Nodes/BRIA AI API nodes/Bria Restyle Portrait
ComfyUI Node

Bria Restyle Portrait

Your face, a brand's look

By Bria-AI·Created 2 years ago·Updated 3 months ago· 95
Bria Restyle Portrait
  • images
  • output_images
tailored_model_id
api_keyBRIA_API_TOKEN
seed123456
tailored_model_influence0.90
id_strength0.70

TailoredPortraitNode (shown as Bria Restyle Portrait) takes a portrait and restyles it into a trained model's look while keeping the person's face intact. It's the identity-preservation trick that makes commercial AI portraits useful: you train a tailored model on a brand's illustration style or a product's aesthetic, then run a real person's photo through this node and they come out the other side in that style, still recognizably them.

This is Bria's answer to the "keep the face, change everything else" problem that people normally solve locally with IPAdapter FaceID + LoRA stacks. The difference is the whole job happens in one API call to /v1/tailored-gen/restyle_portrait, with the identity and style tuning exposed as two dials.

How it works

You feed it the portrait (images), the trained model's ID (tailored_model_id), and your api_key. Internally it sends the face as an id_image_file plus the style model and two control values:

  • id_strength (default 0.7) - how hard to hold onto the person's identity. Crank it up when the restyle keeps drifting the face into someone else; back it off when the face starts looking paste-on.
  • tailored_model_influence (default 0.9) - how hard the trained style model pushes. Lower it if the style is drowning out the photo's original lighting and detail.

Plus seed (default 123456) for reproducible results. It loops over the batch, one call per portrait. One output: output_images, a list of tensors - no surprises there.

Installing it

Part of the BRIA AI API nodes pack (Bria-AI/ComfyUI-BRIA-API). ComfyUI Manager → search BRIA API → install, or:

cd ComfyUI/custom_nodes
git clone https://github.com/Bria-AI/ComfyUI-BRIA-API

Restart ComfyUI; it's under API Nodes. No models download and no GPU is required - it's a cloud call, so you need a BRIA API token from platform.bria.ai and credit on the account. The tailored_model_id must be a trained model on your account (use TailoredModelInfoNode to check it and pull its settings).

Practical notes

The batch failure mode is worth knowing: if one portrait errors, the node substitutes the original input image rather than failing the run, so a "failed" restyle can look like the model just didn't apply the style. Check each output instead of trusting the count. Also, identity preservation has limits - id_strength is a scalar, not a magic wand, and extreme style gaps (photo-real to heavy illustration) will stretch what any single pass can hold. If the face drifts, raise id_strength before you blame the model, and keep seed fixed while you tune so you're comparing like for like.

CategoryAPI Nodes

Inputs (6)

NameTypeDefaultDescription
imagesIMAGE
tailored_model_idSTRING
api_keySTRINGBRIA_API_TOKEN
seedoptINT123456
tailored_model_influenceoptFLOAT0.90
id_strengthoptFLOAT0.70

Outputs (1)

NameTypeDescription
output_imagesIMAGE