EvoLink Seedance 2.0 Fast Text-to-Video (Official)

Video generation is the fastest way to empty an API balance, and Seedance is the model that made the community do the math out loud. So the genuinely useful thing about EvoLink Seedance 2.0 Fast Text-to-Video (Official) is right there in the name: it's the Fast tier, ByteDance's cheaper-and-quicker lane for the same Seedance 2.0 model. The trade is a hard cap at 720p - no 1080p, no 4K - which is exactly the right place to cut when you're still figuring out whether the clip should exist at all.

What it is

Same shape as the full Seedance 2.0 T2V node, trimmed to the cheap lane:

• prompt (required) - the shot description, Chinese welcome, ~500 chars suggested.
• duration - 4–15 seconds, default 5, billed by duration.
• quality - 480p or 720p only. Fast tops out at 720p, and 480p is the "just testing the prompt" price point.
• aspect_ratio - adaptive default, or 16:9 / 9:16 / 1:1 / 4:3 / 3:4 / 21:9.
• generate_audio - on by default, no extra charge.

Outputs are the pack standard: video (connect Save Video or you'll see nothing), task_id, status_info, result_urls (24-hour expiry), response_json. The 30-minute default timeout applies - Fast is quicker than full Seedance, but "quicker" is relative when the render is on someone else's GPU.

Install and key

The shared EvoLink routine. ComfyUI Manager → search EvoLink → install → full restart (close the console, not just the browser tab). Manual:

cd ComfyUI/custom_nodes
git clone https://github.com/deeplearning-goethe/comfyui-evolink

Only requests as a dependency; no models, no GPU. Key from evolink.ai/dashboard/keys (sk-), pasted once into api_key - saved locally, blank after, never in a shared workflow.

The workflow pattern that makes it worth it

The way to use this node is as a gate. Write your prompt, run it here at 480p, judge whether the motion and framing are right, iterate, and only promote the winner to the full Seedance 2.0 node for the 1080p/4K final. You still pay per call - no tier dodges that - but you pay the cheap rate for all the calls that are really just you thinking. That discipline is the difference between "Seedance costs a fortune" and "Seedance costs a bit." Failed and moderation-blocked tasks aren't billed, and unchanged params cache without recharging, which keeps the iterating loop from being a horror show. Just remember the standing caveat: your prompt leaves the machine, and the clip you get back is a metered render, not a download you own.