Kling AI drops Avatar 2.0 with 5-minute clips and cleaner hands

Making an AI avatar talk for 30 seconds is relatively easy nowadays. Keeping that same digital character alive, expressive, and human-like for five full minutes without their hands morphing into claws is an entirely different engineering challenge.

Kling AI just rolled out Avatar 2.0, a major platform upgrade designed to tackle the two biggest headaches in synthetic video: rigid body language and short duration limits.

Natural hands and full 5-minute takes

The marquee headline is video length: Kling AI Avatar 2.0 now supports single-take generation of up to 5-minute videos.

That 5-minute threshold is a big deal because it covers full long-form formats right out of the box, including podcast explainers, in-depth product pitches, full musical performances, and storytelling clips without stitching multiple cuts together.

Beyond sheer duration, the visual performance got a significant makeover. Version 2.0 introduces coordinated body movements, emotional facial expressions, dynamic camera angles, and notably stabilized hand gestures.

The team specifically overhauled hand rendering to fix the unnatural flickers and anatomical glitches that plagued version 1.0, making presentation gestures look far more deliberate.

Benchmarking against HeyGen and Omnihuman

To back up the quality claims, Kling AI released internal benchmark figures evaluated on November 25, 2025.

Using a pairwise preference dataset across multiple ethnicities, languages, art styles, and speech/singing scenarios, Kling reported an overall GSB score of 194% against Dreamina's Omnihuman-1.5, and 126% against industry leader HeyGen.

Notably, the team adjusted their evaluation formula for this round, dialing down the relative weight of lip-sync accuracy while assigning higher importance to visual fidelity and full-body motion.

Why it matters

For marketing teams, corporate trainers, and solo creators building faceless video channels, single-prompt 5-minute generation cuts down hours of tedious timeline splicing. You no longer have to dice an explainer script into 30-second bites and pray the lighting matches between clips.

Across the AI video sector, the goalposts are shifting rapidly. Standard lip-syncing over static portraits has become a commoditized feature. The new frontier is expressive full-body acting, reliable gesture control, and temporal consistency over extended runtimes.

That said, we think Kling's choice to downweight lip-syncing in its internal benchmark raises a slight eyebrow. Hands and body sway matter, but in talking-head video, an out-of-sync syllable breaks the illusion instantly. Whether Avatar 2.0's acting chops truly hold up across multiple accents and pacing will be the real test for production use.