HeyGen Voice STS carries a spoken take onto an avatar
Say a line once and HeyGen uses that timing on an avatar. The speech-to-speech API costs $0.12 a minute and returns audio only.
Every HeyGen update in one feed — release notes, features and changes, straight from the source.
46 updates
Say a line once and HeyGen uses that timing on an avatar. The speech-to-speech API costs $0.12 a minute and returns audio only.
Voice Turbo, a faster and cheaper mode of HeyGen’s voice model, is live on the API and tops Artificial Analysis’s text-to-speech leaderboard.
HeyGen’s new video model covers the rest of the shot beyond avatars, such as product close-ups, b-roll and cutaways, and is live on its platform.
Developers get MIT-licensed avatar demos, a 168-brief motion-graphics benchmark and API voice cloning from recorded speech.
The benchmark from HeyGen, HyperFrames and Kaggle scores AI-generated motion graphics against 168 human-made reference videos.
HeyGen adds a real-estate video workflow and Edit Look, which updates an avatar's appearance once for all future videos.
Real estate agents get three video formats, and existing avatar users can retouch their appearance without recording again.
The updates are live, with 30-minute avatar videos available through both the API and web app.
US agents on the Creator plan get 30 days of White Glove free at launch.
HeyGen's new Video Podcast adds studio scenes, multi-cam cuts, and B-roll instead of just AI-narrated audio.
One prompt now drives avatar, motion graphics, captions and music together, with HyperFrames still customizable on request.
One call turns long footage into up to ten confidence-scored social clips, with automatic, instruction-guided, or timestamp selection.
Agent video framework HyperFrames now sources music, images, SFX and logos in-workflow, free with a HeyGen login.
New endpoints let developers queue up to 100 avatar videos, translations, lipsyncs or uploads and get a single webhook when they're all done.
Run it via npx to let an AI coding agent inspect and correct generated motion straight from the command line.
It recognises nine workflows, from launch videos to captions and overlays, and starts the right one without manual picking.
Running `npx hyperframes cloud render` skips local Chrome, ffmpeg and AWS setup and frees the machine while it renders.
Grok answers can now be turned into an AI avatar video without leaving the chat.
The May 2026 release also brings a real-time Digital Twin, four workflow integrations and a full Android app.
Viewers post a GitHub repo in chat and the avatar reviews it live, flagging bad design decisions as it goes.
The API lets developers script avatar likeness and expression into a coding-agent video pipeline instead of using the app.
It converts an existing design.md into video-native guidelines covering motion and framing for agent-generated video.
An MCP and OAuth integration lets anyone building on Lovable add a talking avatar without leaving the app-building workflow.
Variables for text, colors and clips run parallel batch renders, and HeyGen published a migration guide for teams moving off Remotion Lambda.
Video generation, translation and other endpoints get lower rates, with quality and lip sync unchanged.
The model turns a 15-second recording into studio-quality video with stable identity across camera angles.
Priced at $0.05 per second; HeyGen says it beat Veo 3.1, Kling O3 Pro, OmniHuman 1.5 and Seedance 2.0 on talking-head benchmarks.
A tutorial covers prompting eye contact, pacing, posture, gestures and energy shifts in avatar performances.
Custom Motion lets creators type cues like a lean-in or thumbs up for Avatar V to perform.
The AI-avatar tool integrates with Canva's 190 million-plus users for decks, training docs, social posts and marketing assets.