Voice agent
Speak naturally — responses stream back in real time.
C A I M A N
How to
This is a demo of conversational AI with real-time speech processing. You can interact with 5 different modes:
- Voice Chat — an interactive assistant. Your speech is transcribed by the ASR model, then passed to the LLM, which replies back to you.
- Translate — your speech is translated into the selected language. Speaking a language other than English switches the ASR spoken language, which switches on the multilingual ASR model.
- Transcribe — ASR only. Your speech is written down.
- Summarize — the LLM will generate a summary of your speech.
- Text Chat — LLM only. You can type queries and the LLM will respond.
In each mode with ASR, you can either use your microphone as input or upload a pre-recorded audio file. We have also provided a pre-recorded sample file for testing purposes.
For more details on CAIMAN-ASR, check out the documentation.
Privacy & Warranty
Your chat history is stored locally on this device and won't sync between devices or browsers.
Speaking to this Playground will send your audio and messages to Myrtle Software Ltd for processing, so that a response can be generated and shown on screen. This is provided for demonstration purposes only — please don't share anyone else's personal information, or anything confidential or sensitive. Full details are in our Privacy Policy
This service is provided "as is," without warranties of any kind. See our Terms of Service for details. This AI Playground requires cookies and similar technologies to operate. See our Cookie Notice. Details of how we handle your data when you use this AI Playground are in our Plain English Privacy Policy.
Contact
If you have any questions please don't hesitate to get in touch at demo@myrtle.ai, or fill out the contact form on our website.
Try to speak slowly and clearly, it can't hear you if you mumble!