Choose the brain
Run our hosted model, or point the engine at your own LLM.
Language model
- Thrifty hosted Default
- OpenAI
- Claude
- +Your own LLM endpoint
PRESENCE ENGINE
A face is not enough. The Presence Engine is the stack under every Thrifty Artificial Human: it listens, understands, remembers and answers in real time, so the person on the screen feels like they are actually there.
Built with world-class AI, backed by leading startup ecosystems
BENCHMARKS
0.90 s
From first audio to first frame on Maya-2, our live human model.
42+
Languages an Artificial Human can speak, with lip-sync that follows.
99%+
Identity consistency on streams longer than five minutes. No cuts, no resets.
THE PRESENCE THESIS
Chatbots answer. Presence responds: it hears how you said it, remembers who you are, checks what it knows, and shows its work while it talks. The Presence Engine runs that whole loop on every turn, in one continuous stream.
WHAT WE MEASURE
Every row is measured on live streams from our own models, not on short demo clips. Where a bar has a line across it, that line is the bar you have to clear to feel real.
Start-up speed
Time from the first word of audio to the first frame on screen
0.90 s
under 1 s
Real-time throughput
Smooth full-body video, in portrait or landscape
32 FPS
live needs 25
Identity consistency
Your face stays the same, from minute one to minute sixty
99%+
5+ min streams
Motion smoothness
Gestures, blinks and head turns that move naturally
99%+
5+ min streams
Background consistency
The backdrop holds steady while the person moves
98.4%
5+ min streams
Lip-sync
Lips that match every word, in every language
#1
of models tested
OUR OWN MODEL
Live video needs 25 frames every second, without exception. Most full-body models are nowhere near. Maya-2, the face inside the Presence Engine, clears it with room to spare, and so does its single-GPU sibling Maya-Lite.
Full-body frames per second
Maya-2
Fastest other real-time model
Leading offline full-body models
Typical diffusion avatar models
FRAMES PER SECOND · REAL TIME = 25
ONE ENGINE
The engine is identity-agnostic. Build a photoreal colleague from one photo, a brand mascot, or an illustrated storyteller, and the timing, memory and expression come with it.
ORCHESTRATION
The Presence Engine is the conductor, not the only instrument. Bring the language model, voice and knowledge you already trust. The rhythm, memory and rendering stay ours.
Run our hosted model, or point the engine at your own LLM.
Language model
Stock voices, a clone from one minute of video, or your voice provider.
Voice
42+ languages
Files and websites become answers it can cite, tuned for speed or depth.
Knowledge base
RETRIEVAL
Show visuals, present decks and join meetings, all mid-conversation.
Skills
ENTERPRISE-GRADE
The more present an Artificial Human feels, the more it matters that people know what it is and where their data goes. Both are built into the engine, not bolted on.
Every Artificial Human says what it is. Real people's likenesses are only used with their documented consent.
Set the lines it never crosses, decide what it remembers, and give users a way to opt out.
Our cloud, your VPC, fully on-prem or air-gapped, with a zero data retention mode when it matters.
The Presence Engine is the runtime behind every Thrifty AI Artificial Human. It handles listening, turn-taking, memory, knowledge, skills and live rendering together, so a conversation with an Artificial Human feels natural instead of like a chatbot with a video attached.
Thrifty Studios is where you create and customise an Artificial Human: its face, voice, personality and where it is deployed. The Presence Engine is what runs underneath once it is live. Every Artificial Human built in Studios runs on it.
The face is rendered by Maya, Thrifty AI's in-house human model family: Maya-2 for full-body video and Maya-Lite for single-GPU deployments. The reasoning model is your choice. Use our hosted model or bring your own LLM.
Yes. Give an Artificial Human files such as PDFs, Word documents, slides, spreadsheets and images, or a public website to crawl, and it retrieves from them during the conversation. You can favour speed or depth depending on the experience you are building.
An Artificial Human always discloses that it is an AI and only uses a real person's likeness with their documented consent. You set its guardrails, decide what it remembers, and can run it in our cloud, your VPC or fully on-prem, with a zero data retention mode available.
On your website or app with one script tag, as a hosted page, inside Google Meet as a participant, and on physical devices such as kiosks, robots and holobox displays.