Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.bithuman.ai/llms.txt

Use this file to discover all available pages before exploring further.

A Homebrew-installable command-line app that lets you talk to your Mac in three modes: text, voice, or video. Useful for:
  • Trying bitHuman in 30 seconds without writing any code.
  • Testing your hardware before integrating the SDK.
  • Demoing the technology to colleagues or stakeholders.
  • Learning the SDK — the CLI is itself a working example. See the annotated reference apps at bithuman-product/bithuman-apps for full Mac, iPad, and iPhone consumer code.

Install

brew tap bithuman-product/bithuman
brew install bithuman-cli
Requires macOS 26 (Tahoe) or later, Apple Silicon (M3+), ~3 GB free disk for voice mode or ~7 GB for video mode (LLM + TTS + expression engine weights cache locally on first run). If you previously installed the legacy bitchat formula:
brew uninstall bitchat
brew untap bithuman-product/bitchat
…then run the install commands above.

Three modes

ModeCommandWhat you getFirst-run download
Textbithuman-cli textTyped REPL in the terminal. Pipe-friendly.~2 GB
Voice (default)bithuman-cliSpoken conversation through your Mac’s mic + speaker. Voice cloning supported.~3 GB
Videobithuman-cli videoVoice + a small floating animated face that lip-syncs at 25 FPS. 8 bundled agents. Drag-drop face swap.~4 GB
# Quickest possible test
bithuman-cli text
> hello
Hi there. What's on your mind today?

# Voice mode (the default)
bithuman-cli
🎙  Listening. Talk any time.

# Video mode with the default agent
bithuman-cli video

# Video mode with a custom portrait
bithuman-cli video --image ~/Pictures/me.jpg

# Voice mode with a cloned voice (10–20 s of clean reference audio)
bithuman-cli --voice ~/myvoice.wav

# Custom system prompt
bithuman-cli --prompt "You are a calm tutor explaining ideas in plain words."

Flags

FlagWhat it doesModes
--locale <bcp47>ASR locale + TTS language hint (default: en-US).all
--voice <preset|path>Voice preset name (case-insensitive) or path to a 10–20 s mono audio file. Path = the voice is cloned.voice + video
--image <preset|path>(video only) Bundled portrait preset (Alice, Marco, Captain, Nia, Riley) or path to a JPG/PNG/HEIC. Face is auto-cropped via Vision.video
--prompt <text|@path>Override the LLM system prompt. Inline text or @/path/to/file.txt.all
-h, --helpShow full help.all

Bundled agents (video mode)

Each agent is a face + voice + system prompt bundled together. The CLI uses Diego by default; right-click the floating window to swap.
AgentCategoryVoice
Diego (default)Coacham_michael
NovaEntertainmentaf_alloy
EinsteinMentoram_adam
RiyaCoachaf_heart
LenaCoachaf_kore
RaeEntertainmentam_echo
Dr. MayaBusinessbf_emma
MasonBusinessbm_george

Voice cloning (voice mode)

Bring your own reference clip:
bithuman-cli --voice ~/myvoice.wav
Requirements:
  • 10–20 seconds of clean mono speech, any sample rate.
  • A transcript: either supply it as a sibling .txt file (~/myvoice.txt) or let the CLI auto-transcribe via Apple’s on-device Speech framework (cached after first run).
Cloning is voice-mode only. Video mode uses preset voices (Kokoro) for performance reasons — the GPU is busy driving the avatar’s lip-sync.

What it looks like

In video mode you’ll see a small floating circular window in the top-right of your screen. Right-click for the menu:
  • Choose agent… — gallery of the 8 bundled agents
  • Customize → Change image… — swap the portrait
  • Customize → Change voice… — voice gallery (Kokoro presets)
  • Customize → Change prompt… — system prompt editor with 6 starter templates
  • Quit bitHuman
Drag any image file from Finder onto the window for an instant face swap. Press the spacebar in the terminal to interrupt the bot mid-sentence.

Why ship a CLI?

The CLI exists for three reasons:
  1. The fastest possible “see what this is”. No SDK integration, no Xcode, no signing. Just a brew install.
  2. A working reference for the SDK. Every bitHumanKit feature (voice cloning, agents, drag-drop face swap, video mode) is exercised by the CLI. The source is the canonical integration example.
  3. A diagnostic tool for hardware compatibility. If bithuman-cli video runs cleanly on a developer’s machine, their app will too.

Distribution

Distributed via the public Homebrew tap at bithuman-product/homebrew-bithuman. Releases are notarised by Apple and stapled before tap upload. Manual DMG download (when you don’t want Homebrew): each release on the tap repo also includes a notarised zip artifact.

Troubleshooting

The CLI inherits all the SDK’s behaviour, so the SDK Troubleshooting page covers most issues. The CLI-specific gotchas:
  • brew install 404 after the bitchat → bithuman rename: brew untap bithuman-product/bitchat first, then re-tap.
  • First-launch hang for ~2 minutes downloading models. Add BITHUMAN_VERBOSE=1 to see progress.
  • Audio device refusal on external USB / Bluetooth mics. Test with the built-in mic first; some peripherals report unusual channel layouts the SDK’s voice-processing IO doesn’t accept.

Next