SillyTavern

SillyTavern is a free interface for interacting with language models that is installed on your own computer. It was created for roleplay chats with characters, but it is also suitable for general work with AI models. It connects to cloud-based models via API and to models running locally. The project is open source (AGPL-3.0 license): it started in February 2023 as a fork of TavernAI 1.2.8, and now has more than 300 contributors and around 35,000 stars on GitHub. The latest version, 1.19.0, was released on September 14, 2026.

Features

• Multiple models in one interface. OpenAI, Claude, Mistral, OpenRouter, NovelAI, AI Horde, as well as local KoboldAI/KoboldCpp, Ollama, Text Generation WebUI (Ooba), and TabbyAPI.
• Characters and roleplay chats. Character cards with a description and first message, author’s notes, and “lorebooks” (World Info) — information about the world that is inserted into prompts based on keywords.
• Fine-grained generation settings. Presets, response and context length, temperature, Top P, Top K, repetition penalties, Mirostat, and other parameters — more than in most chat clients.
• Images and voice. Image generation through AUTOMATIC1111 and ComfyUI, text-to-speech (TTS) for responses, and a visual novel mode.
• Convenience. Mobile-friendly layout, a customizable interface, automatic message translation, and third-party extensions.

Installation and Requirements

Node.js 20 or newer is required. SillyTavern runs on Windows, macOS, Linux, Android (through Termux), and Docker; instructions are available at docs.sillytavern.app. Any computer is sufficient for working with cloud-based models. To run models on your own computer, the developers recommend an NVIDIA 3000-series graphics card with at least 6 GB of VRAM. The project does not provide online services or collect user data.

Pricing

SillyTavern is free. You only need to pay for the APIs of cloud-based models if you use them.