Seedream 5.0 Flash · Text to Image

Here's the one line that matters on this node: resolution offers 1k, 1.5k and 2k, and all three cost the same. So there is no reason to ever pick 1k. Default is already 2k, and you should leave it there unless you're in a hurry.

4.35 credits a picture. Flash is the fast tier of ByteDance's Seedream 5.0 line - Lite sits above it on price, Pro above that - and for a lot of work the difference won't show.

What it is

PopNow's node for bytedance/seedream-5.0-flash/text-to-image, generated from the model's own input schema, so the widgets you see are exactly what the model takes: a prompt, an aspect ratio, a resolution tier, an output format. That's it. Run it and the picture comes back decoded as a ComfyUI IMAGE batch.

There's no checkpoint loader, no sampler, no seed. Not because the pack is lazy - because a hosted model doesn't expose those and the schema is the truth. Expect roughly a few seconds to a handful, depending on the queue, and note that nothing touches your GPU.

The four inputs

• prompt - subject, composition, style, lighting, and any exact text you want rendered. Seedream's generation of legible text is one of its selling points, so quote the copy literally rather than describing it.
• aspect_ratio - 1:1 default, plus 4:3, 3:4, 16:9, 9:16, 3:2, 2:3, 4:5, 5:4, 21:9, 9:21, 2:1, 1:2, 3:1, 1:3. Wide enough that you can hit almost any delivery format without cropping later.
• resolution - 1k, 1.5k, 2k. Same price across the board. Use 2k.
• output_format - jpeg (default) or png. Pick png if the image is heading into another pipeline or you care about compression artefacts before an upscale.

Outputs

images (a batch) and task id. Wire images into Save Image, or into your local finishing chain - a cheap hosted draft followed by local detail work is the workflow this node is best at.

Install

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

or Manager → search PopNow → Install. Key from popnow.ai/api/console/keys, entered in Settings > PopNow or as POPNOW_API_KEY in ComfyUI's environment. Restart once more so the node list rebuilds from your account's catalog. Nothing to pip install: the pack declares zero dependencies and uses only ComfyUI's own libraries.

If you're coming from the old single-file popnow_api.py, delete it - its PopNow Generate node is the same type with the same inputs, so your saved workflows keep working.

Where it goes wrong

• PopNow 40201 - credits exhausted. Cheap per image, but a 50-image batch is still a real bill.
• PopNow 400 naming a field - an aspect ratio or format that isn't on the list, or a prompt field sent empty.
• PopNow 40004 - the combination isn't served. Rare on this node; typically a resolution/ratio edge case.
• PopNow 40310 - verify your email address from the sign-up mail.
• content_rejected / generation_failed - the model declined or couldn't finish; failed runs are refunded.
• Missing a model that exists on popnow.ai - restart ComfyUI. The catalog is fetched at startup.

Set Settings > PopNow > Max credits per run if you're going to loop a prompt over a CSV of styles. The quote-before-run check is the thing that makes batch experimentation safe rather than a gamble.

Honest placement

Flash is not the best image model you can touch. Pro is sharper, and locally you can run things you control far more precisely. What Flash is, is the shortest distance between a sentence and a 2K image with no setup at all - the draft stage. Use it to find out whether an idea works, then spend your time and your GPU on the ones that do.