humava Create an avatar

Avatar studio for companies

A voice hums.
A presenter stands up.

HumAva turns a script and a voice into a photoreal presenter, a 3D character or an app-ready avatar — rendered on request, never picked from a stock library.

Create an avatar Open the editor
  • —engines, probed live
  • 21presenter voices
  • 0bytes leave the box on a free render
  1. Photoreal presenter
  2. Stylised 3D character
  3. App avatar · low-poly

01 · The forms

Seven forms a voice can take

Drag the orbit. Every form is a real pipeline on this machine, and its status is asked of the machine as the page loads — if an engine cannot run right now, the card says so instead of taking your order.

01 / 07

02 · How it’s made

  1. 00 · 0–4 kHz

    The voice

    Your script is spoken first — Kokoro on this box, or a cloned voice with recorded consent.

  2. 01 · scan

    The face

    A photo you supply, or a presenter we cast from your brief, is read line by line.

  3. 02 · ~723k tris

    The mesh

    Hunyuan3D or TRELLIS.2 lifts it into geometry you can turn.

  4. 03 · shade

    The render

    Light, skin and surface — locally, or on a frontier engine you approve by name.

  5. 04 · take 01

    The presenter

    The voice drives the face. Filed in your project, with a link.

03 · The voices

Thirty-four voices in eight languages, each with its own figure

Every plate is drawn from that voice’s own recording — its pitch sets one mode, its brightness the other, the way sand finds the still lines on a sounding plate. Hover or tap to hear it. All run on this machine. A language is only listed here after its voice was recorded and heard back correctly by a separate listener; anything else is on request, paid, and needs approval.

Open the voice studio — type a script, pick a voice, change its mood

04 · Takes

Real takes, straight off the renderer

Nothing here is a mock-up. Each frame is an order this platform actually completed, read from the order book as the page loads.

05 · The studio

Tune your avatar

Three steps. A free render starts at once on this box. A frontier render is quoted first and waits for a human Yes before it spends anything.

  1. 1 Form
  2. 2 Brief
  3. 3 Engine & order

Which form?

06 · The board

Your orders, resonating

Live. Each order walks the stations — voice, render, pack, deliver — and they light as it passes. A paid order holds at needs your Yes until somebody approves the spend by name.

Cut a video from your takes in the editor Admin room — approvals, spend, health

07 · Under the bonnet

The engines, and what each one actually is

No engine is described on the strength of a label. Press Check everything now and each row is probed against the live machine — the voice service is asked to speak, the Higgsfield bridge is asked for its balance.

08 · Pricing

Two tiers. One rule: nothing is spent without a name on it.

TIER A · INCLUDED

On this machine

£0 per render

  • All 21 Kokoro voices
  • Wan 2.2 speaking presenter — 640×640, 16 fps, up to 15 s a piece
  • Hunyuan3D, TRELLIS.2 and Pixal3D meshes
  • Character stills from the house pipeline
  • Nothing leaves the box
Start a free render

TIER B · FRONTIER

When it faces your customers

quoted in Higgsfield credits

  • Presenter video: est. 12–36 credits per 8 s
  • Influencer-grade presenter: est. 18–40 per 8 s
  • 3D character: est. 10–20 · full scene: 20–50
  • Character sheet: 4–10 · branded voice: 2–6
  • Parked at needs your Yes until approved by name
Get a quote

Credit figures are estimates from the provider’s observed bands. The approved order records the balance before and after, which is the real number.

09 · Honestly compared

Where we stand against the names you know

Picture quality is where the big studios lead, and we say so. Where HumAva is different is who holds the data and the money.

HumAvaHeyGenSynthesiaD-IDReady Player Me
What it isPresenters, 3D characters and app avatars, made to orderCloud talking-presenter videoCloud presenter video for training and commsTalking heads from a single photoStylised cross-game avatars
Picture quality todayGood on the free tier (640 px); frontier when it mattersHigherHigherComparable for headsStylised, not photoreal
Runs without the cloudYes — the whole free tierNoNoNoNo
3D mesh you own (GLB)YesNoNoNoYes
Spend needs a named YesAlwaysSubscriptionSubscriptionSubscription / credits—
Languages & editor polish34 voices in 8 languages, each proven by ear; a scene editor with captions, music and 16:9 / 9:16 / 1:1 export. More languages on request (paid)Many languages, full editorMany languages, full editorMany languages—

Competitor rows describe their public positioning at the time of writing; check their current plans before you decide.

10 · Questions

Asked often

Is the presenter on this page real output?

Yes. The particles are sampled from a 3D mesh HumAva built in order 98d6f79e, and the voice is a Kokoro take rendered on this machine.

Can you make an avatar of a real person?

Only with their consent, named on the order. It is written onto the delivery note, and the form will not take the order without it.

What do I actually receive?

Depending on the form: an MP4 with sound, a GLB mesh, a WAV voice track, or character stills — filed in your project folder, with a link.

Why are some engines marked research-only?

Their licence forbids commercial use. We use them for internal drafts and the order form refuses them for commercial delivery.

How long does a render take?

A voice takes seconds. A 3D mesh about a minute or two. A speaking presenter a few minutes per short clip on the free tier.