Back home

HeyGen Avatar 6

x+web 0.94
Visit source

Intro

HeyGen Avatar 6 is the talking-avatar model HeyGen introduced on 17 November 2026. The official research page, “Avatar 6: Talking Avatars That Are More Like You,” says it can generate a speaking video from a single photo, optionally plus a short recording as an identity reference, at up to native 2K (quoted as 2880×1620), with length following the audio and no fixed cap. A same-day PR Newswire release describes it as launched that day.

The official emphasis is not only lip sync. Avatar 6 is framed as an audio-driven, context-aware model that produces gestures and expressions that match the script’s tone, aimed at the repetitive or stiff motion common in AI avatars. The research page also reports a side-by-side “Turing test” figure of 84.4% indistinguishable from real footage, plus preference numbers against models such as Seedance 2.5 and Wan 3.0.

Developer docs are at the Avatar 6 model page. This entry records the new model name, not HeyGen the platform as a whole.

Advertisement

Explain

HeyGen describes Avatar 6 as audio-driven and context-aware: it matches mouth movement to speech and also performs gestures suggested by the script. The research page shows action cues written next to lines (a thumbs-up on “Hello,” pointing when the script mentions a counter), performed in that person’s style on cue. Identity consistency is another official claim: footage of the person is used to learn how they look and move. The press release says an avatar can be created from as little as about 30 seconds of video, with optional extra footage.

Resolution and duration are the other advertised limits. The research page says every frame can be generated at up to 2880×1620 and shows continuous 25–28 minute clips from a single audio track, without looping. Official comparisons against other commercial avatar products claim first place on multiple criteria and overall score, including Avatar Arena voting.

The launch materials also stress consent: third-party platforms may generate photo-based or virtual avatars, while realistic videos of real people remain on HeyGen’s platform and require explicit consent. Availability, pricing, and API limits follow the product and docs at the time you use them. Do not treat Avatar 6 as the same name as earlier HeyGen avatar generations or another vendor’s avatar model.

Usage

Read the official research page Avatar 6: Talking Avatars That Are More Like You, then the developer docs or the model picker in the HeyGen product. The usual path is to upload a photo (and optionally a short identity-reference recording and a voice), generate a talking video, and, if you need specific motions, write those cues next to the matching lines rather than only giving a vague style prompt.

When evaluating it, look at whether lips and gestures stay with the script, whether identity holds up on long videos, and whether any use of a real person’s likeness follows HeyGen’s consent rules. 2K output and very long durations depend on the plan and region then offered.

If you only need a voiceover, or an older HeyGen avatar model, that is not this new name. The published benchmark figures are official comparisons; independent checks should use a reproducible setup.