Seed Volcengine T2I Config
ByteDance's Seedream, piped into ComfyUI
- request
Seedream is ByteDance's flagship image model, and the honest truth about ByteDance is that its flagship generation products are API-only, full stop - Seedream and Seedance have never shipped open weights. So if you want to poke at the same image model that powers the Doubao ecosystem, a hosted API is the only door, and SeedVolcengineT2IConfig is how that door connects to ComfyUI.
It builds a text-to-image request for the Volcengine ARK endpoint (ark.cn-beijing.volces.com), authenticated with an ARK API key. Output is a MEDIAHUB_IMAGE_REQUEST that goes to MediaHubGenerateImage, which makes the actual call and returns the IMAGE.
Inputs that matter
- model - three Seedream versions, and the names encode their release dates:
doubao-seedream-5-0-260128(the current flagship),doubao-seedream-4-5-251128, anddoubao-seedream-4-0-250828. Start with the 5.0 unless you're testing regressions. - size -
1K,2K,3K, or4K. This is the standout feature here: Seedream genuinely renders up to 4K, which is far beyond what the other image providers in this pack offer. 4K also costs more and takes longer, so it's a deliberate trade, not a default. - prompt - your text.
- watermark - off by default. Turn it on if you want ByteDance's AIGC watermark baked in (some commercial pipelines require it).
- seed - defaults to
-1, which means "random" in this API's convention. Set a non-negative value for reproducible takes. - response_format -
urlorbase64.
Output: request → MediaHubGenerateImage.
Why you'd reach for it
The 4K output is the reason. If a job needs real resolution - print work, big canvases, detail that survives cropping - none of the other image configs in this pack get there. Seedream's text rendering is also famously decent, which makes it a solid choice when your prompt contains readable words.
Installing it
Same pack install as everything else:
cd ComfyUI/custom_nodes
git clone https://github.com/vantang/ComfyUI-MediaHub.git
/path/to/ComfyUI/python -m pip install -r ComfyUI-MediaHub/requirements.txt
Restart ComfyUI; node under MediaHub/Provider/Seed Volcengine/Image. Dependencies are just requests/Pillow/numpy. You need a Volcengine ARK account and an ARK_API_KEY - that's a ByteDance/Volcengine signup, separate from every other provider here.
Common gotchas
The 4K temptation is real but pricey - a 4K generation is a different cost tier, so don't leave size on 4K for everyday iteration. And the ARK key is scoped to your Volcengine project; if you get permission errors, it's usually the key's project access, not the node. Watch the seed = -1 convention too: the other providers in this pack use different seed conventions, so a workflow you ported from another provider may carry a seed value that behaves differently here.
Inputs (6)
| Name | Type | Default | Description |
|---|---|---|---|
| prompt | STRING | — | |
| model | COMBO | 3 options: doubao-seedream-5-0-260128, doubao-seedream-4-5-251128, doubao-seedream-4-0-250828 | |
| size | COMBO | 4 options: 1K, 2K, 3K, 4K | |
| response_format | COMBO | 2 options: url, base64 | |
| watermark | BOOLEAN | false | — |
| seed | INT | -1-1–4294967295 | — |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| request | MEDIAHUB_IMAGE_REQUEST | — |