What does MOHOLO need to create an avatar?

Nothing from you: pick an avatar from the library and paste one tag. Each avatar is a rigged, photoreal 3D head with its own way of talking.

Is the avatar a video?

No. It is a fully three-dimensional head rendered about sixty times a second by the viewer's own device via WebGL. The server streams only motion data — a few kilobytes per second.

Why is it cheaper than video avatars?

Because there is no rendered video to stream: the viewer's device draws the avatar, so serving costs are roughly a thousand times lower than piping GPU-rendered video per viewer.

Compact photoreal avatars — in browser & in app

A face for your AI.

zoom 1.00 pan 0.000, 0.000 box 74% · 3:4 drag to pan · scroll or pinch to zoom · arrows nudge

Change what she does

Give her a role in one or two sentences: who she is for you, what she knows, how she should behave. She restarts the conversation with it.

Attach a document

Compact, realistic avatars that run in the browser or in your app — so they cost you nothing to serve. Don't pay outrageous per-minute fees for a live avatar when you can install a free, configurable one. Every avatar here is a synthetic face from our library, and what you're seeing is not video: they are fully three-dimensional heads, drawn sixty times a second by your device, compacted with our proprietary geometry, appearance and rig compaction so they download in seconds. No video stream. No render farm. No GPU bill.

Input — pick a face from the library, or generate your own avatar from a photo Runs — in your browser or your app Size — 2–10 MB (depending on quality level) Cost per viewer — zero

Put her on your site

One tag. Any page.

Paste this where she should appear. She listens, answers out loud and renders live in your visitor's browser. The prompt is hers to keep: tell her who she is for you. Free during the beta with the key free.

All parameters and a live example →

From the studio to the library.

Plate I is the entire input

Plate I — Source The capture session an avatar starts from.

AI processing

Plate II — The head, in 3D The result, replaying the exact motion recovered from Plate I — in sync. Grab and spin her while she moves.

No. 01

Choose →

Pick a face from the library. Every avatar is a rigged, photoreal 3D head with its own way of talking, compacted with our proprietary geometry, appearance and rig compaction: full detail, a few megabytes.

No. 02

Drive →

Voice in, expression out. A personalized audio-to-face model produces her lips, blinks and head motion live, from the conversation itself.

No. 03

Render anywhere →

In every modern browser, or embedded in your own app or game engine. No plug-in, no video stream, no per-minute GPU bill — the avatar lives where it is watched. Pick a detail level per device, or leave it automatic.

The beta is small,
the faces are yours.

We are onboarding a handful of products at a time. Leave an address and we will write when your key is ready.

How live avatars compare.

Every other service streams video from a data-center GPU

Capability	Moholo	Anam	Tavus	HeyGen	Synthesia	D-ID	LemonSlice	LiveAvatar	AKOOL	Colossyan
Rendering	Real-time 3D head, no video stream	Real-time	Hybrid	Real-time	Hybrid	Real-time	Public latency claim	Not comparable for live	Not applicable for live	Not applicable for production
Where the pixels are made	Your visitor’s browser or app	Cloud video	Cloud video	Cloud video	Cloud video	Cloud video	Cloud video	Cloud video	Cloud video	Cloud video
Latency	0.55 s text → first word; ~1.5 s speech → first word (measured)	Sub-1s conversation; 180ms server	~500ms / sub-600ms public claims	Low-latency claim; no comparable round-trip figure	471ms p99 inference TTFB	Low-latency claim; no numeric benchmark	—	—	—	—
Languages	24+ hosted; any with your own model	70+	42+	100+	30+	N/A	155+	175+	140+	120+
SDK / API surface	One iframe (WordPress), one script tag, JS API, bring-your-own model	JavaScript SDK, Python SDK, API, embed	HTTP API, React CVI library, Daily/WebRTC rooms	Agents SDK, Agents API/Streams, legacy Talks/Clips Streams	API, hosted avatars	API, embed, Web SDK	REST/OpenAPI, WebSocket	REST API, CLI/MCP, Video Agent, TTS	REST video and templates, REST video generation, Custom avatars, Custom LLM	—
Knowledge / context	Prompt up to 8,000 chars; your own model brings its RAG	Partial / BYO LLM integration	n/a	Knowledge base / RAG	BYO RAG / runtime context	Managed RAG + visual context	Managed agent RAG	Managed knowledge base	Prompt/context + reference URLs	Managed knowledge base
Security	TLS; model keys server-side only; nothing leaves your site in bring-your-own mode; SOC 2 not yet	SOC 2 Type II, HIPAA, GDPR/DPA, ZDR	SOC 2 Type II, HIPAA, GDPR	SOC 2, ISO 27001/27017/27018/42001, GDPR controls	ZDR option on Enterprise	n/a	SOC 2 Type II, ISO 27001/42001, GDPR	SOC 2 Type II, GDPR, CCPA, DPF, EU AI Act	SOC 2 Type II, ISO 27001/42001, GDPR controls	SOC 2 Type II, GDPR
Pricing model	Trial free; $10 / month per character	Subscription + included minutes + usage overage	Subscription + credits/minutes	Subscription + credits + usage overage	Subscription + credits	Subscription + credits/session precharge	API pay-as-you-go	Subscription + credits/minutes	Subscription + video minutes	—
Price per minute	$0 with your own model (Gemini Live ≈ 1–2 ¢, billed to you); hosted on quote	$0.11–0.24	$0.32–0.37	≈ $0.10–0.20	—	≈ $0.35	≈ $0.16–0.21	—	—	—
Best fit	Any website or app; unlimited minutes at a flat fee; nothing to host	Embed live avatar agents, AI humans with memory	Real-time agents and video	Expressive full-body characters	Quick embeds and custom UI	Teams using RTC providers	Programmatic and batch video	Training and L&D video	SCORM and training video	—

Columns for other vendors as published on anam.ai/compare (17 August 2026) and their pricing pages (September 2026); Moholo figures measured on 13 September 2026. Corrections: founders@moholo.co.