Produce long-form avatar videos with Kling Avatar

Kling Avatar is Kuaishou's long-form talking-head video model. It animates a reference image into a speaking character driven by an audio track, producing clips of up to 60 seconds with synchronized lip movement and expressive performance.

You supply a portrait and a voice recording. The model reads the audio for timing, prosody, and emotion, then generates video where the reference character speaks in sync. Popcraft handles routing, polling, and delivery of the finished MP4.

Popcraft's free tier includes credits you can spend on Kling Avatar. Longer clips and Pro-mode renders cost more credits, and paid plans or top-ups extend your monthly allowance.

Popcraft paid plans grant commercial rights to the avatar videos you render. You still need permission to use the likeness of any real person in your reference image — do not generate talking-head videos of people without their consent.

Between 2 and 60 seconds per render on Popcraft. For anything longer, chain multiple clips on the Remotion timeline or stitch scenes together in the AI Agent pipeline.

Both generate talking-head video from one portrait and audio. Kling Avatar is tuned for longer takes up to 60 seconds and offers Standard and Pro quality tiers. OmniHuman 1.5 tops out at 30 seconds at 1080p and adds mask support for multi-face scenes.

Open the Character page, pick Kling Avatar, upload a portrait, attach your audio, choose aspect ratio and quality mode, and submit. Progress streams live and the rendered video saves to your gallery and asset library.