Aabao 图像生成

AabaoImageGen is the odd one out in a video pack: a plain text-to-image node. It calls the Aabao reseller's Sora image endpoint (sora-image models) to generate stills - the kind of thing you can already do locally with Stable Diffusion or Flux for free, but useful here because it's the same account and workflow you're already using for video, and because it produces frames you can feed straight into the pack's image-to-video nodes. Think of it as a storyboard front-end: write a scene description, get four candidate frames, pick one, then run AabaoImage2Video on it.

How it works

Same API-wrapper pattern as the rest of the Aabao nodes: prompt and parameters get POSTed to the provider's /images/generations endpoint, the returned image URLs (or base64) come back, and the node downloads them into your ComfyUI output folder. Because it's the Sora-image model behind the reseller, output resolution is fixed and modest - the size dropdown only offers 360x360, 540x360, and 360x540 (square, landscape, portrait). These are small frames, good for storyboarding and reference, not for final art.

Inputs and outputs

• prompt - multiline text; the default is "一只可爱的小猫" (a cute kitten). Be specific about subject, composition, lighting.
• model - sora-image, sora-image-landscape, sora-image-portrait. Match this to your size choice; there's no automatic lock between them.
• size - the three resolutions above.
• n - 1–4 images per run, default 1. This is the standout: crank it to 4 and you get a quick contact sheet of options for one prompt, which is exactly what you want before committing to a video generation.
• api_provider / api_key - provider is fixed to aabao; key from config.json or the node field.

Outputs are just two: images (an IMAGE batch - so 4 images come out as a batch of 4) and status (a STRING with the result/error message). It's marked as an output node, so it doesn't feed further nodes by default - but you can wire images into an image-to-video node.

Install

ComfyUI Manager → search "ComfyUI_Sora", or:

cd ComfyUI/custom_nodes
git clone https://github.com/xuchenxu168/ComfyUI_Sora
cd ComfyUI_Sora && pip install -r requirements.txt

Restart and make sure aabao_api_key is in config.json or fill the node's key field. No model files to download.

Common issues

• Images come out tiny - yes, 360px-ish is the whole range on this endpoint. Use it for storyboard/consistency work, not hero images; if you need a bigger still, generate locally and bring it in.
• n=4 but only one image - the status output will usually tell you. Some provider/model combos ignore n or cap it; check status for the JSON error.
• Character consistency across the batch - don't expect it. Each of the n frames is an independent draw; if you need the same face across frames, generate one good reference still and drive everything from it via the image-to-video node.

Honestly? For image generation you already have better local options. Where this node earns its keep is as the fast, in-pack way to draft storyboard frames on the same account you're paying for video.