HeyGen’s June updates include new HyperFrames skills, advanced avatar tools, live avatar APIs, Speech Cleanup, cloud rendering, and deeper integrations across developer workflows.
June was a big month. We hit $200M in ARR, roughly doubling in about eight months. We celebrated for about five minutes. Then we shipped 22 launches. Here's what we built.
The next generation of HyperFrames skills
Skills are pre-built agent workflows. In June we shipped the next generation of HyperFrames skills, focused on workflows, bringing the platform to nine in total:
- Website to video gives the agent a URL and gets a finished video back. It reads the site's content, structure, and branding and scaffolds a complete video from it.
- PR to video takes a GitHub pull request and outputs a narrated video walkthrough of the code changes. No one reads changelogs. People watch videos.
- Beat sync inverts how most video is made. Instead of choosing music and editing to match it manually, you give the agent a song and it analyzes downbeats, swells, and bass drops to design the video around the music. One of our developers described it as getting to "better outputs much, much faster."
- Captions / video-use handles the full caption pipeline in one instruction: transcription, timing, brand-consistent formatting, and rendering. Captions aren't optional anymore. Most people watch video on mute. This makes captions the default.
- Launch video takes a product brief (name, features, tone) and produces a finished, publish-ready launch video. Scripting, animation, voiceover, captions, motion all handled.
At the end of June, all nine HyperFrames skills became a unified, intelligent platform. The agent routes automatically based on context. You describe what you need. It goes to the right workflow. This is the shift from "code-based video tool" to "agentic video platform.”
We also added additional capabilities to HyperFrames
- Code Snippets gave the agent a library of pre-built, tested HTML building blocks to pull from. Instead of reinventing animated title cards and data visualizations every time, the agent has proven components ready to assemble. Faster output. More consistent quality.
- Cloud Rendering was the unlock that made HyperFrames production-ready. Before this, rendering required headless Chrome and FFmpeg running locally. Now you push the composition and rendering happens in HeyGen's cloud. That means HyperFrames can live in CI/CD workflows, serverless functions, and scheduled pipelines without anyone's laptop being involved.
- Native music and sfx turned audio from an afterthought into a one-line instruction in the HeyGen CLI. 100,000 background tracks, sound effects, a free tier of tracks, all built in — all free. No external sourcing, no manual integration.
Avatar technology hit new ground
Avatar Look Packs
Look Packs are a new way to create looks for your digital twin that actually look like you. Most AI-generated looks only kind of resemble you, with a slightly different face every time. Look Packs solve that.
Powered by Image N, our new AI image engine built for generating images of people rather than general-purpose scenes, Look Packs give you a polished, persona-based set of consistent looks in one tap. The same person across every outfit, every setting, every context. No more generating one look at a time and hoping it lands. Start with health and wellness, legal, teacher, or fitness, and your twin shows up looking like you every time.
Avatar Realtime API (Broadcast)
This changes what an avatar can be.
Until now, HeyGen avatars were pre-rendered. You write a script, render a video, publish a recording. The Avatar Realtime API makes the avatar a live streaming agent. It can browse the web, see real-time inputs, and respond in the moment. The difference between a voicemail and a phone call.
We demonstrated this by putting an avatar on Twitch 24/7. Viewers sent GitHub repos in chat. The avatar opened them, reviewed the code, and gave live feedback.
Link: https://developers.heygen.com/avatar-realtime
World's longest talking video: 30 minutes, continuous
HeyGen can now generate 30 minutes of continuous talking-head video in a single pass. That's 6x longer than the previous industry ceiling and 10x our own prior maximum.
The technical achievement isn't just the length. It's the consistency. Most AI video systems drift over time: the face gradually stops looking like you, the voice subtly shifts. Our streaming inference framework stays locked on your likeness and voice for the full 30 minutes.
Full online courses. Employee onboarding modules. Recorded presentations. These are now in scope for AI video. Available on Avatar III, IV, and V, on both API and web.
Link: https://developers.heygen.com/models
Cinematic Avatar API
Cinematic Avatar API takes a natural language prompt and up to three Avatar Looks, then returns cinematic footage with your real face in it. Your coding agent can call it directly, which makes cinematic video one line in your pipeline instead of a project on your calendar.
Link: https://developers.heygen.com/cinematic-avatar
New content tool: Speech Cleanup
Record a rough take. Upload it. Speech Cleanup removes every filler word, awkward pause, and false start, then stitches the remaining footage into a single seamless take with no visible jump cuts.
Every other filler removal tool on the market (Descript, CapCut, Submagic, Cleanvoice, Adobe, Riverside) cleans the audio. The fillers disappear from the sound but the video still jumps. Speech Cleanup hides the cut visually. The final video reads as one continuous, unbroken take.
Pricing is transparent: you pay per cut applied. A perfect take gets no charge.
HeyGen is now inside every major developer tool
June was also the month we became native to wherever developers work.
- Claude: HyperFrames became an officially listed MCP connector in Claude. From inside a Claude conversation, ask for a video. Claude calls HyperFrames. No context switching.
- Grok: HyperFrames is also in the Grok connector directory. You can prompt a video from directly inside a Grok conversation.
- Cursor: HeyGen became an officially listed plugin in Cursor's marketplace with three skills bundled together: heygen-avatar, heygen-video, and heygen-translate. They chain automatically. Write a video demo inside your coding environment without switching tools.
- Lovable: Connect it as a personal connector and avatar video becomes a feature of your app, whether it’s for onboarding, product walkthroughs, in-app explainers. You can build an app in an afternoon now. Getting people to understand it is the hard part. Now your app ships with a voice, face, and first impression.
Three major AI platforms in nine days.
Link: https://developers.heygen.com/mcp/overview
We also launched native LinkedIn publishing: create a video in HeyGen, post it to LinkedIn directly. No download, no re-upload, no friction.
And the HeyGen x Stripe integration extended video into the business layer: paywalled content, video-driven checkout flows, monetized product launches, all automated.
Link: https://developers.heygen.com/docs/stripe-projects
Also shipped
Customer support agent: Meet your new in-app support agent. Ask it the everyday stuff, like "how do I create my avatar?" or "where do I change the language?", and get an answer in seconds while you're working. No ticket, no waiting, no digging through help docs. And it keeps up as we ship, so the answer is always current.
Questions? Reach out at [email protected] or find us on X @HeyGen.
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.
