Give it an identity
Choose a face, define personality, and assign a voice that sounds unmistakably yours.
Thevideo,voiceandconversationlayerforAI.
Video, voice, and real-time conversation for any being you can imagine, humans, mascots, game avatars, and brand personalities. Built in one studio. Shipped anywhere.
Built for production
Proof, not just a demo.
Sunesis Studio
One canvas. Every layer.
Voice, video, and conversation, generate, preview, and ship from a single screen. No tabs. No context switching. Just flow.
app.sunesis.ai/studio
Welcome to Sunesis, I can see, hear, and talk back in real time.
Narrate a storyRecord an adGuide a meditation
Sunesis Voice v2Aria · Warm
Generate Generate a character. Talk to it live.
A digital character that sees, hears, and talks back.
Build a character in seconds, choose its face, its look, and its voice, then start a real video call right in your browser. It listens with live speech recognition, replies out loud with natural text-to-speech, and holds the whole conversation in a running chat thread. No clips, no pre-renders. Press the mic and say hello.
- Hears you in real time, replies out loud, no server round-trip
- Full chat thread and live captions as they speak
- Turn on your camera for a face-to-face call, right in the browser
1 · Face
2 · Look
3 · VoiceAria's voice · Warm · Female
Photorealistic · Voice · VideoReady
Aria · Photorealistic
Call Aria, live, right in your browser
How it connects
Build your first character One character. A complete interaction layer.
Let people talk to it
Open a low-latency voice or video call that listens, responds, and takes turns naturally.
Give it work to do
Connect knowledge, memory, and tools so every conversation can lead to an outcome.
Put it where it matters
Ship on the web, in your product, or over the phone from one real-time API.
Voice, video, conversation, and action work from the same character definition.SUNESIS / REAL-TIME STACK
Sunesis Voice
Make every interaction feel considered.
Explore voice, live interaction, and developer tools when this section comes into view.
Flexible ways to build
Start with the moment you want to improve.
Choose the path that fits your team today. Add characters, agents, and developer tooling later without changing the foundation underneath.
{ }
S
</>
01For product teams
Turn one idea into a live interaction.
Move from a brief to a character, voice, or guided conversation without assembling five separate tools.
- Prototype a character in Studio
- Preview voice and video together
- Share a working flow before you ship
- 1 workspace
- to shape the experience
- 184ms
- streaming round-trip
How can I change the delivery address?
I found the order and kept the context for your team.
Memory attached · handoff ready
02For customer-facing teams
Give every conversation a useful next step.
Ground an agent in your knowledge, give it a memory, and let it hand off with the context your team needs.
- Voice, chat, and video in one agent
- Scoped tools and knowledge access
- Human handoff with a clean summary
- 32 languages
- for global conversations
- 24/7
- without losing the human tone
TSPYGOWS
03For developers
One foundation for every modality.
Use the same credentials, events, usage controls, and SDK conventions across speech, characters, live rooms, and agents.
- TypeScript, Python, Go, REST, and WebSocket
- Stream speech and video as the user waits
- Add webhooks, memory, and tools when ready
- 1 API key
- across every surface
- Sub-200ms
- for real-time turns
One workspace · shared credentials · usage you can see.
Developers
Voice, video, and conversation APIs built for production.
Fast to start. Open by design.
Build speech, cloning, avatars, and live transcription into one API workflow. TypeScript, Python, and Go SDKs are ready out of the box, with sub-200ms response times.
- 184ms
- first audio
- 32
- languages
- 99.99%
- uptime
Quickstart
API onlineYour first response in five minutes.
import { SunesisClient } from "@sunesis/sunesis-js"; const sunesis = new SunesisClient(); // Stream lifelike speech, token by token const audio = await sunesis.textToSpeech.convert({ text: "Welcome to Sunesis.", voiceId: "aria", stream: true, });
Use cases
What teams ship on Sunesis.
Four production paths. One API key, one event model, and one usage view underneath them all.
Use case 01
Avatar video
Lip-syncable, emotion-aware TTS for AI avatar products. Your digital character speaks with natural prosody and visible emotion, video + voice fused.
SceneVoiceStatus
Welcome + product tourAria · WarmRendered
Feature walkthroughAria · WarmRendered
Pricing explainerAtlas · DeepQueued
Closing + next stepsAria · WarmDraft
Localised cut, FrenchÉtienne · ClearDraft
Use case 02
Voice agent
Sub-200ms turn-taking over WebSocket. Streaming TTS and ASR so your agent can listen, think, and respond without the awkward pause.
TurnLatencyOutcome
Caller states the problem168msUnderstood
Agent checks the order184msTool call
Caller interrupts142msBarge-in
Handed to a person176msEscalated
Summary written back159msLogged
Use case 03
Audio content & companions
Notes-to-audio, narration, and AI companions that actually talk back. Per-character pricing that scales with usage, not seats.
TrackVoiceLength
Chapter one, narrationNova · Bright12:04
Weekly notes digestSage · Calm04:38
Companion, evening check-inLuna · Soft02:11
Ad read, 30 second cutAtlas · Deep00:30
Meditation, sleep setLuna · Soft18:52
Use case 04
Conversational apps
Voice cloning from 15 seconds of audio. Persistent memory, function calling, and RAG, build agents that remember and act.
SessionToolsMemory
Onboarding assistant3 scopedPersistent
Support triage5 scopedPersistent
In-app tutor2 scopedSession
Account concierge4 scopedPersistent
Renewal outreach3 scopedPersistent
Character, voice & live conversation
Make every interaction feel like it has a point of view.
Create the character and voice together, then deliver a response quickly enough to keep the moment moving.
You define
Start with someone worth talking to.
- Role & point of view
- What it is there to do, and the opinion it holds while doing it.
- Voice direction
- Pace, warmth, and how it should sound when the answer is no.
- Context & guardrails
- What it may act on, and the lines it will not cross.
Aria
Warm · Female · English
- Role
- Product guide
- Voice
- Warm, measured
- First response
- <200ms
- Languages
- 32
- Voice sample
- 15s
Everything above travels together. Change the voice and the character keeps its role; change the role and it keeps its voice.
4.6 MOS voice qualityMeasured naturalness, not a promise on a slide.
One foundation. Many ways to build.
Compose the interaction your product deserves.
Start with speech, a character, a live session, or an agent, and connect them when the experience calls for more. One workspace, one set of credentials, one usage view.
Expressive speech
Generate, stream, or tailor a voice with delivery controls that fit the moment.
DocsCharacter video
Turn a script or an audio track into a speaking character on camera.
DocsRealtime rooms
Run live voice and video exchanges with turn-by-turn events you can act on.
DocsAgent runtime
Grounded conversations with memory, tool access, and an optional on-screen presence.
Docs- 184ms
- to first audio, streaming
- 32
- languages live in production
- 4.6
- MOS measured naturalness
- 99.99%
- uptime across 14 regions
Typed SDKs
TypeScript, Python, Go, REST, and WebSocket examples that match production behavior.
Built to operate
Scoped API keys, audit-friendly workspace controls, and predictable error handling.
Human developer support
Talk to the engineers building the platform when you are designing at scale.
Talk to usThe connected platform
Build one experience. Add depth when you need it.
Start with a single capability, then bring voice, characters, agents, and developer tooling together on the same Sunesis foundation.
SpeechVideo
Surface 01
Sunesis Studio
A shared workspace for shaping voice, characters, media, and live interaction.
ExploreYour order shipped on Tuesday.
Can I change the address?
MemoryHandoff
Surface 02
Sunesis Agents
Conversational roles with memory, grounded knowledge, and carefully scoped actions.
Explore200 OK · 184ms
Surface 03
Sunesis API
One developer foundation for generation, real-time sessions, webhooks, and typed SDKs.
Explore
WarmAria's voice
Surface 04
Sunesis Characters
Real-time personalities with visual presence, a signature voice, and agent intelligence.
ExploreReal people
Made for the humans on both sides.
Creators
Clone your voice from one short take, then narrate anything — and keep that same voice across every language you publish in.
Live conversations
Put a face on the conversation. A visible character that listens, answers, and holds the thread in real time.
Support & sales
One agent across phone, chat, and video — listening properly, and handing over with full context when a person should take it.
- Sub-200ms
- latency across video and voice
- 32
- languages one voice can carry
- 4.6 MOS
- measured, not promised on a slide
Digital Characters
Real faces. Real conversations.
Click any face to hear them speak.
240+ voices, infinite possibilities, voices uploaded and built by the Sunesis community.
The difference
Not another laggy AI platform.
| Capability | The others | Sunesis Labs |
|---|---|---|
| Latency | 500ms+ latency | 180ms latency |
| Voice quality | Robotic voices | MOS 4.6 naturalness |
| Memory | No memory | Persistent memory |
| Pricing / lock-in | Vendor lock-in | Open API, no lock-in |
| Regions / uptime | Single-region | 14 regions · 99.99% uptime |
Trusted by teams shipping voice and video in production
- Northwind
- Koru Health
- Helio
- Northwind
- Koru Health
- Helio
- Northwind
- Koru Health
- Helio
- Northwind
- Koru Health
- Helio
Customer outcomes
Powering voice and video for 2,400+ teams.
From new products finding their voice to public companies redesigning a customer journey, teams use Sunesis to make the interaction count.
Voice agents
“We replaced our entire IVR with Sunesis agents. Call deflection jumped overnight and our support costs dropped by 64%.”
Mara Lindqvist
VP Engineering, Northwind
Digital characters
“Onboarding completion went from 41% to 89% once we added a face-to-face digital character to the flow.”
Dev Patel
Head of Product, Koru Health
Real-time API
“We ran 40k concurrent voice sessions during launch week without a single dropped call. The streaming latency is unreal.”
Sofia Marchetti
Staff Engineer, Helio
In the wild
The product is part of the conversation.
Creators and developers share the moments that make a voice experience feel different.
“Cloned my voice in 15 seconds and it nailed my cadence on the first try. Wild.”
@creatorkai@YouTube
“The streaming TTS feels instant, my viewers think I'm live when I'm not.”
@novastreams@TikTok
“Built a customer support agent in an afternoon. It's better than our offshore team.”
@indiehacker@Twitter
“Finally a voice API that doesn't sound like a robot from 2015.”
@devjules@YouTube
Plans that grow with you
Explore pricing Start building now. Choose a plan when it matters.
Explore credits, production features, and team options on the pricing page, without interrupting the product story.
The real-time backbone
The details that make an interaction feel instant.
Every layer is measured for the moment between a person speaking and the system responding.
- 2.4k
- teams building on Sunesis
- 180ms
- median latency, voice generation
- 32
- languages for live conversations
- 4.6
- MOS, human-level naturalness
One turn, end to end14 regions online
- 01
Speech in
Live transcription as they talk
- 02
Understand
Grounded reasoning and tools
- 03
Voice out
Streaming speech, first token
- 04
On screen
Character video, lip synced
Voice in to response out
180msmedian, measured in production
Streaming voice · character video · grounded agents
Explore Sunesis
Find the right part of Sunesis.
Start with the product, plan, or team information that matters to you.
FAQ
Questions, answered.
Sunesis Voice v2 speaks 32 languages out of the box with any of our 240+ voices. The API supports 30+ languages including English, Spanish, Japanese, French, German, Mandarin, and more, all with native-quality prosody and the ability to clone a voice and use it across languages.