LiveAvatar by HeyGen

LiveAvatar is a real-time digital human interface by HeyGen that outputs 1080p video. The company claims the first frame takes less than 300 milliseconds and there is no upper limit on concurrency. It can clone an avatar from a single photo, making it suitable for AI customer service and live-streaming digital humans.

AI Video Tools#AI Video#Video Generation#Content Creation

Updated: 2026-10-27

What is LiveAvatar by HeyGen?

LiveAvatar is a real-time digital human interface launched by HeyGen. It turns a portrait that can speak, express emotions and make eye contact into an embeddable audio and video stream. Its positioning is different from HeyGen's own video generation products: the latter produces finished videos, while this one only provides a real-time channel that can serve one or tens of thousands of sessions at the same time. The official positioning is "Launch a digital human in five minutes from API Key to launch".

Who is it for?

**Backend engineers building AI customer service or sales assistants**: Need a stable and scalable speaker avatar component instead of maintaining rendering clusters on their own.
**Language learning product teams**: The official mentioned that Master English uses it to provide oral practice services for more than 60,000 students.
**Recruitment and sales training teams**: Zeligate used it to cut the average time to fill each position by 27 hours, and Copient AI used it to deliver over 2,500 hours of sales training.
**Offline display and holographic hardware manufacturers**: Proto Hologram integrated its digital humans into screens in physical spaces.
**Content teams creating digital avatars**: The official stated that content related to Reid Hoffman's digital avatars has garnered over 50 million views on social media platforms.

Key features

Real-time 1080p Rendering

The official states that the median time to the first frame is less than 300 milliseconds, delivering professional 1080p video with optional full-body or half-body framing, and capturing facial details, gestures, and movements.

Unlimited Concurrency

The official states that there is no additional charge for extra concurrent sessions, and the unit price decreases as the scale grows. For enterprise plans, it can be reduced to $0.01 per minute, with the API availability rated at 99.99%.

Avatar Library & Cloning

It comes with a library of over 100 preset avatars. You can also train a custom avatar using a single image or 2 minutes of footage, and push the stream to all users simultaneously after a single cloning.

Two Access Modes

The Full mode comes with a built-in voice stack that integrates speech recognition, large models, and speech synthesis, and is ready to use right out of the box; the Avatar Only mode only takes over the avatar image and streaming, with the voice part handled by your own model, resulting in lower costs.

Engineering details handled internally

Just embed one element on the frontend to play it, with interruption handling and reconnection after disconnection built-in. The official claims that developers can get it up and running in five minutes on their own.

Enterprise-grade support

The enterprise plan offers avatar fine-tuning, dedicated capacity and SLA, single sign-on, an admin dashboard, and audit logs. The official website also provides additional compliance and data protection instructions.

How to use LiveAvatar by HeyGen

**First get the pipeline running with the free tier**: The $0 tier includes the complete API and 1080p preset avatar library, 10 credits per month, limited to 1 concurrent stream and a maximum duration of 2 minutes per session.
The output has a watermark and is suitable for verifying latency and video quality.
**Upgrade once you have real users**: The Essential plan at $99 per month includes 1,100 credits, removes watermarks, offers unlimited concurrency, supports 20 minutes per session, and comes with one 720p custom avatar.
**Go Live with the Business Plan**: $475 per month includes 6,000 credits and one 1080p custom avatar, with a single session lasting up to 60 minutes.
for higher volume, upgrade to the Enterprise Plan to discuss exclusive capacity.
**First calculate the voice cost clearly**: The Full mode is convenient but includes speech recognition and synthesis in a package.
the Avatar Only mode uses your own model for lower costs, but you need to handle interruptions and timing on your own.
**Be sure to obtain authorization for cloned images**: The official repeatedly emphasizes that portraits must be used with the owner's consent, and in regions such as the EU, data retention must be handled in accordance with GDPR.