Pick the plan that matches your output

It is a plugin that installs into your editing software and adds two AI engines to the timeline: one that replaces a performer's face across a shot, and one that generates natural Chinese-language speech. Both are driven from the same panel, and anything you export carries a commercial licence.

Video rendering is unlimited on both plans, because it runs on your own graphics card rather than our servers. Generated audio is metered: Basic includes 60 minutes a month, and Pro is unlimited under a fair-use policy.

Everything you produce while your subscription is active stays licensed for commercial use forever, including client deliveries, paid advertising and monetised platform uploads. If you cancel, projects you already finished remain covered — you simply stop being able to render new ones.

Yes. Upgrade whenever you like and you are charged the pro-rated difference straight away. Downgrades take effect at the start of the next billing period. Cancelling stops the next renewal; there is no refund for the current period, but you keep access until it ends.

Identities in our library ship with a signed release for commercial use. For anything you upload, you are responsible for having the subject's consent, and uploads are screened against a blocklist of public figures. Using the plugin to impersonate a real person without permission, or to produce sexual or deceptive content, breaks the acceptable-use terms and gets the licence revoked.

The speech models were trained on Mandarin as the primary language rather than adapted from English, so tone pairs, erhua, measure words and number reading come out the way a native speaker would say them. Cantonese and several regional accents are included on Pro. We would rather you judge it yourself — the sample project in the download has raw output in it.

A GPU with at least 8 GB of VRAM for 1080p, and 16 GB for comfortable 4K work. Apple Silicon Macs run on the unified memory pool — M2 Pro or better is the sensible floor. Preview scrubbing is real-time at half resolution on most current cards.

It depends on the workflow. Tracking, preview, 1080p replacement and lip sync run entirely on your own machine — nothing is transmitted. 4K synthesis, voice cloning and the dialect packs run on our inference cluster, so those jobs upload your source media and download the result. The plugin labels which steps are cloud-backed, shows an estimated transfer size before each job, and lets you disable cloud processing outright if your footage cannot leave the building. Everything on the Basic plan is local-only.

Roughly the size of your source media in each direction. Ten minutes of 4K ProRes is 30–60 GB up, with the render returning at a similar size; the same shot delivered as H.264 is a small fraction of that. Proxy mode cuts it dramatically — it uploads a low-resolution version and applies the result to your full-quality footage locally. If you are on a metered line, behind a VPN, or somewhere with a data cap, schedule cloud jobs the way you would a large backup: overnight, on a connection you trust.

If you are one person cutting short-form video and you only need the stock voices, Basic covers it. If you deliver to clients, need 4K, want to clone a specific presenter's voice, or run the same edit across several platforms, Pro pays for itself quickly. You can start on Basic and upgrade mid-term without losing anything.