NS WaveSpeed Seedream V4 Edit Sequential

If you've ever wanted to run the same instruction across a stack of reference images - "recolor these ten product shots", "put all four character refs in the same style" - the single-image Seedream edit node makes you queue ten times. This one, NSWaveSpeedSeedreamV4EditSequential from symbiotica-ai/comfyui-nodes, folds that loop into one node: up to ten input images, one prompt, one API call per image, and a batched IMAGE tensor out the other side.

How it works

Same engine as its single-image sibling - ByteDance Seedream V4, hosted on WaveSpeed's API, no local model and no VRAM. The difference is plumbing. Instead of one image_url, you get ten optional inputs (image_1 through image_10), all of them URL strings meant to come from ComfyUI's Upload Image node. Each one you wire in gets edited by the shared prompt at the chosen size_preset, and the results come back stacked as one batch on the output_images output - exactly what a downstream grid, contact sheet, or per-item save loop wants to eat.

max_images (1–15) caps how many results you ask for. The author's tooltip is unusually specific about the catch: the number has to match what your prompt says. If you say "a series of" in the prompt but set max_images to 3, the model gets told one thing and the node expects another, and you'll get a mismatch instead of a nice batch. This node is edit-first - the prompt should describe the edit you're applying to each reference, and the series phrasing keeps the results feeling like one set rather than ten strangers.

The inputs that matter

• image_1 … image_10 - your reference images as URL strings (Upload Image node → here). Leave unconnected and they're just skipped.
• prompt - the edit, phrased as a series: "apply the same neon-noir grade to each image."
• max_images - how many results you expect; align it with the prompt.
• size_preset - the nine Seedream V4 presets, up to 4096x4096.
• seed - 0 means random on this API; set non-zero for reproducible edits.
• client - the WAVESPEED_AI_API_CLIENT socket from the pack's NS WaveSpeed Client node.

Output is output_images (IMAGE, batched). Wire it to a save loop or an image comparer.

Install and keys

This ships in the same pack as all the NS WaveSpeed nodes, so it's one install:

# ComfyUI Manager: search "Symbiotica", or:
cd ComfyUI/custom_nodes
git clone https://github.com/symbiotica-ai/comfyui-nodes.git symbiotica
pip install -r symbiotica/requirements.txt

Restart, add the NS WaveSpeed Client node, and put your wavespeed.ai key in its api_key field (or once in Settings → Symbiotica, or as WAVESPEED_API_KEY). Wire client into here. No model files, no heavy deps.

Common issues

Prompt says three, max_images says ten. You'll get a result that ignores your count or errors. Treat the two fields as a contract and keep them in sync.

Nothing comes out because the client node is missing. Every WaveSpeed node in this pack requires that client wire - the key lives there, not on this node. If you see "API_KEY is empty", that's the cause.

And the standing caveat for the whole family: this is a paid cloud call, your references leave your machine, and ByteDance's content filter applies to whatever you send (external-api-nodes.md). Great for batch production work where consistency across a set is worth the money; a waste for a single throwaway edit, where the plain Seedream edit node is the cheaper, simpler route.