ALPHA personas and avatars

A persona is who a talking preset can put on screen

Personas, at Personas, lists who ALPHA can cast when a preset puts someone on screen. The page says it in one line: a persona is cast only when it is active, with its locked still, its voice and its disclosure, and a digital twin only with a written release on file.

The list has two parts. This workspace holds your own spokespersons and digital twins. Nagent library holds the presenters and creators Nagent provides, which any workspace can cast.

Each row shows the persona's name, its id, its kind and its state: Can be cast, Draft, Waiting for approval, Cannot be cast yet, Revoked or Retired. When a persona cannot be cast, the row says why, for example that a twin has no written release on file, or that its release expired on a date. Under the name sit its disclosure line, its release and its last day, and how many lanes are ready.

There are two kinds of persona you can make:

A spokesperson can be cast by any preset that puts someone on screen, except a twin preset. A twin preset, such as Founder twin, casts only your workspace's own twins. A twin is never cast anywhere else.

A workspace admin makes and manages personas. Everyone else sees the list and picks from it under Who is on screen in the studio's Settings.

Step 1. Create the draft

Click New persona. Choose Brand spokesperson, a face the workspace owns or Digital twin, a real person with a written release. Give it a Name, which is how the storyboard names them, and a Disclosure line, which is shown on every video, such as "AI presenter". Click Create draft.

A new persona is a draft. Nothing is made at a vendor and nothing is spent until it is approved.

Step 2. Complete it

Click Complete on the draft. The form holds:

• Name and Master still. Pick the still from your library. The still is the face every avatar keeps.
• Avatar (Beta). How the avatar is made has three choices. From the master still makes the avatar from the photo. From words, with the master still as its reference asks you to describe them, as the avatar should look: age, look, clothing and setting, in up to 1,000 characters; the master still keeps the face. From their own footage is offered on a digital twin only, and takes 15 to 600 seconds of the person facing the camera, an MP4 or WebM up to 32 MB, picked from your uploads with Pick footage.
• Voice engine and Voice id at the engine, for a voice you already hold at Fish Audio, Kinetiq or ElevenLabs. Leave them empty to make a voice here instead (next section). A voice id typed here replaces a voice made here.
• Wardrobe, Setting and Delivery, in words, and the Disclosure line.

Click Save.

A twin made from footage carries a note worth reading before you pick the footage: the twin is made only with a release that covers their voice, since their voice is cloned from this footage as the twin is made. Once it is made, they give consent on camera from a link you send them.

Step 3. File the written release, for a twin

A digital twin is a real person, so their signed release goes on file before the persona can be approved. Under Written release:

1. Pick The signed document, from this workspace's library. Upload it to your library first if it is not there.
2. Tick What it covers: paid social, organic social, website, email, internal, and their voice. Tick their voice if the twin will speak in the person's own voice, cloned from a recording or from their footage. Without it, their own voice is refused and the twin cannot be made from footage.
3. Fill Territories, as country codes, such as IN, AE, and Last day it covers.
4. Click File the release.

The row then reads Release to that date. When the last day passes, the persona can no longer be cast, and a cloned voice is deleted. Filing a new release starts again.

Step 4. Give it a voice

Under A voice made here (Beta) you can make a voice instead of naming one by id. Nothing here shows you a vendor's name or id; the voice is simply the persona's.

Design a voice. Describe it, for example "Warm, unhurried, a light Mumbai accent, mid thirties", pick a Voice gender (Any, Female or Male) and an Accent, as a locale, such as en-IN, and click Design voices. ALPHA offers a few voices, each with a preview you can play. Click Use this voice on the one you want. An offer is good for a day; after that, design again. A designed voice is nobody's own, so it needs no release.

Clone their voice, on a twin only. Pick A recording of them speaking, an MP3 or WAV from this workspace's library and click Clone their voice. This needs a release on file that covers their voice; until one does, the button says so.

Use the voice from their footage, on a twin made from footage. The voice arrives when the twin is made, and the row says so until it has.

A cloned voice, or the one a twin brings, is a real person's. It is deleted when the persona is revoked, when the release expires, or when a new release stops covering their voice. The persona then says its cloned voice was deleted when its release ended, and cannot be cast until it has a voice again.

A spokesperson can also carry a Long-form presenter (Beta): a stock face and voice that present long videos, such as a learning series. That section is described in VIDYA; a digital twin does not offer it.

Step 5. Ask for approval

Click Ask for approval. The row reads "Waiting for approval from another admin of this workspace, or from Nagent. The person who asked cannot approve it." The approval goes to Actions like any other.

What the approval does:

• The persona becomes active, and its still, voice and disclosure are locked. A face and voice picked for a long-form presenter are approved with it.
• The avatar is made, the way the draft says: from the still, from the words, or from the footage. Making an avatar is paid work, so ALPHA checks the price against its budget first and records the spend once. An avatar with no price set yet is not made, and the persona says why.
• Each lane that can put the persona on screen gets its own copy. The row counts lanes ready as they finish; a lane that is not set up for your workspace stays pending with a note.

A spokesperson can be cast as soon as its lanes are ready. A twin has two more steps.

Step 6. For a twin: bindings and consent

A real person's likeness is bound to a vendor only after two approvals. On an active twin with no lane ready, click Ask for bindings. The row then shows Vendor bindings: the workspace waiting, Nagent waiting, and says both must approve before a real person's likeness is bound. Another admin of your workspace approves one, Nagent the other.

A twin made from footage also needs the person's own consent, given on camera. The row reads Their consent to the twin, on camera with one of four states: not asked yet, waiting for them, given, or refused, so the twin cannot be used.

1. Click Get the consent link. A link appears in the row, with a note: send this to the person. They record their consent on camera at the avatar maker's page. It works for 24 hours; ask again for a fresh one. It covers their face only; their voice is the release's.
2. Send the link to the person yourself. Nagent does not email them.
3. They record the statement. ALPHA reads the answer on its own and the row updates; it is not told at once, so allow a little time.

The twin is ready only when its look is trained and the consent is given; ALPHA reads both as it binds the twin. A refused consent fails the binding, and no new link is offered.

What a persona looks like in a video

A synthetic presenter always carries a disclosure layer with the persona's disclosure line. That holds for a spokesperson, a library creator, and a persona moved by a motion clip (see the studio page for motion transfers). A digital twin carries its release instead, and renders only while the release stands.

In the studio's Settings, Who is on screen lists the personas this preset can cast; a persona that cannot be cast yet is not offered. Beside it, Library voice lets you keep The persona's own voice or pick one of Nagent's library voices, each with a short brief and a sample to play.

The AVATAR desk's Cinematic persona preset (Beta) puts your spokesperson to camera and then in one cinematic scene made from its look, which speaks no line. A cinematic scene is a flat price whatever its length, up to 15 seconds.

A persona made from its still may speak straight from that still: the model makes the picture and the voice together, in its own voice rather than the persona's calibrated one. The disclosure layer stays.

Revoking a persona

Revoke is on every persona you manage. It asks why, and warns that this is final: new videos stop using the persona, open jobs stop, and every video made with it is marked do not reuse. Click Revoke with the persona's name to confirm, or Keep.

Revoking a twin also deletes its cloned voice and its vendor copies. The reply counts the assets flagged and the jobs stopped. Mission control counts releases ending within 30 days, so you can renew before a persona stops.

Where to go next

• ALPHA, the video studio: the presets that cast a persona, and how a job runs.
• VIDYA: a persona as the presenter of a learning series.
• Approvals and the inbox: where persona approvals and bind approvals are decided.