The Technology Powering Soulbits

An open-source AI engine ecosystem built with local-first setups in mind, backed by scalable cloud with maximum focus on privacy.

At a glanceClient SDKsGo · JS

SOULBITS SOFTWARE

A complete ecosystem - built for local execution

Soulbits Engine

Soulbits Engine is modular middleware that connects AI models into a unified, multi-modal character experience. Each module — LLM, TTS, STT, Countenance, Movement — can be used independently or combined. The real magic happens when they work together: the AI Harmonization Layer coordinates modules in real-time to produce a cohesive character that talks, moves, and emotes.

AI Harmonization Layer achieves true multi-modality between modulesSupport for a growing variety of existing AI technologies and providersUnified Event System with standardized APIs (OpenAPI and AsyncAPI)Extensible plugin system for easy integration into any application

View on GitHub Read the Docs

Supported Backends

LLM Module:

Kajiwoto.AICharacter AIKindroid AIText-Generation-WebUIOpenAI

TTS Module:

Harmony Speech EngineElevenlabs

STT Module:

Harmony Speech EngineOpenAI

Countenance Module:

Text-Generation-WebUI

Movement Module:

Text-Generation-WebUI

Available Plugins

VNGE Plugin: Visual Novel Game Engine integration for interactive character dialogue

VTube Studio Plugin: Real-time avatar control for VTubers with AI-driven expressions

ProjectP Plugin: Advanced motion capture and animation workflows

Download the Engine

Install the Soulbits Engine on your own computer — the same core that runs your companion in the Cloud, on your own hardware.

WindowsEXEComing Soon

LinuxAppImageComing Soon

macOSDMGComing Soon

Screenshot Placeholder — Soulbits AI Link UI

Soulbits App

The Harmony AI App brings the full companion experience to your pocket. Available as custom iOS IPA and Android APK builds, it works completely offline — your companion goes everywhere you go, with no cloud dependency required.

Full offline capability - no internet connection requiredCustom iOS (IPA) and Android (APK) buildsComplete companion experience: chat, voice, and personalitySyncs with Soulbits Engine sessions across devices

View on GitHub Read the Docs

Download the App

AndroidAPKComing Soon

iOSIPAComing Soon

Screenshot Placeholder — Soulbits AI App UI

Soulvoice Engine

Harmony Speech is a high-performance AI Speech Engine that enables faster-than-realtime speech generation, zero-shot voice cloning, and speech recognition using state-of-the-art open-source AI voice technology. It provides OpenAI-style APIs for seamless integration.

OpenAI-style APIs for Text-to-Speech, Speech-to-Text, Voice Conversion, and Speech EmbeddingMulti-model and parallel model processing supportToolchain-based request processing with internal re-routing between modelsExtensible architecture for adding future open-source AI speech models

View on GitHub Read the Docs

Supported Model Architectures

OpenVoice V1 OpenVoice V2 / MeloTTS OpenAI Whisper (FasterWhisper)Harmony Speech V1

Screenshot Placeholder — Speech Engine UI

ARCHITECTURE FLOW

How it works

User Input (Text or Voice) → Capture and Processing via Soulbits App

Soulbits Engine retrieves historical context, RAG backstories, and system instructions

Payload is routed to the configured AI backend (Local PC hardware via Soulbits AI Link or secure Cloud)

LLM generates response → Processed via Soulvoice Engine for emotive voice synthesis

Real-time streaming output (Audio, Text, and Live2D/3D animation triggers) delivered to the user interface

View the architecture diagram

Your input flows through the Soulbits App and Engine, which routes it to the AI backend you choose — local on your own hardware or fully managed in Soulbits Cloud — and streams the result back to you.

Supported Backends & Models

Total architectural freedom. Choose the model stack that fits your privacy goals and hardware capabilities.

Large Language Models (LLM)

Llama 3, Mistral, Gemma, Phi-3, and OpenAI-compatible API endpoints (Run locally via Ollama, LM Studio, or KoboldAI).

Text-To-Speech (TTS)

StyleTTS2, XTTSv2, AllTalk, ElevenLabs, and cloud-hosted neural providers.

Speech-To-Text (STT)

OpenAI Whisper (Local faster-whisper implementations) and browser-native web speech APIs.

Vision & Image Gen

Stable Diffusion (XL/v1.5) via Automatic1111/ComfyUI web APIs for dynamic companion expressions and image generation.

Download Soulbits

Get the companion on your phone, or run the full engine on your own computer.

Download for MobileAndroid and iOS

AndroidComing SooniOSComing Soon

Download for DesktopWindows, Linux and macOS

WindowsComing SoonLinuxComing SoonmacOSComing Soon

Building something with Soulbits? Guides and API references are on the way — in the meantime, the source code and a Discord invite are on the Developers page.