LiveAvatar API billing: credits, conversation modes, and total cost

When researching HeyGen live avatar API pricing, distinguish generated video from a LiveAvatar conversation. LiveAvatar supports interactive sessions. Treating that product as a sequence of ordinary video-generation jobs misstates what its API does.

FULL and LITE use different units

LiveAvatar documents a managed FULL mode and an avatar-only LITE mode. FULL includes speech recognition, an LLM, TTS and WebRTC management. LITE supplies the avatar video while you provide the conversational stack and manage WebRTC. Compare the architecture you intend to use in the official mode overview.

Its billing documentation, checked on 5 September 2026, specifies two credits per minute for FULL and one for LITE. A credit balance therefore buys different session durations in the two modes. Divide included credits by the applicable rate before comparing minute allowances.

Check what happens after the allowance

LiveAvatar's billing guide says a session needs sufficient credits for at least a minute to start. It describes per-minute session deductions and an overage setting: when overage is disabled, exhausted credits stop sessions and prevent new ones. Confirm the current subscription, rate unit and enabled settings in your account before estimating a bill.

A complete worksheet includes the base subscription, included minutes, overage, simultaneous sessions, maximum session length and avatar-creation charges. For a rendering mode, add the speech recognition, language model, synthesis and infrastructure you supply.

Realtime Avatar's published rates

Our live-call product includes speech recognition, the language model, voice and avatar rendering. Published self-serve overage is $0.08–$0.095 per minute, or $4.80–$5.70 per hour. The Developer plan is $24 per month with 600 included call minutes and $0.085 per minute overage. Photo-avatar creation beyond the included count has a published price of $1.

Included credits are shared with generation uses, so spending them on avatar or asset generation reduces what remains for calls. The pricing calculation guide separates subscription cost, allowances and marginal overage, with a calculator for your own usage assumptions.

Compare an actual conversation

Cost is one input. Test startup, a fixed spoken question, an interruption and explicit session ending on the same browser and network. Keep failed and queued attempts in the report. This article makes no measured claim that one provider is faster than another.

Use our developer evaluation checklist and timing methodology. The Next.js starter gives you a runnable Realtime Avatar integration to include in that trial.