No description
- JavaScript 35.2%
- TypeScript 23.2%
- CSS 22.1%
- Python 11.2%
- Svelte 7%
- Other 1.3%
|
|
||
|---|---|---|
| .claude | ||
| .github/workflows | ||
| AI | ||
| public | ||
| scripts | ||
| src | ||
| tests | ||
| .gitignore | ||
| actionplan.md | ||
| ai-instruct.txt | ||
| BROWSER_TESTING_GUIDE.md | ||
| index.html | ||
| index.html.legacy | ||
| landing.js | ||
| MIGRATION.md | ||
| package-lock.json | ||
| package.json | ||
| playwright-server.js | ||
| playwright-server.mjs | ||
| playwright.config.mjs | ||
| README.md | ||
| requirements.txt | ||
| SPEECH_RECOGNITION_ANALYSIS.md | ||
| style.css | ||
| svelte.config.js | ||
| tsconfig.json | ||
| tsconfig.node.json | ||
| vite.config.ts | ||
DEPRICATED APPLICATION try instead: https://github.com/Unity-Lab-AI/Talking-with-Unity.git https://www.unityailab.com/Talking-with-Unity/
Talk to Unity in Plain English
Talk to Unity is a single web page that acts like a friendly concierge. The landing screen double-checks that your browser has everything it needs (secure connection, microphone, speech tools). Once every light turns green, a voice assistant named Unity wakes up so you can talk out loud and hear it answer back.
What you need
- A recent version of Chrome, Edge, or Safari. (Firefox still lacks the speech tools we use.)
- A secure address — either
https://on the web orhttp://localhostwhile testing. - Speakers or headphones so you can hear Unity talk.
- A microphone and permission to use it.
If any of these are missing, the landing page will highlight what to fix and share a short tip.
Fastest way to try it
- Download the project or clone it with Git.
git clone https://github.com/Unity-Lab-AI/Talk-to-Unity.git cd Talk-to-Unity - Start a basic web server. Python is an easy option:
python -m http.server 8000 - Open
http://localhost:8000in a supported browser and loadindex.html. - Watch the landing page run its checks. When everything shows Ready, press the button to enter the AI lab and start talking to Unity.
Understanding the landing page
- System lights: Secure context, speech recognition, speech playback, and microphone access each get a color. Green means you are good to go. Amber means you will see clear instructions to fix the problem.
- Status message: The box at the top always tells you what Unity is waiting on. When all lights are green it invites you into the lab.
- Help links: Quick links back to the Unity AI Lab website and this repository sit near the top so you can jump out for more info.
Using the voice assistant
- Enter the lab once the landing checks pass.
- Click Unmute microphone to let Unity listen.
- Speak naturally. Unity will transcribe, respond with text, and also read its answer aloud.
- Use the on-screen controls to mute, stop, or reset the conversation if you want to start over.
All of this runs in the browser — no extra servers or databases are required.
Customizing the experience
- Prompt and personality: Edit
ai-instruct.txtto change Unity’s default instructions. - Landing behavior: Adjust the readiness checks or UI copy inside
landing.js. - Voice interface: Modify
AI/app.jsfor how Unity listens, thinks, and speaks. - Styling: Tweak the look across the landing page and chat area in
style.cssandAI/style.css.
File map
| File | Purpose |
|---|---|
index.html |
Loads the landing page, pulls in the styles, and boots both scripts. |
landing.js |
Handles the pre-flight checks and controls the landing layout. |
style.css |
Shared styling for the landing screen. |
AI/app.js |
Runs the live voice conversation. |
AI/style.css |
Styles for the in-lab chat interface. |
ai-instruct.txt |
Text prompt that shapes how Unity responds. |
Need more help?
- Unity AI Lab: https://unityailab.online
- GitHub issues: https://github.com/Unity-Lab-AI/Talk-to-Unity/issues
Share feedback, swap in your own prompt, or dress up the visuals — just keep it simple and have fun chatting with Unity.