HeyGen says Avatar V can learn a person's motion from a 15-second recording. Making a usable digital presenter takes more than that clip: the depicted person must give separate consent, the account must permit the intended video and file export, and the finished result needs their approval. This guide takes an adult creator making their own avatar, or an authorized team working with that adult, through one short English test video. It follows HeyGen's Avatar V instructions, which display a 9 September 2026 date. BIG CHANGE checked the documentation on 23 October 2026; we did not train an avatar, render a video or inspect an account charge.

The big change

  • What changed: Avatar V uses a short recording of a real person as a motion reference for a video-based avatar Look. HeyGen describes 15 seconds as enough source footage to start, followed by avatar creation and video generation.
  • Why it matters: A creator can prepare a reusable on-screen version of themselves, while a team can do so only with the depicted person's direct consent and approval of the intended use.
  • What to watch: The footage length is no promise about elapsed work or output quality. Export entitlement, identity and voice rights, Look type and the actual credit charge determine whether the first video is usable.

Before you record

Agree on the subject, workspace operator, one intended audience and purpose, a short English script, and who may approve the finished video. If the avatar depicts someone else, that adult must personally record the separate consent video in the creation flow, even if the operator has general team permission. The subject should also approve uploading their footage, any base image or voice sample, and the planned destination. Store these files and the private record below where only authorized people can access them.

Check the actual workspace before training, creating a Look or rendering. It needs Avatar V, a suitable Digital Twin slot, enough balance for the proposed actions, and paid local-file download at the resolution you need. HeyGen's export guide describes downloading as a paid-user feature. Its plan guide lists a 720p sharing link for Free and 1080p export for Creator; eligible higher plans list 4K. A sharing link does not complete this guide's local-file check. Record the account's actual plan, region, slot, export control, resolution, price and terms. If any required access is missing or uncertain, stop before spending.

HeyGen lists 48 credits per minute as the standard Avatar V video-Look rate, including Premium Credits on legacy plans. That is a published rate, not a quote for your project. Account rates and separate setup, Look or voice actions can differ. Set a maximum for setup and one test render, then read each displayed cost before confirming it. A slot or included balance does not establish that the whole test is free.

1. Record motion footage and separate consent

Record a continuous, well-lit clip of the subject facing the camera. HeyGen's input guidance calls for a clearly visible face, eye contact and expressive movement; include hands if their movement matters. Review the source locally for one visible person, clear audio, permitted background details and rights to any included material. Fifteen seconds is the documented source-footage entry point, not a quality guarantee. Do not use someone else's footage or synthetic footage as identity evidence.

During avatar creation, the same person must record the consent text shown by HeyGen. This is a different clip from the motion source. HeyGen requires the provided words read exactly, clear face and audio, a clip under 30 seconds, and a matching identity. It accepts MP4, MOV or WebM and offers webcam, phone/QR or upload routes. For a remote subject, let them submit their own clip through the offered route. A team note or a screen recording of another person's consent does not substitute for their recording. Record the consent submission and eventual acceptance privately; do not publish the consent file.

Expected result: two distinct, rights-cleared inputs and an account that can attempt avatar creation. If consent is rejected, read the specific reason on the avatar card before recording again.

2. Create and inspect the Digital Twin

In Avatars → Clone a Real Person, follow the video-based Digital Twin flow with the reviewed motion source and separate consent. Record the submission time and avatar identifier privately. Wait for the card's actual status. An upload acknowledgment does not mean the identity was accepted. HeyGen's error guide explains identity mismatches, unreadable consent and footage faults; correct the named problem with the subject or contact support.

Choose a video-based Look for Avatar V and inspect its appearance. A photo may be offered as a base-image input, but a photo-only Look follows a different model route. Use one approved Look and one motion reference for the first test. A standalone voice clone is optional. Skip it unless the subject separately approves cloning their voice and the workspace shows acceptable access and cost. Check the voice actually assigned to the avatar; motion, appearance and voice are distinct choices.

If the likeness or voice is wrong, hold the test before public use. Avoid deleting the avatar merely to clear a preview problem: HeyGen's Digital Twin FAQ says rebuilding a slot may be limited to once per billing cycle.

Expected result: an accepted avatar card and one reviewed video Look, with the selected voice and motion reference recorded. A pending or rejected card remains a hold.

3. Approve and generate one test

Choose Use in video on the reviewed Look to open Studio. Confirm that the selected avatar, video Look and motion engine actually say Avatar V. In Advanced Settings, choose the intended motion-reference video if there is more than one. Add the subject-approved script in one scene and inspect the assigned voice, pronunciation, background, resolution, music and other visible assets. HeyGen's Studio guide permits an audio preview, but full animated lip-sync is only visible after generation.

The following is a complete, optional first script. Replace it with the subject's own approved words if needed; it makes no product endorsement.

Prompt
Hello. This is a short test of my digital presenter. I am checking whether the voice, words and movement match the script I approved. This video is a generated test, and I will review it before anyone else sees it.

Read the actual credit estimate and balance immediately before Generate. Have the subject or authorized owner approve this exact script, Look, voice, motion reference, audience and charge. Generate once and record the project ID and time. If the page closes or the result is uncertain, inspect Projects, project status and the balance before deciding whether any retry is needed. A corrected render needs a fresh cost decision.

Expected result: one completed project with an observed credit change, or a precise failure or pending status. An editor preview is not a finished-video check.

4. Play the file and obtain the subject's decision

Watch the completed project in full with audio. Compare its words with the approved script and inspect face, gestures, mouth timing, voice, background, labels and artifacts. Ask the depicted person to approve the actual output for the agreed audience. If it misrepresents them or fails the intended use, mark it hold or revise within a newly approved budget.

Use Projects → ⋯ → Download, or the download control inside the project, at the resolution verified before training. Recheck the export control and any displayed charge. Open the saved file locally and play all of it; record its file name, duration, size, resolution and, if useful, a SHA-256 hash. A project view or sharing link alone does not prove that a playable file was saved. If export entitlement has changed, hold and resolve access before calling the task complete.

Check any destination-specific watermark, caption and disclosure requirements before sharing. Keep the reusable identity private until its owner approves who can use it and for which scripts. Approval of one video does not grant permission for later messages or audiences.

The decision is ready for the approved audience, revise, or hold. Ready requires accepted consent, matching identity/Look/model/voice, the subject's approval, a fully played local file, and a reconciled charge. If a charge looks wrong, compare the project and account history before another submission. If access or use becomes unauthorized, have the workspace owner restrict it and follow HeyGen's account and privacy controls.

Copy this private creation record

Fill this in privately as you work. A blank field means the check is pending.

  • Date; workspace/operator; depicted adult; approved audience and purpose:
  • Subject's rights, separate consent route and approval of uploads and intended use:
  • Plan/region, Digital Twin slot, starting credit balance and allowed setup/render maximum:
  • Before training: paid local-file export entitlement, required and available resolution, plan price/terms, observed control or stop reason:
  • Motion source duration, local review, permitted inputs and private file location:
  • Separate in-product consent clip submitted and accepted; time and avatar-card status:
  • Avatar ID; video Look; motion reference; selected model; selected voice; optional standalone voice clone skipped or separately approved:
  • Exact approved English script and approval time:
  • Displayed setup, Look, voice and one-render costs; final generation estimate:
  • Generate time, project ID/status, observed balance change; Projects/balance readback before any uncertain retry:
  • Full project playback: words, face, gestures, lip timing, voice, background and artifacts:
  • Downloaded file name, duration, bytes, resolution, complete local playback and optional SHA-256:
  • Subject's decision on actual output; destination disclosures; approved audience:
  • Final decision: ready / revise / hold; reason and next authorized action:

Keep footage, consent media, account identifiers and voice samples out of any public copy of this record.

Sources and further reading

HeyGen's Avatar V guide describes the model, video-Look restriction and Studio path; its displayed 9 September 2026 date is a document date, not a verified launch date. The consent instructions define a separate subject-recorded clip. Input guidance, the Digital Twin FAQ and error codes cover Look inputs, slots and rejection. The credit and plan guide, Studio guide and download guide explain the remaining account-dependent steps. These are provider instructions, not BIG CHANGE test results. The account UI, pricing and export entitlement may change.