CompNav · Open-source projects
Amica
Talk with customizable 3D AI characters by voice or text; characters can express emotions and use visual input.
What it is
Amica centers on an interface for interactive 3D characters. Users can import VRM avatars, adjust voices, and chat by voice or text. Characters can express emotions and work with visual input; the chat model, speech recognition, and speech synthesis come from separate services or local backends. It is better understood as a configurable virtual-character workspace than a companion app with one fixed persona.
Core mechanisms and capabilities
Import a VRM 3D character, adjust its voice, and converse by text or natural speech; AI responses can include emotional expressions.
Visual input, speech recognition, and speech synthesis can be connected to the character experience, alongside hosted APIs and several local model services.
Try the experience in a web or mobile browser, or use the Tauri-based desktop application to bring the character into a local app.
Editorial take
Amica merits inclusion because it brings 3D avatars, voice interaction, and multiple model backends into an interface that can be tried online or run locally. It suits people who want an AI character with a visible presence and voice. Most code is MIT-licensed, but models, images, and other assets retain their authors' separate terms. Users also need to connect models and speech services, and the project notes that some setup documentation is still being expanded.
Who it suits
Suited to people who care about character appearance, voice interaction, and avatar customization and are willing to configure models, speech services, or VRM assets.
Deployment and technical requirements
Amica offers a browser experience and source code that can be run locally; its desktop app is packaged with Tauri. For more control, configure a hosted model API or local service. The web trial and local deployment are separate routes.
Runtime requirements
The official web experience lets you preview the character interface. Running locally requires installing the project and configuring `.env.local` as described in the README. Hosted models require an API key; local models or speech tools require their own services and model files.
A local deployment needs a chat model and any desired speech services configured. The README lists hosted connections such as OpenAI and OpenRouter, plus local or compatible backends including Ollama, LM Studio, llama.cpp, and KoboldCpp. Hosted services require API keys and may charge; local inference requires model downloads and suitable compute resources.
Limitations and considerations
Amica does not include a hosted model service. Chat, voice, and vision capabilities depend on the selected backends, and some hosted features require credentials or a paid account. Most code uses the MIT license, while 3D models, images, and other assets follow their authors' separate terms.
Official sources and further reading
Information checked: Oct 26, 2026.