HERE COMES LUMI0%

Try Lumi

Give your character a life beyond the model file

Lumi
ANIWAFFLE / LUMI02

We turn one identity into aliving character system.

ANIWAFFLE API
VOICEFACEMEMORY
STUDIO
LIVING CHARACTER
FOR DEVELOPERS

It all starts with our voice API.

Read the docs Get an API key
Lumi inside a character game
DAY 15Make a plan with Lumi

LUMI

The café is still open… What should we do?

72Bond84Mood61Ideas

Real-time response

A character feels present when it can answer as quickly as you speak.

How long before the first word

US0.3s
CLOUD TTS A0.9s
CLOUD TTS B1.4s
CLOUD TTS C2.2s
Sample data · same scale

What the same line costs

US1×
CLOUD TTS A9×
CLOUD TTS B14×
CLOUD TTS C22×
Sample data · same scale

Turn a single recording into a voice you can use anywhere.

01
Reference
02
Cloned

voice responds with natural emotion and expression, automatically.

Lumi Voice

Reference · illustrative

Choose an emotion

New line, synthesized

Emotion lives in the finer details of a voice. Our model is built to preserve them.

AniWaffle, 48 kHz
AniWaffle, 48 kHz
Most other AI voices, 24 kHzTheir output has a 12 kHz ceiling, leaving the high-frequency band above it empty.
Most other AI voices, 24 kHz

Most other AI voices stop at 12 kHz. AniWaffle preserves the high-frequency detail above it.

Whisper
ねえ、聞こえてる？ ……大丈夫、ここにいるから。ゆっくり眠って。

[ See the Voice API ]

Our Avatar API gives it more life than you think.

Read the docs Talk to her now

Turn a VRM into a real-life character in seconds.

Read the docs Talk to her now
02

It speaks and reacts

Your Agent's voice, and a face that keeps up.

It speaks in your Agent's real voice, and what it feels comes through in that voice and on its face at the same time.

Laughing Let down Not impressed Annoyed Curious

Five posts on a phone, five different faces.

03

Create once.
Bring her everywhere.

One Agent ID. The same Lumi.

Create her in Studio and publish her once.
That one Agent ID drops her into your app, your website,
your desktop or your game, with the same character,
voice and avatar every time.

One Lumi. Across your apps, websites and games.
Lumi across a smartwatch, AI charm, smartphone and handheld character-raising game.

Avatar API

Try
Talk to her now
Docs
Read the docs
Product
See the Avatar API

Start creating.

Download Studio
[FAQ]

Questions, answered.

More questions? Contact us

No. The React SDK loads, verifies, caches and renders the avatar published with the Agent, with lip sync, expressions, motion and gaze timed to its live audio. One package, no separate avatar runtime.

Yes, including both in one sentence. Voice runs on open and closed source routes, with the Japanese vocabulary most voice models lack. Expressive voice tags on routes that support them.

Voice and avatar come in one plan. You do not pay for a voice API and then an avatar on top. Plans and limits are on the pricing page. See pricing

Yes. @animated-waffle/client is a framework-independent browser client for your own UI. @animated-waffle/server is a typed Node.js SDK. The public /v1 API, HTTP and WebSocket, gives lower-level control.

Memory is optional and scoped to one end user by the stable endUserId your backend passes, so one person's context never mixes with another's. Monitoring shows the latest 30 UTC days of Sessions.

The voice engine runs on the device. Your reference voices and the lines it speaks stay there. Sessions with an Agent are a separate matter: what Monitoring keeps for a Session is in the docs. Read the docs

Who we are

AI characters for everyone, powered by our API.

Learn the API
RESERVED FOR DEV
DEVELOPER PLATFORM / API

One character.
One API.

Publish the complete identity once, persona, memory, voice, avatar, emotion and lip sync, then bring that same character into any product.

ANIWAFFLE SDK● LIVE
const lumi = await aniwaffle.characters.get("lumi")

await lumi.connect({
  voice: "realtime",
  avatar: "live2d",
  memory: true
})
READ THE API DOCS ↗
Questions

Before you ask.

Do I need to be a developer?

No. Create a talking Agent, choose its voice and Avatar, then test the conversation live, all in the Dashboard. The SDK is there when you want the Agent inside your own app.

Can I bring my own voice?

Yes. Upload clean recordings or record the speaker in the Dashboard, or import a voice from your ElevenLabs or Fish Audio account. Built-in voices are ready without any setup.

Where does the Avatar file live?

Not in your app. The connected session supplies short-lived Avatar assets automatically. Applications select the published Agent, not an Avatar file.

Is my API key safe in a browser?

Workspace keys never leave your backend. Browsers use short-lived session tokens instead, each valid for 15 minutes and good for exactly one conversation.

Your character. Live. Expressive. Real.

Bring your character to life, right now.

Start building