VolcengineImg2Img(NYJY)

If you've already met Seedream4Img2Img in this pack, this node will feel like deja vu - because in the source, Seedream4Img2Img is just a subclass of VolcengineImg2Img with a different menu name. Same inputs, same API call, same outputs. The pack gave the generic wrapper a dedicated Seedream 4.0-branded twin, and both work identically.

It's part of ComfyUI_NYJY (aidenli's pack), sitting on the pack's AIModelBridge layer that routes everything through the Volcano Engine Ark API. The img2img nodes are the "reference + prompt" generation: you feed an image, and Seedream re-renders it according to your instruction.

How it works. Every image in your input batch is converted to base64 and sent to the Ark images.generate endpoint alongside the prompt; the returned images are decoded back into a ComfyUI tensor. The default model is doubao-seedream-4-0-250828, and the model field being free text means the newer doubao-seedream-4-5-251128 from the pack's own list works too. Where this differs from the txt2img node is purely the images input - everything else is shared plumbing.

The inputs that matter:

• images - your reference batch. This is the input the whole node exists for.
• prompt - what changes about the input image: restyle, change the setting, redraw the subject.
• ratio_or_size - the eleven options (2K aspect presets like 1:1 2048x2048, plus 1K/2K/4K). Note the output size is chosen here, not inherited from the input.
• override_with / override_height - set both for exact pixels instead of a preset.
• max_images - 1–15; how many variations to generate per call.

The outputs. Image (a batch if max_images > 1), plus Width and Height from the first result. Standard trio across the pack.

The gotchas. The family traits apply: an API failure returns a black 512×512 tensor while the real error lands in the ComfyUI console - black output means "read the log". You need a Volcano Engine Ark key, and non-China sign-up has historically been the blocker, not the install. And since this is paid per generated image, treat max_images as a cost control. For precise instruction-editing (removing a watermark, changing one element), Seededit3 in the same pack is the more appropriate tool than this node.

Install. Ships with the pack:

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

or ComfyUI Manager → "Install via Git URL" → https://github.com/aidenli/ComfyUI_NYJY. The dependency that matters is volcengine-python-sdk[ark].

When to reach for it. Identical situation to Seedream4Img2Img - use whichever name you found first. If you're loading a workflow that references the generic name, this is the node behind it.