Nodes/ComfyUI/ByteDance Image
ComfyUI Node Runs on cloud

ByteDance Image

ByteDance's Seedream 3.0 in ComfyUI — the retired version

By Comfy-Org·Created 4 years ago·Updated about 22 hours ago· 129,495
ByteDance Image
    • IMAGE
    model
    prompt
    size_preset
    width1024
    height1024
    seed0
    guidance_scale2.50
    watermarkfalse

    ByteDance's Seedream models are the closed flagship of the company that also quietly releases half the ecosystem's load-bearing open infrastructure - and Seedream is firmly in the closed, API-only column. ByteDance Image is the ComfyUI wrapper for the text-to-image tier, and the first thing to know is the same caveat as several nodes in this batch: it's deprecated. Marked is_deprecated in core, superseded by the newer ByteDance Seedream node (the V2-generation one that's got the most search traffic of the whole ByteDance family). If you're starting fresh, use that one; this page is for recognizing and maintaining old workflows.

    It's a partner node, of course: your prompt posts to Comfy's proxy at api.comfy.org, ByteDance's (BytePlus) servers run the model, and credits come off your Comfy account. Notably, the request hits a synchronous image-generation endpoint (/proxy/byteplus/api/v3/images/generations) rather than a submit-then-poll flow, so the node returns when the image is ready. No VRAM, no model files.

    The input set is a plain, no-frills text-to-image layout. model is a single hardcoded option - seedream-3-0-t2i-250415 - which is exactly the sort of thing the newer DynamicCombo-based Seedream node fixed. prompt, then size_preset: a dropdown of recommended sizes, where picking Custom unlocks the width and height fields (512–2048, step 64, and the tooltip is explicit that width/height only take effect when size_preset is Custom). Optional seed, guidance_scale (default 2.5, "higher = follows prompt more closely"), and watermark - a boolean that appends an "AI generated" watermark, buried under advanced. Given that ByteDance's regulatory posture is one reason Seedream is API-only, that watermark toggle existing at all is mildly surprising; leave it off unless you want the marker.

    Output is a single IMAGE tensor, already decoded, ready for Save Image or downstream nodes. Nothing unusual there.

    The grounded judgment. Seedream 3.0 is a genuinely strong Chinese text-to-image model with excellent prompt adherence and text rendering, and if you're on an older workflow that uses this node, it still works - deprecated in ComfyUI's parlance means "kept for compatibility," not "removed." But the fixed model string, the preset-or-custom size dance, and the missing DynamicCombo all argue for migrating to the current Seedream node when you touch the graph. For new work, this is the version you'll only meet inside an old workflow JSON - and now you know what it is. If you're here because you googled "ByteDance Image," you probably actually want the newer Seedream node.

    Categorypartner/image/ByteDance

    Inputs (8)

    NameTypeDefaultDescription
    modelCOMBO1 options: seedream-3-0-t2i-250415
    promptSTRINGThe text prompt used to generate the image
    size_presetCOMBOPick a recommended size. Select Custom to use the width and height below
    widthINT1024512–2048Custom width for image. Value is working only if `size_preset` is set to `Custom`
    heightINT1024512–2048Custom height for image. Value is working only if `size_preset` is set to `Custom`
    seedoptINT00–2147483647Seed to use for generation
    guidance_scaleoptFLOAT2.501–10Higher value makes the image follow the prompt more closely
    watermarkoptBOOLEANfalseWhether to add an "AI generated" watermark to the image

    Outputs (1)

    NameTypeDescription
    IMAGEIMAGE