The latest AI news and updates of HeyGen — AI video generation platform specializing in AI avatars, digital twins, and video creation workflows for professionals and marketers. Covering HeyGen's latest product updates, launches, and company news from the past 90 days.
HeyGenSep 10HeyGen open-sourced a reference integration for real-time AI experiences, combining OpenAI's GPT-Live-1 with HeyGen's LiveAvatar and HyperFrames. The repo includes a Japanese tutor demo where the avatar triggers animated term cards and recap panels via tool calls. This starter framework provides the wiring for interactive, voice-first AI agents with on-screen visual overlays.
HeyGenSep 9HeyGen launched Professional Voice Clone in preview on the HeyGen API. This feature uses a dedicated voice model trained on twenty minutes of a single speaker to generate speech that matches the user's voice. The release includes an audio test comparing real speaker takes against the AI-generated clone to demonstrate the realism of the voice model.
HeyGenAug 26Harvard Business School built interactive AI avatars of its professors using HeyGen LiveAvatar for its HBS Foundry startup program. These avatars conduct practice sessions for pitches, sales calls, and board meetings on demand. The $699 course is currently live at over 100 universities, scaling faculty expertise through interactive, on-demand instruction.
HeyGenAug 22HeyGen launched Look Packs, a feature that updates an avatar’s outfit and setting while maintaining the original face. This tool ensures identity consistency across professional video content, including real estate, legal, fitness, education, and wellness. The system allows for varied visual contexts while keeping the same spokesperson visible to clients in every video.
HeyGenAug 22HeyGen launched an Upscale Video app that converts blurry listing footage into 4K resolution. The tool supports MP4, MOV, and WEBM formats, providing output resolution options of 1080p or 4K and frame rates ranging from original to 120fps. This update improves the visual quality of real estate listing videos to increase viewer retention.
HeyGenAug 22HeyGen launched Brand Kits to centralize visual and verbal identity. The system stores logos, fonts, exact hex codes, and a pronunciation glossary for avatars. This configuration ensures every generated video maintains brand consistency without requiring manual retakes or repeated adjustments across projects.
HeyGenAug 13HeyGen details the architecture of its Avatar V model, which processes the entire reference video at every layer to improve realism. This method preserves specific physical details like skin texture and teeth, while capturing micro-expressions and natural talking rhythms. By avoiding single-frame compression, the system maintains identity consistency across generated video content.
HeyGenAug 12HeyGen launched the Market Update skill for real estate video creation. The tool generates short, avatar-led videos from a single prompt, such as a script or listing link. A three-step confirmation process allows refinement of the script, voice, and format before the system produces a finished, ready-to-post video in approximately 12 seconds.
HeyGenAug 11HeyGen launched HeyGen for Real Estate, providing three video formats—market updates, home tours, and listing spotlights—that generate polished videos using an agent’s own AI avatar and voice. The platform eliminates the need for cameras or editing. New US Creator plan subscribers who sign up within 12 hours receive 30 days of HeyGen White Glove service free.
HeyGenAug 3HeyGen reports that an employee used a LiveAvatar AI sales agent to manage his inbox during 8 weeks of paternity leave. The agent handled 2,741 conversations, secured 132 new customers, and generated $3 million in pipeline. This deployment maintained full sales coverage without human intervention throughout the leave period.
HeyGenJul 29HeyGen launched Video Podcast, a tool that converts documents, links, or ideas into two-host video shows. The system automatically generates studio scenes, multi-camera cuts, and B-roll to produce a complete, publishable video. This feature moves beyond audio-only generation by creating a fully visual podcast format ready for immediate distribution.
HeyGenJul 23HeyGen updated its Video Translation API to support batch processing for entire content channels. The system localizes videos into up to 10 languages simultaneously, delivering finished channels in under 24 hours. Precision mode handles avatar inference and lip-sync for high-quality output, automating the translation workflow for large-scale video libraries.
HeyGenJul 23HeyGen launched Video Agent, a prompt-driven tool inside its platform that uses the HyperFrames framework to generate complete videos from a simple text request. The agent assembles avatars, motion graphics, captions, and music into a finished cut. This integration makes the framework’s production capabilities available as a hosted, conversational product for creating product intros and explainers.
HeyGenJul 23HeyGen launched an Audio to Video API endpoint that lip-syncs pre-recorded audio onto avatars or still images in a single call. The system accepts audio URLs or asset IDs, replacing script-based generation. It supports batch processing for multiple narrations on a single visual, with output controls for resolution, aspect ratio, and captions via asynchronous webhook notifications.
HeyGenJul 22HeyGen launched a Storyboard flow in HyperFrames that adds a three-pass pre-production stage to video generation. The agent creates a plan, a wireframe sketch, and a final build, requiring user sign-off at each step. Comments on specific storyboard cards trigger frame-level revisions, and the system saves the board as version-controlled markdown files.
HeyGenJul 22HeyGen launched companion mode for its HyperFrames framework, allowing AI agents to pitch five video angles, storyboard, and sketch frames for review. This workflow requires user approval at each step before the agent builds the final video. The process shifts video production from a single-turn order to a directed collaboration, preventing post-render surprises.
HeyGenJul 18HeyGen launched an AI Clipping API that converts long-form video into short, social-ready highlights. The endpoint automatically selects and scores key moments, or follows natural language instructions to target specific topics. It supports custom aspect ratios, burned-in captions, and webhook notifications for asynchronous processing, enabling programmatic creation of short-form content from long recordings.
HeyGenJul 18HeyGen added a media library to its HyperFrames framework, featuring 10,000+ music tracks, 75,000+ images, sound effects, and logos. Assets save locally to the machine, allowing agents to reuse files across projects without re-searching. This update removes the manual sourcing bottleneck, enabling fully automated video production runs.
HeyGenJul 17HeyGen launched batch APIs that support the submission of up to 100 videos, translations, lipsyncs, or uploads in a single request. The API returns a batch ID immediately and triggers a webhook once all items reach a terminal state. This update enables asynchronous processing for high-volume video workflows, replacing the need for individual API calls per item.
HeyGenJul 14HeyGen released a /figma skill for its HyperFrames framework that imports Figma designs directly into video compositions. The tool pulls brand tokens, converts frames into editable components, and translates Figma Motion timelines into GSAP animations. All assets are frozen locally, ensuring renders remain stable without live API calls, allowing agents to build and revise video from design files.
HeyGenJul 13HeyGen launched a CLI that provides terminal-based search and download access to its music and sound effects catalogue. The tool processes natural language queries on macOS, Linux, and Windows. This release integrates with the HyperFrames framework, enabling agents to automate beat-synced video production by pulling audio assets directly from the terminal.
HeyGenJul 12HeyGen released a new guide for its HyperFrames motion-graphics skill, which automates the creation of design-led motion graphics from natural language prompts. The skill generates charts, logo stings, and stat cards by autonomously handling all motion design decisions. It pulls public data from APIs or uses provided files like Figma vectors to create animations without manual editing.
HeyGenJul 12HeyGen now supports transparent background video output by setting output_format to "webm" on the /v3/videos API endpoint. This generates a video file with a real alpha channel, removing the need for green screens. The feature is available across the HeyGen API, CLI, and MCP for avatars trained with matting.
HeyGenJul 11HeyGen launched the /talking-head-recut skill, which automatically overlays synced motion graphics like titles and stat count-ups onto talking-head videos. The agent researches topics, writes scripts, casts avatars via CLI, and generates the final video. Revisions are handled through natural language instructions, allowing for the creation of professional video explainers without manual editing or camera work.
HeyGenJul 10HeyGen released a HyperFrames skill that generates launch videos directly from Figma design files. By pasting a Figma link and invoking the /figma command, the agent reads the layout to preserve exact hex colors, fonts, and frame designs. This workflow eliminates manual exports and plugins, allowing for direct conversion from design mocks to finished video assets.
HeyGenJul 9HeyGen released a music-to-video skill for its HyperFrames framework that generates beat-synced videos from audio tracks. The analyzer maps BPM, downbeats, phrases, and drum fills to lock visual cuts to the music. The agent uses typography and design systems to render the video, with revisions handled through natural-language notes that preserve the established timing map.
HeyGenJul 8HeyGen released a HyperFrames skill that converts open GitHub pull requests into narrated video explainers. The agent analyzes the diff, commits, and reviewers to generate a 30-to-90-second video featuring real code animations, captions, and voiceover. This single-generation workflow replaces manual diff reading with a visual brief, ending with the author’s avatar and an awaiting-review status.
HeyGenJul 7HeyGen launched a tool that converts any website URL into a finished promotional video. The AI agent analyzes the site’s CSS, fonts, and colors to write a script, add narration and music, and render the final MP4. Built on the open-source HyperFrames framework, the tool offers the first conversion free of charge.
HeyGenJul 2HeyGen open-sourced Keyframes, a visual editing system for GSAP animations within its HyperFrames framework. The tool provides timeline diamonds, arc motion, and gesture recording for precise control. AI agents can now inspect their own motion output and apply corrections with a single command, automating the refinement of complex motion graphics without manual code adjustments.
HeyGenJul 1HeyGen Research developed an autonomous agent that optimized the Wan 2.2 video decoder, achieving a 3x inference speedup on a single H100 GPU. The agent performed 18 experiments using Claude Code to reduce latency from 14.47s to 4.79s. HeyGen published the optimization harness and the resulting decoder code on GitHub for public use.
HeyGenJun 30HeyGen updated its avatar models to support single-pass video generation up to 30 minutes, a 10x increase over its previous limit. The new streaming inference framework processes video in chunks to maintain identity, lip-sync, and motion consistency without manual stitching. This capability is available now for Avatar III, IV, and V models via the web platform and API.
HeyGenJun 28HeyGen released a new generation of HyperFrames skills featuring nine creation workflows that AI agents now automatically route to based on context. Instead of manual selection, agents interpret natural language requests to identify the correct workflow for tasks like product launches, music videos, or captioning. This update automates intent detection and tool selection for agentic video production.
HeyGenJun 25HeyGen released HyperFrames skills that enable AI agents to automatically place graphics and captions within videos based on the message content. These skills translate professional editing patterns into repeatable, brand-adjustable agent actions. Agents now autonomously handle the placement of overlays and captions, reducing manual editing time for professional video production.
HeyGenJun 24HeyGen launched Look Packs, a feature powered by a new image engine that maintains consistent facial identity across AI avatar generations. The tool generates a full set of polished looks from a single photo, ensuring digital twins retain the same appearance in every setting. The first Look Pack is free upon creating a Digital Twin.
HeyGenJun 23HeyGen released a rhythm-aware skill for its HyperFrames framework that automatically syncs video composition to music. The system analyzes audio tempo, energy spikes, and rhythmic intervals to align visual cuts and transitions with the beat. This skill automates musical timing, allowing agents to build video around a song rather than manually editing footage to fit a track.
HeyGenJun 19HeyGen launched cloud rendering for its open-source HyperFrames framework, moving video composition processing to HeyGen infrastructure. The npx hyperframes cloud render command eliminates local dependencies like Chrome and FFmpeg. This offloads rendering tasks to the cloud, keeping local machines free so AI agents can continue building while video renders complete.
HeyGenJun 18HeyGen published an MCP server to the Cursor marketplace, enabling coding agents to generate product launch videos directly from a project's README. The integration extracts features and pricing from the repository to produce finished videos without manual filming or editing.
HeyGenJun 17HeyGen launched HyperFrames and a custom HeyGen connector on Grok, automating video production directly within chat responses. The integration provides tools for generating avatar-based videos and programmatic motion graphics. Both connectors authenticate via OAuth and draw on existing HeyGen plan credits to render MP4, WebM, or MOV files without leaving the Grok interface.
HeyGenJun 16HeyGen integrated semantic audio search into its open-source HyperFrames framework. Background music and sound effects can now be found by describing the desired vibe in natural language, with the system returning ranked tracks from HeyGen's library that integrate directly into video projects. This streamlines soundtrack selection for agentic video production workflows.
Frequently asked questions