About

What this is

Video Gen Now turns a written prompt, a still image, or a set of reference images into a finished video clip. You pick a model, describe the shot, and get an MP4 back — usually inside a minute. There is no timeline, no node graph, and nothing to install.

It exists because the good video models are all API-only. Getting a clip out of them normally means an API key, a billing account with each provider separately, and a script to poll for the result. This site is the part in between: one page, one balance, three models.

Which models run here

Every render goes to a third-party model over its provider API. We do not train or host models ourselves.

• Seedance 2.0 — ByteDance's flagship. 480p through 4K, four to fifteen seconds, and a synchronized audio track generated from the same prompt at no extra cost.
• Seedance 2.5 — the long-form variant. Up to thirty seconds in a single take and up to thirty reference images, in exchange for a 720p ceiling and a higher price per second.
• MiniMax H3 — 2K on every mode at a flat per-second price, and unusually stable subject motion across a long take. No audio track.

All three support text to video, image to video, and reference to video. The model comparison goes through the trade-offs with the actual numbers.

How paying works

There is no subscription. You buy credits, and each render costs credits.

Credits track what a render actually costs us on the provider's hardware, which is why the price moves with resolution, duration, and aspect ratio rather than sitting at a flat rate. A wider frame is more pixels, and more pixels cost more. The exact cost is shown on the button before you submit — nothing is charged based on an estimate.

If a render fails, the credits go straight back to your balance. You are not charged for our queue's bad day.

What we do not do

• We do not claim rights over your renders. What you generate is yours, subject to the terms and to the model provider's own policies.
• We do not use your prompts to train anything. They go to the model provider to produce your clip, and they stay in your library so you can find them again. See the privacy policy for exactly what is stored and where.
• We do not pad the samples. Every clip on the model pages was rendered here, on the model the card names, with the prompt printed underneath it. Nothing on those pages is stock footage or another tool's output.

Status

Video Gen Now is new and run by a small team. That means the roadmap moves quickly and it also means you will occasionally find rough edges. If you hit one, open a ticket from support in your account and it will reach a person.