Seedance 2.0 · Text to Video

Type a prompt, get a cinematic clip with sound. That's Seedance 2.0 Text to Video, and it's the plainest way to find out what ByteDance's video model can do without installing a single weight file. It's also the node that keeps the argument about API nodes alive, because text-to-video is the most expensive way to use a video model: everything has to be generated from nothing, so nothing is shared.

There are no Seedance weights. If you want this model inside ComfyUI, someone has to call it, and the rate on this node starts at 10.5 credits per second (~$0.105/s). A 5-second 720p clip is about 52 credits. A minute of 4K is a different category of purchase.

What it actually is

PopNow is the vendor that operates api.popnow.ai, so this pack is the API's own ComfyUI front end rather than a third-party wrapper - the models, the schema and the billing come from one place. The pack generates the node from Seedance 2.0's input schema, quotes each run with jobs/quote before it starts, holds at most the quoted amount, and submits with an idempotency key so a network retry can't double-charge you.

That plumbing is worth knowing because it's the difference between a metered API and a slot machine. The general trade for this whole category is cost per call, data leaving the machine, and a content filter you can't patch out - the model refuses what ByteDance refuses, and there's no abliteration for video.

The inputs a beginner sets

• prompt - what to generate: subject, action, camera, mood. Seedance responds well to one clear camera instruction ("slow dolly in, shallow depth of field") rather than a pile of adjectives.
• aspect_ratio - 21:9, 16:9, 4:3, 1:1, 3:4, 9:16, or adaptive. adaptive lets the model pick, which is usually the wrong call once you know your delivery format.
• resolution - 480p, 720p, 1080p or 4k, default 720p. Super-resolution tiers aren't offered here; what you pick is what you get.
• duration_seconds - the billed quantity. The dropdown runs 4 to 15 seconds, plus -1 for smart duration, where the model decides how long the shot wants to be. -1 is genuinely useful for a first pass: it stops you paying for five seconds of a shot that wanted two.
• generate_audio - on by default. Native, synchronised sound is Seedance's headline feature and the thing local video models still struggle with.
• enable_web_search - off by default, and off is right unless you're rendering something that needs current information. Runs with it on are billed at the web-search rate.
• bitrate_mode - standard or high. High for a final render you'll actually publish.
• seed - starts on fixed, so re-queueing the same graph reuses the seed. Randomize the control-after-generate widget for new takes; each one bills.

Outputs

video goes to Save Video; last frame gives you an IMAGE of the final frame - handy for chaining into another node as a first frame - but only when return_last_frame is switched on, otherwise anything downstream is blocked with a nasty-looking (and harmless) execution message. task id is the string you hand to PopNow's console or support.

Install and key

Install from ComfyUI Manager by searching PopNow, or:

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

Restart, then add your key in Settings > PopNow (or export POPNOW_API_KEY before launching ComfyUI). Restart once more so your account's catalog fills the node list. No separate pip install is needed: the pack ships with an empty dependency list and uses ComfyUI's own aiohttp/torch/PyAV.

Where people get burned

• The badge is a rate, not a total. 10.5 credits per second of output. Multiply before you commit to a 15-second 4K pass.
• PopNow 40201 is "you're out of credits" - a top-up, or drop the resolution.
• PopNow 40004 means the combination isn't served (4K with a weird aspect ratio, say). Change the setting the message names.
• Cancelling doesn't stop the job. It stops your wait. The run continues on PopNow, is charged if it succeeds, and the file lands in your API console logs.
• Set Settings > PopNow > Max credits per run. Hard ceiling checked against the quote before anything is held. Do this on day one, not after the surprise.

The community verdict on Seedance inside ComfyUI has been consistent since 2.0 landed: impressive output, real price - the recurring complaint was how few clips a monthly budget buys, and the recurring reply is that this is simply what the render costs. Treat it as a tool you rent for shots you can't get locally, not a default for every idea.