Seedream 5.0 Lite · Image Editing

Lite is the tier that hands you a fixed list of eight width*height presets and writes a 2K image, for $0.0345. One to ten reference images in, one edited frame out, no filter on the result.

Be honest about the tier, though: Lite is the most expensive per-image option in this pack's Seedream 5.0 line and the least flexible. Flash edit is $0.029 with aspect ratios and a transparency path; Pro edit is a sliver more with an enable_thinking pass. Lite's edge is exact pixels - the same argument as Seedream 4.0's edit node, with one fewer preset slot and a newer model behind it. Reach for it when your output has to land on a specific canvas, or when you want the uncensored path on a frame whose dimensions you already know.

What it does

Standard API-node shape: each connected image is encoded, PNG where it fits under the 10 MiB upload ceiling and JPEG if not, then uploaded to SpicyAPI. The node takes a quote, submits the job, and polls. The answer arrives as a ComfyUI IMAGE batch - from your graph's point of view, a normal image, ready for Save Image or the next local stage. No weights, no VRAM, no sampler.

References are growing sockets on image_urls: wire one in and another appears. Up to ten, each one adding to the quote, and the prompt should name what each contributes in the order you connected them.

Inputs

• prompt - what to produce from the sources: subject, composition, style, lighting. Editing instructions work better when they say what must stay untouched as well as what changes.
• image_urls - one to ten source images. This is a multi-reference node; the obvious use is a person plus a wardrobe plus a setting, assembled into one frame.
• size - the eight presets, 2048*2048 by default. Set this first if you're building a series.
• output_format - jpeg or png.

Outputs: images and task id, the latter a STRING that names the request for SpicyAPI's console logs or for support when something fails.

No seed on this node, so you can't reproduce a roll exactly. Keep what you like.

Install

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

Then restart ComfyUI, open Settings > SpicyAPI, paste a key from spicyapi.ai/console/keys - or set SPICY_API_KEY if it's a headless box - and restart again so your account's model list replaces the bundled one. ComfyUI Manager: search SpicyAPI. The pack ships no Python dependencies and pulls no models; the dependencies = [] in its metadata is not a typo.

Cost, and how not to get surprised

$0.0345 per image. The badge on the node shows the current rate and it updates as you change options; the real number for a given run comes from a quote taken before submission. Protect yourself with Settings > SpicyAPI > Max cost per run: any run quoted above that figure is refused before it's charged. Two things people trip on: multiple reference images raise the quote, and control after generate starts on fixed on purpose - seeds being stable means queueing twice doesn't bill you twice, and you have to switch it to randomize deliberately for a new draw.

Troubleshooting

No SpicyAPI key is set - the key isn't where the pack looks (SPICY_API_KEY, or user/spicyapi/config.json written by the settings panel). 40310 - verify your email. 40004 - no deployment serves that option combination; change what the message names, nothing charged. An error ending in request_id ... is the one to paste into a support ticket. Missing models mean a restart. And remember that hitting Cancel stops you waiting, not the server-side job, which completes and is billed if it succeeds.