ByteDance Seedream V5.0 Pro (Official Edition)
Seedream V5.0 Pro, official route, with image reference support
- image
- api_config
- image
- request_id
- urls
The most capable Seedream node in the pack: ByteDance's V5.0 Pro text-to-image through the official API route, with an optional reference-image input, format control and a watermark toggle. This is the top of the Seedream line, and the node that uses it fully.
Everything about this one is the premium lane. Pro is the tier you pay for when the output has to be excellent; the official route is the direct vendor path; and the reference-image input means you're not limited to text - feed it a character or a style anchor and it'll generate from that. If you've been using the Channel Pro node and hitting its text-only wall, this is the upgrade. The extra inputs are exactly the difference: image reference, size at 1K or 2K, output_format, and watermark.
How it works
Output node in the BizyAirPlus cloud layer. With BizyAirPlus ON in the action bar, the node uploads your optional reference image and forwards the prompt, size, format and watermark setting to Seedream 5.0 Pro via the vendor's direct API, then returns the finished image to your workflow. Closed model - no weights, no GPU. The image input is the standout: it turns a pure text-to-image node into a reference-guided generator, which is where Pro models shine.
The inputs that matter
- prompt - free-form text; the primary driver.
- image - optional reference image for image-guided generation. Wire it from a Load Image node or another generator's output.
- size - COMBO:
1Kor2Kresolution. - output_format -
jpegorpng; PNG if you'll keep editing. - watermark - off by default; flip on only if you actually want ByteDance's mark on the output.
Then the usual helpers: api_config (per-request base URL/API key override) and skip_error (placeholder on failure instead of a workflow abort).
The outputs
image - IMAGE tensor for saving, upscaling or downstream use. urls - direct result link. request_id - the support trace ID.
Installing it
Ships in BizyAirPlus - ComfyUI Manager search "BizyAirPlus", or:
cd ComfyUI/custom_nodes
git clone https://github.com/siliconflow/BizyAirPlus.git
cd BizyAirPlus
python -m pip install -r requirements.txt
Restart, toggle BizyAirPlus ON, key from bizyair.ai (stored at ~/.BizyAirPlus/apikey.ini, or BIZYAIR_API_KEY). Real dependencies are the bizytrd and bizyair-cloudberry pip packages - light install, nothing to download.
Gotchas
The standard cloud-node failure list applies, and the ComfyUI console always has the real error. (And yes, BizyAir is legitimate SiliconFlow software - the malware talk traces to an old Manager auto-install mixup, fixed.) Budget-wise, this node is the most expensive image generation in the pack, so the workflow that pays is: draft on lite, reference on Pro only when a concept survives. Keep the size at 1K until the prompt is right, and remember the image reference plus a short "make it X" prompt is often stronger than a long text-only prompt.
Inputs (7)
| Name | Type | Default | Description |
|---|---|---|---|
| prompt | STRING | Prompt | |
| imageopt | IMAGE | Reference image for image-to-image generation | |
| sizeopt | COMBO | 1K | Image size: 1K or 2K resolution |
| output_formatopt | COMBO | jpeg | Output image format: jpeg or png |
| watermarkopt | BOOLEAN | false | Add watermark to the generated image |
| api_configopt | BIZYAIR_OPENAPI_CONFIG | 单次请求覆盖base_url和api_key | |
| skip_erroropt | BOOLEAN | false | 开启后遇到错误不中断工作流,输出对应类型的错误占位符 |
Outputs (3)
| Name | Type | Description |
|---|---|---|
| image | IMAGE | — |
| request_id | STRING | — |
| urls | STRING | — |