NS WaveSpeed Seedream V4

Seedream is ByteDance's flagship image model, and "flagship" is doing real work here: it's API-only, with no open weights and - per the bytedance panel - no exceptions in the company's whole product line. This node is WaveSpeed's hosted version of Seedream V4's text-to-image mode, which is the only way it's ever getting into your ComfyUI graph.

It's part of the Symbiotica pack's neuralsins/WaveSpeed family. Where the Nano Banana nodes in the family top out at 1:1–21:9 and offer 4K or 8K tiers, Seedream V4 does things in its own register: resolution presets that go up to a 4096×4096 native square, and 21:9 built in. It's the node to reach for when you want ByteDance's image quality and a genuinely large canvas.

How it works

prompt in, output_image (IMAGE) out, via WaveSpeed's /api/v3/bytedance/seedream-v4. The controls are a deliberately short list:

• prompt - the image description. Seedream is a strong instruction-follower with good text rendering; give it a real sentence.
• size_preset - nine resolution presets, and they're the story here: 2048x2048 (1:1) (default), 2304x1728 (4:3), 1728x2304 (3:4), 2560x1440 (16:9), 1440x2560 (9:16), 2496x1664 (3:2), 1664x2496 (2:3), 3024x1296 (21:9), and 4096x4096 (1:1). That last one is the native-4K-square draw - no upscale pass required.
• seed - defaults to 0 (not -1 like the other nodes), with a seed control widget. Set a value for reproducibility.
• enable_sync_mode - defaults to off, matching the video nodes: submit and poll, because these big renders take a while.

Installing

cd ComfyUI/custom_nodes
git clone https://github.com/symbiotica-ai/comfyui-nodes.git symbiotica
pip install -r symbiotica/requirements.txt

Restart or use ComfyUI Manager (search Symbiotica). Feed an NS WaveSpeed Client node with your WAVESPEED_API_KEY into client.

Common issues

• enable_sync_mode defaults to off. Unlike the image nodes, you get a task id and a poll, not an inline wait. If you expected the node to just return an image, flip the toggle on.
• 4096×4096 is heavy. Native 4K output means a big tensor and a real wait (and a real meter). Use the smaller presets for drafts.
• Seed defaults to 0. If you're getting the same-ish image every run and wondering why, that's the fixed seed. Set a different value or treat 0 as your starting point.
• The model's filter, not the node's. ByteDance's hosted product is moderated; it refuses what it refuses, and no setting bypasses that. (The bytedance panel's NSFW note makes the same split clear - the open ByteDance releases are another story entirely, but Seedream itself is a hosted, filtered product.)

The honest take: this is the node for ByteDance's specific strengths - big native resolutions and 21:9 - with the family's usual caveat that every run is a paid call. If your deliverable is "square, huge, from a text prompt," this may genuinely be the best single node in the pack for it.