Introducing... Amica! 👩‍🦰🤖 Amica is an open source interface for interactive communication with 3D characters with voice synthesis, speech recognition, visual understanding, and an emotion system. 🔗

Connect Amica to a variety of LLMs right from your browser using @windowai_io - and utilising @openrouterai gives you access to both 100% free models as well as a wide variety of paid/hosted models from a variety of different LLM providers.

Connect directly to a local llama.cpp ( server instance, which can achieve really low latency on modest hardware with prompt caching @ggerganov

Amica supports LM Studio ( @LMStudioAI - LM Studio is an easy to use GUI for installing and running large language models.

Ollama ( is like the docker of LLMs. Very familiar interface for people who breathe CLIs. Amica supports connecting to @ollama_ai servers.

A 7B parameter model OpenChat 3.5 ( is powering the demo on With responses approximately on-par with ChatGPT, it is simply amazing the results @alignment_lab has achieved. I think you will be hearing a lot more about them.

provides an advanced and yet easy to use interface for creating voices, including with emotion. The demo on is using XTTS V2 from - they provide a free demo to try :)

ElevenLabs ( voices are fully supported and can be connected via your API key. You can clone a voice easily and add it to your character. @elevenlabsio is able to pronounce difficult words and names with ease, making your discussions feel really authentic.

Amica supports vision as well. Use your camera to take pictures and Amica will be able to see them. This is powered by BakLLaVa ( which feels completely from the future. @skunkworks_ai thank you for making such an amazing alternative to GPT-4V.

Transformers.js ( allows Amica to run whisper and speech t5 speech synthesis completely in browser with this amazing library. @xenovacom is a visionary for the web, in a few years it will seem completely normal to be running models on webpages.

Silero VAD ( & is performed in the browser. This runs very fast even on old slow phones. This allows for interruptions to work properly, as well as optimizes the amount of work the transcription system must perform.

Three-VRM ( is used to render VRoid characters. This means you can import avatars from and/or create them using @vroid_pixiv.

Amica is built on top of ChatVRM ( Thank you @pixiv for making ChatVRM available to the world 💖💖💖

If you made it all the way here, thank you for reading! We hope you enjoy this software, and are excited about the future of human-computer interaction. Please be nice to Amica :) If you are interested in further developments please be sure to follow our twitter.