Nodes/SpicyAPI/Seedream 5.0 Lite · Text to Image
ComfyUI Node

Seedream 5.0 Lite · Text to Image

A 2K frame from a short prompt, at eight fixed sizes

By spicyapi-ai·Created 21 days ago·Updated 2 days ago· 2
Seedream 5.0 Lite · Text to Image
    • images
    • task id
    ◄prompt►
    ◄size2048*2048►
    ◄output_formatjpeg►
    ◄seed0►

    Promise 3 to 3,000 characters, pick one of eight preset sizes, get one 2K image for $0.0345. No aspect-ratio dropdown, no resolution tier, no seed - dimensions you choose from a list, and a result that's always 2K-class.

    That's an odd set of tradeoffs in a pack where Seedream 5.0 Flash does 1K-to-2K across fifteen ratios for $0.029. So use Lite for what it's actually good at: a specific canvas. Twenty different jobs want 2048*2048; a horizontal banner wants a 3136*1344 strip and nothing else will do. If your deliverable has fixed dimensions, picking from eight known-good pixel counts removes an entire class of "why is this 1 pixel off" annoyance. If it doesn't, take the cheaper Flash node and be done.

    What's going on

    Predictably: your prompt and the size go to SpicyAPI, a hosted ByteDance model renders it, and the file comes back as an ordinary IMAGE in your graph. There's nothing local about this at all - no checkpoint, no VRAM, no install time - which is the whole selling point of the pack. It's a rendering service with a node-shaped receipt, and the receipt matters: the rate is on the node, and the charge is quoted before the job is sent.

    The three inputs

    • prompt - 3 to 3,000 characters, plain description of subject, composition, style and lighting. It's the only creative control you have, so write it properly.
    • size - one of eight width*height presets, 2048*2048 by default, out to a widescreen 3136*1344 strip.
    • output_format - jpeg by default, png when you want the bigger lossless file.

    And the outputs: images, the IMAGE you wire to Save Image, and task id, a STRING carrying the SpicyAPI request id. Ignore the second unless you need to trace a specific run with support. There's no seed widget, so two identical runs are not guaranteed to match - save the good one.

    Install (it's the same two lines for every node in the pack)

    cd ComfyUI/custom_nodes
    git clone https://github.com/spicyapi-ai/comfyui-spicyapi
    

    Restart ComfyUI. Open Settings > SpicyAPI, paste a key from spicyapi.ai/console/keys, and restart once more so your account's catalogue replaces the bundled snapshot. On a remote or headless machine, set SPICY_API_KEY in the environment instead - that keeps the credential out of the workflow file and out of any PNG metadata, so you can share graphs and images freely. ComfyUI Manager users can search SpicyAPI and install it that way. Nothing else installs: no pip packages, no model downloads.

    Cost and the safety rails

    $0.0345 per image, all sizes. The node's price badge updates as you change options, so you can see the rate for what you've built before running. Settings > SpicyAPI > Max cost per run is the hard stop: a run quoted above it is refused before submission and nothing is charged. The pack also starts every seed on fixed with control after generate set to fixed - deliberately, because every run is billed, so a double queue shouldn't silently buy you a second image. Flip it to randomize when you actually want a different draw.

    When it breaks

    No SpicyAPI key is set covers all the key problems, including a key pasted with a stray space. 40310 means the account's email isn't verified. 40004 means no hosted deployment serves the option combination you asked for; the error names the field, and again, nothing is charged. Errors containing request_id ... should be sent to support with the id intact. A model that's on the website but not in your node menu just needs a ComfyUI restart - the list is pulled from your account at startup. And cancelling a run in ComfyUI doesn't cancel it upstream; it finishes and is billed if it succeeds, with the output waiting in your console for 14 days.

    CategorySpicyAPI/Image/ByteDance

    Inputs (4)

    NameTypeDefaultDescription
    promptSTRINGDescribe the image to generate: subject, composition, style, and lighting.
    sizeCOMBO2048*2048Output image size in pixels, written as width*height.
    output_formatCOMBOjpegFile format of the returned image.
    seedINT00–2147483647This model takes no seed, so the number is never sent. Changing it makes ComfyUI run the node again: set 'control after generate' to randomize for a new result on every run; every run is billed.

    Outputs (2)

    NameTypeDescription
    imagesIMAGE—
    task idSTRINGThe SpicyAPI task id, for the console logs or support.