A personal avatar starts with your likeness. In October, we doubled down on the part that usually gives it away, your voice.
As a personal avatar company, we know it starts with your likeness, and that the job is to make it the most realistic version of you. But the face is only half of it. When someone can tell the real you from your digital twin, it is usually the voice that gives it away. A flat read, a pause in the wrong place, or the sound of a studio booth in a scene that was never filmed in one.
So October was a voice month. We shipped HeyGen Voice, our in-house model, which debuted at #1 on the Artificial Analysis leaderboard, making it the best voice model in the world by the industry's leading independent benchmark. Alongside it came a professional voice clone, a faster and cheaper voice mode, and speech-to-speech voice mirroring that keeps your delivery and changes only the voice. Around it came 30 new public LiveAvatars, higher-resolution and longer HeyGen Video generations, a refreshed template library, and HyperFrames Studio on the desktop.
Here is what's new, layer by layer.
Voice that sounds like you without recording every line
On October 9, we launched HeyGen Voice, our in-house voice model, which debuted at #1 on the Artificial Analysis leaderboard. Three more voice releases followed in the same month. Together they cover the range from a good voice to your voice to your exact performance.
- Narration that keeps your tone, pacing, and expression. HeyGen Voice preserves the characteristics that make a voice recognizable, and its scene-aware mode adapts the sound to the video's environment instead of carrying the original recording's room and microphone into every scene. Hand it a new script and the narration comes back sounding like you, not like a studio booth.
- A deeper clone for recurring content. Professional Voice Clone, a paid add-on launched alongside HeyGen Voice, trains a more thorough version of your voice. It is built for longer narration, weekly series, and personal brands where sounding like the speaker is the point.
- Faster and cheaper speech for pipelines. HeyGen Voice Turbo, announced October 21, is a lower-cost, lower-latency mode of HeyGen Voice through the API. The Turbo endpoint is free to try through November 30 on shared capacity, so treat it as a way to experiment rather than a guaranteed production tier.
- Perform the line once, change the voice, keep the delivery. Voice Mirroring, our speech-to-speech mode launched October 28, converts a recorded performance to a new voice while preserving the words, pauses, timing, and emphasis. When a whisper or a beat of silence carries the message, you no longer have to coax it out of text-to-speech by re-prompting. The API returns audio, and generating the matching avatar video is a separate step.
Presenters, with 30 new avatars and six emotions each
On October 7, LiveAvatar added 30 public avatars, ready to use without creating a custom one first. Each comes with six emotions, full-body presentation, and more natural expressions than the previous library.
- Pick the presenter the content calls for. A friendly host for customer service, an animated tutor for a course, a more measured presenter for compliance training. The expressive range is built in, so a serious script and a warm one don't need two different avatars.
- Or use yourself. For teams that want a personal avatar, the two customer stories at the end of this post show what a single recording session can sustain.
Footage with sharper shots and longer takes from HeyGen Video
HeyGen Video, our general-purpose video model, launched on September 30. October brought two upgrades to it.
- 2K output. From October 6, the API generates at 2K, so product shots, ad footage, and supporting B-roll hold up when they sit next to a 4K avatar scene instead of looking like the lowest-resolution element in the cut.
- Up to 30 seconds in one continuous take. From October 23, a complete short scene or ad-length shot comes back as one generation rather than several stitched clips, with fewer seams for a viewer to notice. Supported duration depends on the generation mode and settings, so not every request runs to 30 seconds.
Assembly that starts from something finished
The editing layer got a refreshed template library and a new agent-driven editor. The goal in both cases is the same. Begin with a video that already looks done, and save manual editing for the final adjustments.
- Templates that make avatar footage social-ready. A new batch of templates rolled out through October alongside a B-roll editor, stronger previews, and better routing between picking a template and editing it. Choose the visual treatment you like, drop in the avatar footage, and most of the work is already done.
- HyperFrames is now on your desktop. HyperFrames Studio is a desktop video editor built around an agent. Describe the video, your agent builds the first cut, and you direct the changes in the same window by drawing on the frame, commenting on an element, or editing the timeline yourself.
For developers
Most of October's voice and video releases arrived through the API first, and one of them reached agents directly.
- HeyGen Video in HeyGen MCP. From October 16, a connected assistant can generate a HeyGen Video clip through MCP. Describe the clip in chat and it comes back in the same conversation, with no separate integration to build.
- Voice Turbo, free through November 30. The lower-cost, lower-latency HeyGen Voice endpoint is open for experimentation on shared capacity until the end of November.
- 2K, 30-second takes, and speech-to-speech, all via API. The HeyGen Video upgrades and Voice Mirroring described above are available to API customers today.
Meet Holly Xiao, Head of Marketing at HeyGen. With deep expertise in product and growth marketing, Holly has led marketing teams at Drift, Envoy, and Canvas, crafting narratives that fuel business growth through clear positioning and storytelling. At HeyGen, she’s helping redefine how businesses use AI-powered video to scale enterprise communication and engagement.
