The Journal
ProductAug 09, 2026 · 3 min read

Video is in beta: up to fifteen seconds, starting from your own stills

Motion is now in the studio on paid plans: clips of up to fifteen seconds at up to 1080p, priced per second from the same credits as your stills, and able to start from a frame you already made. Here is exactly what it does, and what it does not.

Hana Ishikawa

Hana Ishikawa

Product Lead

Video is in beta: up to fifteen seconds, starting from your own stills

Video has been the most requested thing in our inbox, and the thing we were most reluctant to ship badly. A video tool that cannot hold your character's face is a demo, not a product.

So the first release leans on what already works: the stills you have made. Start from a frame you already like, describe the motion, choose a camera move, and the clip begins from that frame.

What ships today

Motion lives at Studio › Video and is open on every paid plan — Keeper, Vault and Keystone. The free plan does not include video.

There are three models, all from third-party providers and called through OpenRouter: Seedance 2.0 Fast and Seedance 2.0 from ByteDance, and Hailuo 2.3 from MiniMax. Clips come out at 16:9, 9:16 or 1:1, with no audio.

It is in beta, which for us means three things: it works, it is not fast, and the details may change before it is final.

Motion should inherit the look you already built. Nobody wants to grade a character twice.

— release note

Starting from a still

The most reliable way to keep a character in motion is to start from a frame where she already looks right. Use one of your stills as the first frame and the clip opens on it — her face, the light and the grade are in the picture before anything moves.

Start frame — Any still from your Vault, used as the opening frame. Optional, but it is what holds a face.

Motion — What happens in the shot, described in plain language.

Camera — Push in, pull out, orbit, crane, handheld or static — or leave it to the prompt.

Length — 4, 5, 6, 8, 10, 12 or 15 seconds on Seedance; 6 or 10 seconds on Hailuo.

Resolution — 480p, 720p or 1080p on Seedance; 720p or 1080p on Hailuo.

Fig. 1 — three frames from one clip that started from a still. The face holds; the motion is allowed to blur.

Fig. 1 — three frames from one clip that started from a still. The face holds; the motion is allowed to blur.

The limits

Fifteen seconds and 1080p are the ceiling today, set by what the models we call accept. Without a start frame, a clip is generated from the prompt alone, and how well it holds a specific face is up to the model.

There is no audio, no frame interpolation and no way to extend a clip. A third-party model also does not promise the same clip twice from the same inputs, so when a take is good, keep it.

Cost and credits

Video is priced per second and drawn from the same credit balance as your stills — there is no separate video allowance. The price is shown before you render, and a clip that fails, or is refused by the model's safety filter, is not charged.

2 credits / s — Seedance 2.0 Fast

3 credits / s — Seedance 2.0

4 credits / s — Hailuo 2.3

So a five-second clip on Seedance 2.0 Fast is 10 credits, five seconds on Seedance 2.0 is 15, and the longest clip on the menu — fifteen seconds of Seedance 2.0 — is 45.

What it does not do

No audio, no clip extension, nothing above 1080p and nothing longer than fifteen seconds. We would rather say that plainly than put dates on a roadmap.

Every clip you render is kept in your Vault at the resolution it was rendered at.


Motion is live on every paid plan today. If you already have a still you trust, that is the one to start from — the whole point is that you do not have to build anything new.

FILED UNDER

  • Video
  • Character keys
Hana Ishikawa

Hana Ishikawa

Product lead at ockeeper. Writes about the Vault, Character Keys and how the studio is meant to fit into the way you already work.