Concurrent Submit | Zhenzhen Doubao SeedEdit2.0
Doubao SeedEdit 2.0, batched — with the built-in logo stamp
- image
- task
Doubao SeedEdit is ByteDance's image-editing model - you give it one image and an instruction, it returns an edited version. This concurrent wrapper around ComflySeededit adds two things worth caring about: batching (queue a pile of edits without freezing the graph) and an honest-to-god watermark system built right into the node.
That last part is unusual enough to call out. Most API packs leave watermarking to you. This node has add_logo, logo_position (右下角/左下角/左上角/右上角 - the four corners), logo_language (中文/英文), and logo_text. Flip add_logo on, type your brand, and every output comes back stamped. If you're generating listing or social content at volume, that's a feature you'd otherwise build a whole post-processing step for.
Mechanically it's the pack standard: shared image pool (30 workers), COMFLY_IMAGE_FUTURE task output, and you collect with Concurrent Collect Images (ComflyConcurrent_Image_Await) - up to 30 tasks, returned as image_1…image_30 in slot order.
The inputs
- image - the source image. Required; there's no text-to-image mode here. This is an edit node, full stop.
- prompt - what to change. Multiline.
- scale - edit strength, 0 to 1, default 0.5. This is the dial you'll actually touch: low scale = subtle touch-up, high scale = aggressive rewrite. Start at 0.5 and move in 0.1 steps.
- seed - -1 (random) to 2147483647.
- add_logo / logo_position / logo_language / logo_text - the watermark suite described above.
- api_key and skip_error - the usual.
Install
cd ComfyUI/custom_nodes
git clone https://github.com/T8mars/Comfyui-zhenzhen
Or ComfyUI Manager → search Comfyui-zhenzhen → restart, token into api_key (ai.t8star.org overseas; api.seedance.nz for the domestic discount).
Gotchas
- Scale is a strength knob, not a slider you max out. Crank to 1.0 and SeedEdit will happily redraw most of the image. For "keep everything, change one thing," 0.3–0.5 is the sane zone.
- Chinese-language defaults. The logo position/language enums ship in Chinese (
右下角,中文). If the UI looks odd, that's why - the dropdowns work exactly as labeled, just pick the English option if you want it. - Every edit is a paid API call with an upload. Batch math matters: N edits = N generations + N image uploads. And as with everything in this pack, a timeout isn't proof of death - check the async-task page before resubmitting.
Inputs (10)
| Name | Type | Default | Description |
|---|---|---|---|
| image | IMAGE | — | |
| prompt | STRING | — | |
| scale | FLOAT | 0.500–1 | — |
| seed | INT | -1-1–2147483647 | — |
| api_keyopt | STRING | — | |
| add_logoopt | BOOLEAN | false | — |
| logo_positionopt | COMBO | 右下角 | 4 options: 右下角, 左下角, 左上角, 右上角 |
| logo_languageopt | COMBO | 中文 | 2 options: 中文, 英文 |
| logo_textopt | STRING | — | |
| skip_erroropt | BOOLEAN | false | 开启后,节点失败时不报错、按旧行为返回默认空结果;关闭时(默认)失败直接抛出错误。 |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| task | COMFLY_IMAGE_FUTURE | — |