Bria Reimagine

ReimagineNode (shown as Bria Reimagine) is Bria's general-purpose regeneration node from the V1 generation pipeline. You give it a prompt, optionally point it at an image that defines the structure, and it produces a fresh image that follows your words. It's the node that predates the FIBO/V2 structured-prompt family, and while Bria's docs steer you toward V2 for max control, this one is still the quickest path to "new image from a prompt" in this pack - one API call, ~12 steps by default, done.

How it works

The node POSTs to /v1/reimagine on Bria's engine with your prompt, seed, step count and the fast flag. The interesting part is the optional structure_image: feed it a photo and crank structure_ref_influence (default 0.75) to tell the model how hard to stick to that image's composition while redrawing the content. Drop it and you're doing plain text-to-image. It also accepts the tailored-model inputs, so a trained brand model (see TailoredModelInfoNode) can steer the output style with tailored_generation_prefix - the node literally prepends that prefix to your prompt before sending.

Inputs worth touching

• api_key and prompt are the only required fields. Everything else is optional, which makes this a friendly first node in the pack.
• seed defaults to -1 (random). Set a fixed seed to reproduce a result.
• steps_num defaults to 12 - low for a diffusion model, but Bria's V1 pipeline is distilled to run at these step counts. Leave it alone unless you know why you're changing it.
• fast (0/1) and structure_ref_influence (0.0–1.0) are the dials you'll actually fiddle with: fast for quicker previews, structure_ref_influence down if the source composition is dominating too hard.
• content_moderation (0/1) - Bria trains on licensed data and moderates by default; flip it on only if you're deliberately testing the filter.

Output

One output: output_image, a standard IMAGE tensor. Wire it to a Preview or Save node. Nothing else to juggle - the whole thing is fire-and-forget.

Installing it

It's 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, then find Bria Reimagine under the API Nodes category. No model files, no VRAM pressure, no GPU even - this is a cloud call, so you need internet and a BRIA API token from platform.bria.ai. Each run is billed to that account, so treat it as metered, not free.

Common issues

The big one is forgetting the token - the field is a plain string with no default here, and leaving it blank or as the BRIA_API_TOKEN placeholder gets you an API error. Beyond that, expectations: V1-era Bria output is serviceable but a step behind today's open models - a common community complaint was that Bria images needed real upscaling before they looked sharp. If your reimagine comes out soft, don't fight it in-prompt; upscale downstream (this pack has an Image Enhance node for exactly that) or move to the FIBO nodes for cleaner generations.