Skip to content
TalkLive
Sign in
Live now on our own GPUs

A living avatar from one photo, on a call in seconds.

Upload a photograph. TalkLive gives it breath, blinks and a voice, and puts it on a live video call for your AI agent — lips in sync, interruptible, and yours to keep.

Create your avatarSee it in useFree trial: five minutes on air, watermarked. No card until you go live for real.
Otto, a TalkLive avatar, listening on a live support call Listening On air 04:12
I can see it. The share stalled on your side; give me a second while I check the ticket.Otto · help desk · speaking
Otto · help deskone photo · neutral idle · 67 fps · lips 320 ms behind the voice
1photo
is all it takes: a front-facing picture, or a five-second clip.
320ms
from the voice to the lips. Faster than you can notice.
<5s
for the face to join a call, warm on a GPU we run.
0lock-in
your photo, your voice, your agent. Export or delete any time.
About

Video calls should have a face on both sides — even when one side is an agent.

TalkLive Studio is the SEAMS AI alternative to D-ID. It turns one photograph into a photoreal avatar that breathes and blinks while it listens, and speaks in sync with any voice, for any AI agent you connect — support, front desk, coaching, teaching.

It runs on infrastructure we own: LiveKit for the call, our voice-worker pool for the agent, dedicated GPUs for the face. Your customers' calls never pass through a third-party avatar cloud, and the bill is the one on your plan.

The call
LiveKit WebRTC, self-hosted
The face
MuseTalk lips on dedicated GPUs
The idle life
LivePortrait from a neutral driver
The agent
yours — any brain behind one contract
Sign-in & billing
LunaOne: one account, one invoice
Consent
on file before any face or voice is cloned
Features

Everything a face needs to hold a real conversation.

Not a talking picture. A participant.

One photo, a living avatar

Upload a front-facing photo. The studio finds the face, generates a neutral idle loop — breathing, blinks, mouth closed — and stages it warm in under three minutes.

Lips in sync, interruptible

The mouth follows the voice within a third of a second. Talk over it and the face stops at once and listens; the idle life continues between sentences.

Any agent behind the face

Bind the avatar and a voice to your own agent — its brain, tools and memory stay yours. Swap the face in a click; the conversation carries on.

The voice survives the face

If a GPU ever drops mid-call, the caller keeps hearing the agent within two seconds and the call never dies. A second renderer takes the next call.

Consent first, disclosed always

A likeness consent is on file before a face or a voice can be cloned. Every session opens with the AI disclosure. The free trial is watermarked.

Metered air time, nothing hidden

You pay for minutes a face is on a call. Staging an avatar, training a voice and idle time in the library are free. Usage by day, in the studio.

Use cases

Wherever a person would sit across the table, put a face there.

Built with real teams on our own products first.

IT help desk

Otto, our support engineer, answers screen shares and tickets with a face and a name. Callers stop repeating themselves and start pointing at the screen.

live · seams.ai

Hotel front desk

A concierge that greets guests in their language from the booking, day and night, without a queue at reception.

live · hotelli

Coaching & care

A steady, patient face for check-ins that people would rather not type. Consent and disclosure built into every session.

pilot · lifecraft

Online classes

A professor's likeness delivering the lecture live and taking questions, while the professor teaches the next cohort.

coming

Sales & demos

A product specialist who is always awake, demos on screen share, and hands off to a human the moment it matters.

coming

Events & hosts

A host that opens the conference, introduces every speaker and answers the lobby — the same face for every attendee.

roadmap · full body
Docs

Read how it works before you trust it with a call.

Plain-language guides and the API reference. Short, honest, kept current.

Guide

Your first avatar in ten minutes

Choosing a photo, what the idle loop does, approving the preview, staging it warm.

Guide

Connecting your agent

The one contract an agent implements, the dispatch metadata, voices and languages.

Reference

API v1

Avatars, sessions, agents, voices, consents, usage. Every endpoint with its errors and limits.

Pricing

Pay for air time. Everything else is free.

Plans are managed on your LunaOne account, with your other SEAMS apps on one invoice.

Starter
$9 / month
Creators, one avatar.
  • 60 live minutes a month
  • $0.20 a minute after that
  • 1 avatar, stock voices
  • Watermark off
Start with Starter
Most chosen
Pro
$29 / month
Professionals, three avatars, the API.
  • 300 live minutes a month
  • $0.15 a minute after that
  • 3 avatars, one cloned voice
  • API access, usage and session reports
Choose Pro
Business
$99 / month
Teams, ten avatars, priority GPU.
  • 1,500 live minutes a month
  • $0.10 a minute after that
  • 10 avatars, cloned voices
  • Webhooks, priority GPU, a phone line
Talk to us about Business

Free trial: five live minutes, watermarked, no card. Air time is counted while a face is on a call, rounded to the second. Enterprise: pooled minutes, an SLA, dedicated GPUs and an on-prem option — quoted.

Getting started

Three steps, and the fourth is a phone call.

  1. 01

    Sign in with LunaOne

    One account for every SEAMS app. Pick a plan or start the five-minute trial; the studio opens from your LunaOne dashboard.

  2. 02

    Upload a photo, approve the face

    Front-facing, good light, mouth closed. Watch the idle loop, choose its mood — neutral, listening, warm — and stage it.

  3. 03

    Connect an agent, go live

    Bind a voice and your agent, press Go live, and share the link. Interrupt it, swap the face, end the call. The usage page shows the minutes.

Your avatar is one photo away.

Five minutes on air, free, to see it speak in your voice.

Create your avatar