Skip to content
bitHuman
Join DiscordSign up

AI avatars in our cloud, on your servers, on the device or fully offline

The same avatar runs in each mode. What changes is where it renders, where the conversation runs, and what reaches bitHuman.
Contact salesBuild an app

Side by side

  • bitHuman cloud

    Avatar renders
    On bitHuman's servers
    Conversation runs
    bitHuman's voice service, or your own provider keys
    Internet
    Required for the whole session
    Best for
    The fastest start: web pages, apps and APIs

  • Your servers

    Avatar renders
    On your own Mac or Linux machines
    Conversation runs
    The CLI's local brain, your services, or bitHuman's
    Internet
    To start a session
    Best for
    On-premises, data-center and private-cloud deployments

  • On the device

    Avatar renders
    On the user's device, or a WebGPU browser tab
    Conversation runs
    Your app's choice; with the web embed, on bitHuman's servers
    Internet
    To start a session
    Best for
    Apps, kiosks and embedded screens

  • Fully offline

    Avatar renders
    On your Linux PCs and terminals
    Conversation runs
    Agreed with sales for your site
    Internet
    Off the internet once set up
    Best for
    Kiosks, ATMs and trade shows; Business and Enterprise

Creating an avatar from a portrait always happens in the bitHuman cloud. Once created, its model file downloads to your machines and runs there.

What reaches bitHuman in each mode

Available for Business & Enterprise: fully offline · Upgrade to Business

AI avatars in the bitHuman cloud

The fastest start for web pages, apps and APIs. The avatar renders on bitHuman's servers.

How it runs

What reaches bitHuman
The session's audio and conversation, to run it. Conversations with an agent in the bitHuman cloud are kept as transcripts until the agent is deleted, unless conversation logging is switched off for that agent. Today bitHuman switches logging off on request; there is no dashboard switch or API field for it. Agent owners can view their transcripts in the dashboard, export them as JSON and read them with the API. They can't delete a single transcript; deleting the agent deletes its transcripts.
Models
Essence 2, Expression 2 and the first-generation models

Contact salesRead the cloud guideCompare all four

AI avatars that render on the device

On the iPhone, iPad, Mac or Android device in front of the user, or in a WebGPU browser tab.

How it runs

Where it renders
On the iPhone, iPad, Mac or Android device in front of the user, or in a WebGPU browser tab
What reaches bitHuman
With your own voice and language services, usage metering only — never audio, video or conversation text
Internet
To start a session; rendering continues through a network drop of up to 5 minutes
Models
Essence 2 and Expression 2

With the web embed, the conversation runs on bitHuman's servers, even when the avatar renders in the tab.

Contact salesRead the on-device guideCompare all four

AI avatars on your own servers

On your own Mac or Linux machines, on-premises or in your cloud account. A standard Linux PC needs no GPU.

How it runs

Where it renders
On your own Mac or Linux machines, on-premises or in your cloud account; a standard Linux PC needs no GPU
Where the conversation runs
Your choice: the CLI's local conversation brain, your own speech and language services, or bitHuman's
What reaches bitHuman
A credential check when a session starts, the avatar download, and usage reports with no audio, video or conversation text
Internet
To start a session; rendering continues through a network drop of up to 5 minutes
Models
Essence 2 and Expression 2; Essence 1 on the CLI and Python

Usage reports contain no audio, video, images or conversation text. When you run the engine on your servers or device with your own agent, conversation text never reaches bitHuman; usage reports carry only session id, durations, product and install id. Hosted bitHuman agents store transcripts wherever the avatar renders, unless conversation logging is turned off.

Contact salesRead the self-hosting guideCompare all four

Run AI avatars completely off the internet

Pay once per character. No per-minute charges.

Offline license is only available to Business and Enterprise clients who want to run avatars completely locally, off the internet — e.g. kiosks, trade shows, ATM machines, embedded screens.

Available for Business & Enterprise

Contact sales

Linux PCs and terminals

Linux x86_64.

$899 per character, paid once

Packs of 5 and 25 on Business, 100 on Enterprise. Works while your Business or Enterprise plan is active.

Connects about once a year

Offline devices connect before the license file runs out: about once a year on an annual plan, at least every 30 days on a monthly plan.

What runs offline

  • Essence 1

    Photoreal people, pre-built avatars

    Offline
    Comes later · Linux x86_64

  • Essence 2

    A photoreal person from one portrait

    Offline
    Comes later

  • Expression 2

    Any character from one portrait

    Offline
    Now · Linux x86_64

An offline license runs characters fully offline on Linux x86_64: Expression 2 now; Essence 1 and Essence 2 come later.

Creating the avatar from a portrait happens in the bitHuman cloud; the finished avatar model then runs on your machines.

Linux PCs and terminals (x86_64). Phones, Macs and browsers stay online: the Swift package, the Android SDK and the web embed check your credential when a session starts, and an on-device session keeps rendering through a network drop of up to 5 minutes.

Business & Enterprise

Linux PCs and terminals; arranged through sales. Your licenses work while your Business or Enterprise plan is active. If the plan ends, they keep working for 30 days, then the characters stop loading until a plan is active again.

1 character $899 · 5 for $2,990 · 25 for $9,900 (Business) · 100 for $24,900 (Enterprise). Unlimited: by quote.

Prices and plans

Bring your avatars in-house

Compare all four Security & privacy

Tell us where it needs to run

Contact salesBuild an app