Bria Image Increase Resolution

Some nodes take a whole essay. This one takes a dropdown. Bria Image Increase Resolution is a generative upscaler with exactly one meaningful setting - upscaleFactor, which is 2x or 4x - plus output format and quality. You feed it an image, it comes back bigger, and the actual upscaling model runs on Runware's servers. No model file to download, no VRAM budget to worry about, no waiting on a tiled-diffusion chain to fit 4K in 8 GB.

Worth being clear about what "generative upscaler" means here, because the community's biggest upscaling argument is two people doing different jobs. This is the "more detail" family, not the "more pixels" family. A 2x or 4x generative pass will invent texture and sharpen soft edges; it will also, on a face you care about, change it slightly, because adding detail is generation. If your source is already sharp and you just need it bigger, a cheap pixel upscaler (Lanczos, ESRGAN) is the correct answer and this node is the wrong tool. If you want detail recovery on a soft image and don't want to babysit a local generative upscale, this is exactly the thing.

Inputs and output

• image - required IMAGE. The thing you want bigger.
• upscaleFactor - the whole show: 2 or 4. 2 doubles both dimensions, 4 quadruples them. Pick 4 only when you actually need the pixels; it costs more per request and there's more room for the model to invent things you didn't ask for.
• providerSettings.bria.preserveAlpha - keeps an existing alpha channel in the output. Only matters if your input already has transparency.
• outputFormat - JPG/PNG/WEBP, default JPG. For an upscale you're going to keep, PNG is the honest choice; JPG default is fine for previews but recompresses the very detail you paid to add.
• outputQuality, ttl - leave alone.

Output is a single IMAGE socket. Wire it into Save Image (or back into a downscale/composite node - upscaling before a hi-res composite is a classic move).

Install

Standard Runware pack install, once for the whole family:

cd ComfyUI/custom_nodes
git clone https://github.com/Runware/ComfyUI-Runware
pip install -r ComfyUI-Runware/requirements.txt

or ComfyUI Manager → search Runware. Restart, then put your Runware API key in Settings → Runware API key (or RUNWARE_API_KEY). Dependencies are light (runware-sdk, pillow, soundfile) because nothing runs locally. Your image goes up as a PNG request and the upscaled result comes back as an IMAGE tensor.

The practical gotcha

The cost-per-image model means the price of a 4x pass is real, and it's higher than 2x. For a one-off hero image that's fine; for a batch of 50 you might feel it. Also - and this is the one people hit - the node returns a single image, not a batch of upscale candidates, so don't wire a batch of input frames in expecting a stack back. Feed it one image, get one image, and keep a local pixel upscaler around for the "I just need more pixels" cases this node is overkill for.