Android / Ollama-compatible chat / Voice

Dual AI Chat

Dual AI Chat is an Android chat client for user-configured Ollama-compatible servers. It combines single-model chat, two-AI back-and-forth chatter, voice input, TTS, avatars, settings profiles, optional web search, and debug logging in one app.

Docs: User Manual | Privacy Policy

        Use Normal mode for one-model chat or Chatter mode for alternating turns between Base and Chatter Partner.
Voice Input, Auto Voice Input, and Text-to-Speech make hands-free interaction possible.
Profiles, avatar images, search APIs, and Debug Mode are managed locally on the device.

      

Screenshots

Dual AI Chat main chat screen — Chat screen with avatar presentation and live message history.

Dual AI Chat conversation view — Conversation view used for both Normal and Chatter modes.

Dual AI Chat settings screen — Settings screen for server connection, models, voice options, avatars, profiles, and Web Search.

Key Features

Connects to your Ollama-compatible server: On startup and when settings are opened, the app calls /api/tags to refresh model lists and the connection-status tile.
Normal and Chatter modes: Base and Chatter Partner can each have their own model, name, system prompt, and speech settings.
Voice interaction: When Voice Input is enabled, pressing Send with an empty input field starts speech recognition. TTS and Auto Voice Input are also available.
Settings profiles: Save, load, and delete profiles, including the avatar images associated with them.
Web Search and Debug Mode: Add search results from custom or Brave-compatible APIs into the chat context and inspect API request/response logs.
Avatar system: Configure c0 (background), c1 (base), c2 (blink), and c3 (talking) slots separately for Base and Chatter Partner.

Usage Notes

You can use the built-in API from LLM tester with llama.cpp or any other Ollama-compatible server, whether local or remote.
Server URLs and search APIs are chosen by the user. If you use remote servers or third-party APIs, those providers process the request data under their own policies.
In the current implementation, conversation history is kept in memory during the active session and is not persisted across app restarts.