Seedream 5.0 Flash · Text to Image

A ticket booth for images, and a cheap one. Write up to 5,000 characters, pick one of fifteen aspect ratios from 1:3 to 3:1, pick 1K, 1.5K or 2K, and pay $0.029. That flat rate across resolution tiers is the interesting part: the node never asks you to think about whether the bigger render is worth it, because it isn't more expensive.

Flash is the light tier of ByteDance's Seedream 5.0 line - the one you use when you want a clean, prompt-following frame without the Pro tier's deliberation. There's no enable_thinking switch here, no negative prompt, no seed. It takes the prompt and the shape and returns a picture. That's a feature if your prompt is already good and a limitation if you were hoping to argue with it.

Mechanism, briefly

The node posts the prompt and format options to SpicyAPI, waits for the hosted model to finish, and downloads the result as a normal ComfyUI IMAGE. There's no local model, no VRAM, no checkpoint: this runs on a machine you'll never see, which is exactly why it works on a MacBook or a CPU-only box. The trade is a per-image charge instead of electricity, and the obvious one - your prompt leaves your machine, and the resulting image is whatever the hosted deployment decides to give you.

Inputs

• prompt - subject, composition, style, lighting, and any literal text you want rendered, in quotes. Long prompts are genuinely supported here, so put the detail in.
• aspect_ratio - 1:1 by default, and the full set spans portrait 1:3 through panorama 3:1. Pick the ratio your final destination wants; cropping a 3:1 down to square throws away the reason you paid for a big render.
• resolution - 1k, 1.5k or 2k, default 2k, all the same price. Take the 2k.
• output_format - jpeg or png. PNG if the image is going somewhere that cares about compression, JPEG if it's going straight out the door.

Output: images, an IMAGE for Save Image or anything downstream, plus a task id STRING naming the SpicyAPI request.

Install

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

Restart ComfyUI, then Settings > SpicyAPI and paste a key from spicyapi.ai/console/keys. On a server, export SPICY_API_KEY in the environment instead - the key never lives in the workflow file, which is why you can share a graph without leaking it. Restart once more after adding the key so the node list is rebuilt from your account. ComfyUI Manager: search SpicyAPI. Zero dependencies, zero downloads.

Pricing and the queue button

$0.029 per image, shown on the node and quoted before the job is submitted. The pack's safety rails are worth knowing explicitly: seeds and control after generate start on fixed, so queueing the same graph again reuses the cached result instead of buying a duplicate; and the Max cost per run setting in SpicyAPI's settings stops anything above your ceiling before submission. A cancelled run is not a refunded run - the work continues on the server.

Known failure modes

No SpicyAPI key is set if the key isn't configured or is mistyped. 40310 if the account's email hasn't been verified - check the sign-up mail. 40004 when no deployment serves the combination you requested; the message names the offending option and nothing is charged. Errors ending in request_id ... are the ones support wants, so copy the id. New models missing from the node menu is a restart problem, not a bug. And if you're routing traffic through a proxy, HTTPS_PROXY needs to be right or requests fail with a proxy-or-firewall message.