TalkingHead
met4citizen/TalkingHeadReviewed by AI
Checked by RepoPicks Radar on
What is TalkingHead?
TalkingHead is a JavaScript class that puts a full-body 3D character on a web page and makes it speak in real time, with lips, face and gestures following the words. It reads text aloud through a text-to-speech service (Google Cloud by default, or ElevenLabs, Azure and others) and has built-in lip-sync for English, German, French, Brazilian Portuguese, Finnish, Lithuanian and Bengali. It is meant for people who write web code: you add it from NPM or a CDN and supply your own avatar file and service keys. A test page runs online.
How easy is it to install TalkingHead?
B · installer plus one simple step · Web
On the web there is nothing to install: the test app runs at met4citizen.github.io/TalkingHead and the class loads from a CDN, but the README states that speech and AI "will not work" until you paste your own API key in the settings menu, which is the one extra step. Using it inside your own site is developer work: you also need a 3D avatar file with a Mixamo rig and ARKit and Oculus blend shapes.
TalkingHead on GitHub
- Stars
- 1,669on Nov 10, 2026
- Downloads
- —all releases
- Growth
- +7stars in 6 days
Stars over time
Facts
- Runs on
- Web
- You need to install
- Node.js
- To build from source
- Node.js
- Disk space
- 100 MB–500 MB · estimate
- Accounts or keys
- Google Cloud Text-to-Speech, ElevenLabs, Microsoft Azure Speech, OpenAI, Google Gemini, Grok (x.ai)
- Cost
- Free, API costs extra · The code is MIT and free. Making the avatar talk normally needs a text-to-speech account: the README recommends Google Cloud TTS (free up to 4 million characters a month, paid above that) or Microsoft Azure / ElevenLabs. The author's separate HeadTTS module can run voices in the browser for free instead.
- License
- MIT (Permissive)
- Last change on GitHub
- Nov 4, 2026
TalkingHead risk signals (4)
- Attentionapi key in browser code
The beginner path tells you to "Simply download the file, add your Google TTS API key", while the options table warns about ttsApikey: "NOTE: I recommend that you don't use this in production and never put your API key in any client-side code." A key placed that way can be read by anyone who opens the page.
AI review
- Infoexample avatars non commercial only
README: "brunette.glb" is under CC BY-NC 4.0, and "avatar.glb" (Avaturn), "avatarsdk.glb" and "vroid.glb" were created "for non-commercial use". The project itself is MIT, but the bundled sample avatars, animations and sounds are not all free for commercial projects.
AI review
- Infotest app sends data to external ai services
index.html defines proxies for OpenAI chat, moderations and audio transcriptions, Vertex AI, Google TTS, ElevenLabs, Microsoft TTS and Grok. If you enable those features, your text and microphone audio go to those companies.
AI review
- Infoauthor calls it a side project
FAQ: "This is just a small side-project for me, so I don't have any big plans for it." The repository is active (last commit 3 days before the check, v1.8.2) but is maintained by one person.
AI review
Evidence is shown in its original language.
What Radar could not check about TalkingHead
- How well the lip-sync looks and sounds in each supported language
- The real monthly cost once you pass Google's free text-to-speech allowance
- Whether a 3D avatar you create elsewhere really has the required Mixamo rig and the 52 ARKit plus 15 Oculus blend shapes
- How it performs on older phones and tablets, since 3D rendering runs in the browser
- What each external AI or speech service keeps of the text and audio you send