Agnes Image 2.1 T2I Config

Agnes is the wildcard in the MediaHub lineup. MiniMax, Seedream, and Wan are names the community knows. Agnes is a hosted image/video API with barely any community chatter behind it - you won't find a crowded subreddit thread about it, and that's exactly the kind of provider this pack exists to bridge. AgnesImage21T2IConfig is the node that packages your text prompt into a request for Agnes's Image 2.1 model.

The key thing to remember: this node describes the image. It doesn't generate one. It builds a MEDIAHUB_IMAGE_REQUEST that you feed into MediaHubGenerateImage, which is the node that actually calls the Agnes API with your key and returns the IMAGE tensor. Get used to that split - it's the whole architecture of this pack, and every provider config node works the same way.

Inputs that matter

• prompt - your text prompt. Simple string, no ComfyUI prompt-engineering tricks needed; this goes to Agnes's server as-is.
• width / height - pixel dimensions, defaults 1024×768, both clamped between 64 and 4096 with a step of 8 (so effectively multiples of 8).
• response_format - url or base64. This decides how the API returns the image. url is the default sensible choice; base64 is useful if you'd rather not have the node make a second download request, but the end result in ComfyUI is the same IMAGE either way.

The single output is request (MEDIAHUB_IMAGE_REQUEST), which goes straight into MediaHubGenerateImage. There are no other outputs - no image, no preview. If you run this node alone, you get a request object and nothing to look at, which is by design.

Why you'd reach for it

Because it's cheap and hands-off. You're not paying for a GPU, not managing checkpoints, not waiting on model downloads. If Agnes's Image 2.1 quality fits the job and you have a key, this is a few clicks to a hosted image inside a ComfyUI graph. It's also a nice way to A/B a closed API against your local stack without leaving ComfyUI - generate locally, generate on Agnes, compare, all in the same workflow.

Installing it

Install the pack, then find the node under MediaHub/Provider/Agnes/Image:

cd ComfyUI/custom_nodes
git clone https://github.com/vantang/ComfyUI-MediaHub.git
/path/to/ComfyUI/python -m pip install -r ComfyUI-MediaHub/requirements.txt

Restart ComfyUI. Requirements are just requests, Pillow, and numpy - no weights to place, no VRAM to reserve. You will need an Agnes account and API key, because nothing about this runs locally.

Common gotchas

The number one mistake is expecting an image out of a config node. Wire it to MediaHubGenerateImage with your key or you get nothing. The other one is dimension mismatch: Agnes validates sizes, and values outside the 64–4096 range (or not aligned to the step) will bounce back as an API error. If you hit an auth error, check that the key you wired in is the Agnes key and not one from another provider - the MediaHub API Key node doesn't know or care which provider a key belongs to, and it won't stop you from mixing them up.