HeyGen Avatar 4 vs 5: Avatar IV and Avatar V compared

HeyGen Avatar 4 (Avatar IV) is the default rendering engine in HeyGen's API and works with every avatar type; Avatar 5 (Avatar V) is an opt-in engine for Digital Twin looks that list it, which HeyGen calls its highest-fidelity engine, using cross-reference-driven animation. You choose one per video with the engine field, and Avatar V costs more per minute.

Every HeyGen fact here comes from HeyGen's own developer docs and API pricing article, read on 2026-09-28: Models, Avatar V, Avatar IV, the Digital Twin guide and API pricing explained. "Highest fidelity" is HeyGen's description, not a test we ran.

What is the difference between HeyGen Avatar IV and Avatar V?

Coverage versus fidelity, in HeyGen's words. Its Models page tells new integrations to default to Avatar IV for broad coverage and opt into Avatar V "for the highest-fidelity motion and lip-sync". Avatar IV takes studio avatars, Digital Twins, photo avatars, arbitrary images and prompts. Avatar V takes only digital_twin looks that are eligible.

How do I use Avatar V in the HeyGen API?

Check the look first, then ask for the engine:

• Fetch GET /v3/avatars/looks/{look_id} and read supported_api_engines. Avatar V is available when avatar_v appears there; requesting an engine the look doesn't list returns 400.
• Send POST /v3/videos with type: "avatar", your avatar_id, script and voice_id, plus "engine": { "type": "avatar_v" }. Omit engine to get Avatar IV.
• Optionally pass reference_look_id: a digital_twin look from the same avatar group to use as the animation reference. Without it, the Digital Twin references itself.
• Drop expressiveness: it is Avatar IV-only, and sending it with Avatar V causes a validation error.
• Poll GET /v3/videos/{video_id} until status is completed.

How much more does Avatar V cost than Avatar IV?

HeyGen's API pricing article lists Digital Twin video at $7.20 per minute on Avatar V and $4.83 per minute on Avatar IV, charged by actual seconds generated; Avatar IV photo avatars are $2.31 per minute. The full per-engine table, with Avatar III, is in HeyGen API pricing.

Where does Avatar III fit?

Avatar III (avatar_iii) is the third v3 engine, also opt-in: HeyGen describes it as a dedicated photo-to-video pipeline, and it is the only one of the three with 4K output (except for photo avatars). A separate, older Avatar III on the legacy v1/v2 endpoints is not the same engine.

Does Sume have a similar quality choice?

Sume's avatar video, POST /v1/avatar-1.0/talking-video, has a quality field instead of engines: standard, plus (the default) or max, at 720p, the documented resolution. Each tier has its own per-second rate: $0.184/s standard, $0.245/s plus, $0.55/s max (no product image). The fields are on Generate avatar video; a digital twin on Sume is covered in Clone yourself with AI.