Seedream 4 Batch Generation

Most batch generation in ComfyUI is just the same node run N times - N separate generations that happen to share a prompt. Seedream 4 Batch Generation is different: it's a single API call that asks the model for a set of images that hang together. ByteDance's API calls this sequential image generation, and the node exposes it as max_images. You get a thematically related series - a character in different poses, a location at different angles, a storyboard of one scene - from one request, which is both faster and cheaper than firing off separate calls. This is the node to reach for when you want a narrative batch, not variations.

It's a Seedream 4 (seedream-4-0-250828) node, so the usual rules apply: API-only model, paid per image, no local GPU (the KB's ByteDance panel explains the whole "products stay on the API" posture). What's unusual here is the batch semantics.

How it works

The node builds a payload with sequential_image_generation: auto and a max_images option, then sends it to the images endpoint. "Auto" means the model decides how many images the sequence deserves, up to your cap. Your prompt is the thread that ties them together. If you connect image1 and/or image2, they act as references that the whole batch should stay consistent with - useful when you want a set built around an existing character or style. The response images are downloaded and stacked into a single batched IMAGE tensor.

Inputs and outputs that matter

• prompt - describe the shared subject and the range you want covered ("same character, three poses on a plain background").
• max_images - cap on how many the model can produce, 1–10 (default 3). Remember each one is a paid image.
• width / height - output size, default 2048×2048, up to 4096.
• image1 / image2 - optional references for consistency.
• seed - reproducibility, -1 for random.

Output is one IMAGE tensor holding the whole batch - split it with a batch node if you need each image separately, or save the stack directly.

Install

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

Restart, drop your BytePlus API key in config.ini or set BYTEPLUS_API_KEY. Manager: search "ComfyUI Seed API Integration".

Where people get burned

Two things to keep straight. First, "auto" means the count is the model's call - max_images is a ceiling, not a guarantee, so don't wire expectations around getting exactly three. Second, this is thematic batch generation, not a batch of variations. If what you actually want is four versions of the same prompt to compare, the Seedream 4 unified node's parallel mode is the right tool; this node is for related-but-different. And the usual cloud gotchas apply: a black 512×512 image is the pack's error fallback (check the console), and a 10-image batch is a real line item, so check your BytePlus usage dashboard if the bill surprises you.