Skip to content

Models & products

HeyGen Avatar 6 launches with gestures that follow the script

Edited and reviewed by AI Ping Editorial Team

3-line summary

  1. Script-aware gestures arrive for new and existing avatars.
  2. The company’s 84.4% is a transformed comparison score.
  3. Independent tests cover short clips; output and duration limits vary.
PracticalMarketersDesignersDevelopers
Conceptual fictional presenter gesturing beside a script stand and camera
AI-generated conceptual image: a fictional presenter coordinating speech and explanatory gestures · AI Ping

HeyGen launched Avatar 6 on November 17, introducing an avatar model that matches gestures to the meaning of a script. Existing avatars and looks can use the new engine. People producing presenter videos gain written control over how the on-screen speaker moves and expresses a line. The model is available through HeyGen and its API.

The product page describes pointing, counting on fingers and directional gestures tied to speech. HeyGen says the release improves on repetitive motion in earlier avatars. Users can specify emotion, expressions and gestures in greater detail, starting from a photo and optionally adding a short personal recording as a reference. Reference footage can supply the person’s expressions and movement.

HeyGen also lists MCP access from external AI tools and conversations involving two avatars. Its product page restricts likeness use to the person or someone with clear permission, and allows removal requests.

Two developer routes accept reference images or video plus supplied speech audio, or an existing HeyGen avatar look with a script and voice or recording. A lower-expressiveness setting produces smaller gestures and calmer expressions. Specific written instructions override that setting.

Both paths default to 1080p and offer 2K. Their documented 16:9 outputs differ, however: 2880×1620 for direct reference-media generation and 2560×1440 for the existing-avatar path. Realistic human reference video in the direct route also requires enterprise access and applicable consent conditions.

Conceptual reference headshot, pose photographs and a speech recorder
AI-generated conceptual image: reference media and supplied speech represented by photographs and a recorder · AI Ping

What the 84% figure measures

HeyGen’s 84.4% metric doubles the 42.2% of 270 decisive answers selecting generated footage as real in same-sentence, same-audio pairs. Evenly split choices would score 100%. Indeterminate answers are excluded. It does not mean 84.4% of respondents were fooled.

Company test: actual and generated video choices and metric formula
Editorial chart: paired choices and metric formula · HeyGen data, AI Ping

Independent evaluator New Measure asks creators to compare two providers blindly on the same script. Its report dated November 18 ranks Avatar 6 first for personal avatars and stock presenters, using October 27–November 16 votes. Votes were collected before the model’s public launch. Creator feedback included requests for fewer gestures and more natural expressions.

HeyGen typically took about three minutes. In 97% of completed comparisons, both videos were at most 35 seconds long. Scripts were capped at 500 characters and participants were self-selected. New Measure cautions against extending the results to long videos or insufficiently tested languages.

On monthly web subscriptions, Creator costs $29 with 600 credits, 1080p exports and Avatar 6 Turbo. Pro costs $49 with 1,000 credits and Avatar 6 Pro. Free users receive three watermarked videos per month, up to one minute each. Creator and Pro Avatar 6 videos are limited to 30 minutes.

Motion and gesture control has its own allowance: three uses monthly on Free and 20 on Creator. Pro is listed as “Maximum usage,” without a numerical allowance. Enterprise uses custom terms. These are web subscription conditions; API pricing is linked separately.

Sources