix, and this node is how you reach it from ComfyUI. Leon_Pruna_API_Node is the plainest of the pack's text-to-image wrappers: a prompt, an aspect ratio, a seed, and one API key, and it returns a full-resolution image without your GPU breaking a sweat. Zero VRAM, zero model downloads, zero local inference. Every pixel is generated on the provider's hardware and shipped back to you.

Like the other HyprLab-based nodes in this pack, this is really a thin OpenAI-compatible client. It inherits from HyprLabImageGenerationNodeBase, which means the workflow underneath is: build a payload with model: "p-image" and your settings, POST it to https://api.hyprlab.io/v1/images/generations with a Bearer token, then decode the response into a ComfyUI image. The base class handles retries with exponential backoff (five attempts max), downloads the image if the API returns a URL, and always hands you an RGBA tensor.

The inputs that matter

Honestly, this is the pack's most boring node, and that's a compliment. The required fields are prompt, output_format (png/jpeg/webp), seed, api_url, api_key, and response_format. Two things to actually touch:

• api_key - the tooltip says "Your HyprLab/Pruna API key". If your key came from HyprLab, use it as-is; the endpoint is HyprLab's, and it's the aggregator that decides how to reach the Pruna model.
• aspect_ratio - 1:1, 3:4, 4:3, 9:16, 16:9, 2:3, 3:2, 21:9, 9:21. That's the whole creative control surface. There's no negative prompt and no style preset; you get what the model gives you.

Outputs are the standard triple: image (wire to SaveImage/PreviewImage), image_url, and seed. Set the seed to something specific to reproduce a result.

Installing

Same pack as every Leon node - 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 ComfyUI. Dependencies are light (tenacity, requests, pillow, torch, numpy), and there are no model files to hunt down because the model never touches your machine.

Common issues

The two usual suspects apply. Left-at-default api_key of YOUR_API_KEY_HERE is the #1 cause of a 401. And because response_format defaults to url, the node has to download the generated image before it can hand you a tensor - if that download fails or times out, the retry loop kicks in; wait out the five attempts before you start debugging. If you're getting 4xx errors with the right key, check the console output, where the node prints the exact request and the API's error text. This node is the "just generate me something decent" option - for the same pack's editing or upscaling, use its siblings Leon_Pruna_Image_Edit_Node and Leon_Pruna_Image_Upscale_Node.