HeyGen Avatar V — Create Avatar

HeyGen Avatar V Create - Digital Twin Training API

What is HeyGen Avatar V Create?

heygen-avatar-v-create is the training endpoint for HeyGen's Digital Twin avatars. Pass a short reference video and the API enrolls a new lifelike avatar that mirrors the subject's body language, expressions, and delivery style. The endpoint returns an avatar_id (and avatar_group_id) immediately; HeyGen completes training asynchronously over a few minutes, after which the avatar_id becomes usable with the heygen-avatar-v model to generate talking-head videos at scale.

Key Features

• •Synchronous API: returns avatar_id + avatar_group_id in ~3.5 seconds
• •Trains from a single ≥15-second reference clip (mp4, public URL)
• •Captures full-body motion, hand gestures, and natural micro-expressions
• •One identity supports up to 100 different "looks" via avatar_group_id
• •Optional consent_reroute_url for branded consent flows
• •Pairs directly with heygen-avatar-v for downstream video generation

Best Use Cases

Confirmed in testing: the endpoint returns an avatar_id immediately and consent_status: "skipped" when no reroute URL is provided. Training itself completes asynchronously on HeyGen's side. Use this model to scale founder-led video, internal training, personalized outreach, and influencer-style campaigns — anywhere you need a reusable digital likeness instead of repeatedly filming the same person. The Digital Twin path is also ideal for multilingual content: train once, generate in any voice.

Prompt Tips and Output Quality

The single most important parameter is video_url. Shoot the reference footage in 1080p (or higher) at 30fps, keep the subject centered, use even lighting, and record a continuous take of 2–5 minutes for best fidelity. The clip must be at least 15 seconds and served with Content-Type: video/mp4 — shorter or wrongly-typed clips fail with "Footage is too short or too long". Avoid clothing with large logos, harsh shadows, and ambient noise.

How to Use HeyGen Avatar V Create on Segmind

heygen-avatar-v-create trains a HeyGen Digital Twin avatar from a single reference video. It's a fire-and-forget endpoint — the API returns an avatar_id in seconds, then HeyGen completes training asynchronously before you generate videos with heygen-avatar-v.

Quick Start

Upload your reference clip to a public URL (S3, GCS, Cloudflare R2) and POST:

{
  "video_url": "https://your-bucket.s3.amazonaws.com/founder-reference.mp4",
  "name": "Founder Demo Twin"
}

The response includes avatar_id, avatar_group_id, and status: "training". Save both IDs — avatar_id is what you pass to heygen-avatar-v.

Reference Footage Requirements

• •Duration: ≥15 seconds; HeyGen recommends 2–5 minutes for best quality.
• •Format: mp4 served with Content-Type: video/mp4 (S3 sets this automatically when you upload as .mp4).
• •Resolution: ≥1080p at 30fps; 4K/60fps is even better.
• •Take: One continuous recording — no cuts, no splicing.
• •Performance: Look into the lens, speak naturally, keep hand gestures in-frame.
• •Environment: Even lighting, quiet room, plain background.

Recommended Settings

Tested defaults:

• •name: a clear label like "Founder Demo Twin" or "Marketing Persona A"
• •avatar_group_id: leave blank on first creation; pass the returned value when adding more looks
• •consent_reroute_url: optional — only needed if you run a branded consent flow

After Training

Training takes a few minutes server-side. Poll heygen-avatar-v with the returned avatar_id (or wait ~10 minutes) until you get a clean generation. Once trained, the same avatar_id can be reused indefinitely across any number of heygen-avatar-v calls.

Cost reference

$1.25 per training run. Subsequent video generations are billed separately at $0.10/sec via heygen-avatar-v.

How long until the avatar is usable?

A few minutes — poll heygen-avatar-v with the returned avatar_id until generations succeed.

What does it cost?

$1.25 per training run.

Can I add more looks to one identity?

Yes — pass the existing avatar_group_id on subsequent calls.

Do I need a consent video?

HeyGen requires consent for production use; pass consent_reroute_url to route subjects through a branded consent page.

What format must the reference be?

Public URL serving mp4 with Content-Type: video/mp4, ≥15 seconds, single continuous take.

Can I delete an avatar?

Yes, via the HeyGen dashboard — note deletion resets monthly on most plans.