LayerUtility: Jimeng Imgae to Image API (Advance)
Jimeng I2I API — image editing via ByteDance's Jimeng/Volcengine service
- image
- image
An image-to-image wrapper around Jimeng (即梦, sometimes marketed as "Dream Map" or "Instant Dreaming"), which runs on ByteDance's Volcano Engine cloud. If you don't recognize the service, that's normal outside China - this is a mainland-Chinese cloud API, distinct from the Google/DeepSeek/Zhipu APIs elsewhere in this pack, and it's the one node in this family where signup friction is the real barrier rather than the API itself.
How it works
Send an image plus a text prompt, get an edited image back, the same shape as GeminiImageEdit elsewhere in this pack but hitting a different backend entirely. Currently the schema only offers one model, jimeng_i2i_v30.
The inputs and outputs that matter
image- the source image to edit.prompt- your edit instruction.model-jimeng_i2i_v30, currently the only option.scale(default 0.5, range 0–1) - the README doesn't spell out its exact mechanics beyond stating the default, but as with most image-to-image APIs, expect it to trade off between hewing close to your source image and following the prompt more aggressively - push it up if edits feel too timid, down if the result strays too far from your source.time_out(default 300 seconds, up to 3600) - how long the node waits for a response before giving up. Raise it if requests are timing out rather than failing outright.seed- for (attempted) reproducibility across calls.
Output is a single image.
How to install it
Recommended: ComfyUI Manager, search "ComfyUI Layer Style Advance". Manual:
cd ComfyUI/custom_nodes
git clone https://github.com/chflame163/ComfyUI_LayerStyle_Advance.git
Then install_requirements.bat / install_requirements_aki.bat, or pip install -r requirements.txt plus repair_dependency.bat. Restart. No model to download locally - generation happens on Volcano Engine's servers.
Getting credentials, the part that's genuinely more involved than this pack's other API nodes: create an account on Volcano Engine and apply for an API AccessKeyID and SecretAccessKey. Volcengine accounts typically require phone verification, and depending on which services you activate, sometimes real-name or business verification too - budget for some signup friction if you're setting this up from outside China. Once you have both credentials, rename api_key.ini.example (in the plugin's root directory) to api_key.ini, and fill in your AccessKeyID and SecretAccessKey in the corresponding volcengine_ fields.
Common issues & troubleshooting
Requests fail outright, or authentication errors. Confirm both the AccessKeyID and SecretAccessKey are filled in - Volcengine's API requires both, not just one, and a partially-filled credential pair fails the same way a missing one does.
Requests time out. Raise time_out past the 300-second default first - cross-region API calls to a mainland-China-hosted service can be slower than the equivalent request to Google or DeepSeek, depending on where your ComfyUI instance is physically located.
Can't get past account signup, or verification steps you weren't expecting. This is a normal part of using a Chinese cloud provider's API from outside its home market, not something this node's install steps can smooth over - it's specifically why this node is more setup-heavy than the pack's Gemini or DeepSeek equivalents.
Not sure this is worth the setup versus GeminiImageEdit. If you don't already have a reason to specifically want Jimeng's output style, GeminiImageEdit elsewhere in this pack is a meaningfully lower-friction path to the same category of image-editing task.
Inputs (6)
| Name | Type | Default | Description |
|---|---|---|---|
| image | IMAGE | — | |
| model | COMBO | 1 options: jimeng_i2i_v30 | |
| time_out | INT | 3001–3600 | — |
| scale | FLOAT | 0.50–1 | — |
| seed | INT | 00–1000000000000000000 | — |
| prompt | STRING | — |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| image | IMAGE | — |