Avatar V needs 15 seconds of you; a Sume avatar needs one still

HeyGen Avatar V needs a 15-second webcam recording of you and gives back a digital twin. Sume Avatar 1.0 needs one reference photo, or a text prompt, and creates a reusable avatar for a one-time $0.95; it does not learn gestures from a recording, so pick it when a still and a script are all you have.

What HeyGen says it needs

The Avatar V page says '15 seconds to create your avatar' from a short webcam clip. It lists 175+ languages and dialects with phoneme-level lip sync, a G2 ranking as the most realistic avatars, and consistent character across scenes. It says there is no cap on video length or quality. No price appears on the page; it points to a separate pricing page.

The page does not describe any numbers for training time, and we do not guess them here.

What Sume needs

Sume has three ways to make an avatar: a text prompt, structured traits (ethnicity, sex, age), or a reference image. The image must be a fetchable public HTTPS URL. If you only have a recording, export one clear frame from it and use that as the photo.

Each request is a job. When it completes you get an avatar handle, which you pass to POST /v1/avatar-1.0/talking-video with a script.

Cost of the second video onward

After creation, you pay for seconds of video. At plus quality with no product image the rate is $0.245 per second, so a 45-second clip is 45 x $0.245 = $11.03 (rounded). Ten such clips plus the avatar is 10 x $11.025 + $0.95 = $111.20. The same avatar handle is used in all ten.

When each fits

• You want your own movement and voice captured: a recording-based twin.
• You want a presenter who is not a specific real person, or only have a photo: a Sume avatar.
• You need to run it from a script or spreadsheet: Sume exposes it as an API route with Idempotency-Key.
• You need videos longer than 60 seconds: split the script across jobs.

What to do

Create the avatar with a still, then generate a 10-second test clip at standard quality first: 10 x $0.184 = $1.84. If the face and voice read right, move up to plus.