Bria Text2Image Base

Text2ImageBaseNode (shown as Bria Text2Image Base) is Bria's V1 text-to-image generator, and it's the most configurable node in the pack. Where the newer FIBO nodes lean on structured JSON prompts for control, this one gives you the familiar diffusion dials - steps, guidance, aspect ratio, negative prompt - plus some genuinely useful extras: two ControlNet-style guidance slots and an image-prompt input for style transfer. If you want to understand what Bria's API can do before committing to the FIBO workflow, start here.

How it works

It POSTs to /v1/text-to-image/base/3.2 with everything you set, then downloads the result. The controls map onto standard diffusion concepts you already know from local generation:

• guidance_method_1 and guidance_method_2 - two independent ControlNet slots, each with its own method (controlnet_canny, controlnet_depth, controlnet_recoloring, controlnet_color_grid), scale, and image. Wire a Canny edge map into the first slot and a depth map into the second and you get composition control the way you'd use multiple ControlNets locally. Leave the image unconnected and the method is simply ignored.
• image_prompt_mode (regular / style_only) with image_prompt_image and image_prompt_scale - reference-image conditioning. regular follows both content and style; style_only steals the vibe and drops the subject matter. This is the closest thing to img2img here.
• medium (photography, art, none) - a coarse style bias; none (default) leaves it to the prompt.
• prompt_enhancement (0/1) - Bria rewrites/expands your prompt internally. Handy when your description is thin; turn it off when you need your exact wording respected.
• steps_num (default 30), text_guidance_scale (default 5), aspect_ratio (default 4:3), seed (default -1 = random), negative_prompt, content_moderation (0/1).

Outputs

One output: output_image, a plain IMAGE tensor for Preview or Save. Nothing exotic.

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 lives under API Nodes. No models to download, no GPU pressure - it's a cloud call requiring a BRIA API token from platform.bria.ai (paid credit, one billable generation per run).

Notes from the field

Community reports on Bria's V1 pipeline were mixed - usable quality, but a common complaint was that images came out soft and needed real upscaling to look sharp, so budget for an upscale step (the pack's Image Enhance node does this in one call). ControlNet guidance here is a big leg up over plain text2img for anything with layout requirements, since the reference images do the structural work and the prompt just styles it. And remember the same gotcha as every node in this pack: leave api_key as BRIA_API_TOKEN and you'll get an auth error, not a hint.