rpg companion sillytavern in 2026: AI Angels vs SillyTavern

SillyTavern is an open-source front-end that connects to large language models (LLMs) via APIs, allowing users to create and converse with RPG companions. These companions are defined by character cards—JSON or PNG files containing a name, description, personality traits, example dialogue, and scenario context. Unlike generic chatbots, an RPG companion in SillyTavern maintains consistent lore, remembers past interactions (when using a model with sufficient context length), and can be assigned multiple personas for the user. The platform supports advanced features like group chats, narration, and automatic worldbuilding through lorebooks. Because SillyTavern is model-agnostic, users can switch between providers such as OpenAI, Anthropic, or local models via KoboldCPP or Oobabooga. This flexibility makes it a powerful tool for solo RPG enthusiasts who want a dynamic, AI-driven game master or party member. However, setup requires technical know-how: users must obtain their own API keys and configure model parameters like temperature and context size to optimize roleplay quality.

To create an RPG companion in SillyTavern, start by downloading the latest release from the official GitHub repository and following the installation guide for your OS (Windows, Linux, or Android via Termux). Once launched, access the web UI and navigate to the Character Management tab. Click 'Create Character' and fill out the character card fields: name, avatar (optional), description (what the character looks like and their role), personality (trait list or paragraph), scenario (the current situation), and example dialogues (few-shot prompts to guide tone). For an RPG companion, include specific commands or actions in the example messages, like '[The dwarf grumbles and pulls out a map.]'. Next, configure the AI model: you can connect to ChatGPT, Claude, or a local model via the API tab. Set the context size to at least 4096 tokens to hold game state. Finally, test the companion by sending a few messages—adjust temperature (0.7–1.0 for creativity) and repetition penalty as needed. Advanced users can import character cards from sites like Character.AI or Chub.ai, but manual tweaking ensures the companion fits your campaign.

Lorebooks are SillyTavern's most powerful feature for RPG companions. They function as injectable knowledge bases: when a keyword from the lorebook appears in the conversation, the corresponding entry is inserted into the model's context. For example, a lorebook for a fantasy kingdom might contain entries for 'King Aldric', 'The Crimson Forest', and 'Dragon’s Breath Potion'. Each entry includes a key (the trigger word) and content (a paragraph of lore). To create a lorebook, go to the Lorebooks tab, click 'New Lorebook', and add entries with keys and associated text. You can assign a lorebook to a specific character or to the entire chat. For RPG companions, lorebooks ensure the AI remembers geography, NPCs, item effects, and quest history without overloading the character card. Best practice: use short, factual entries (50–100 words) and avoid contradictory information. SillyTavern supports multiple lorebooks per chat, so you can separate world lore, character backstories, and rule systems. Stanford's Center for Research on Foundation Models notes that contextual injection improves model consistency ([CRFM](https://crfm.stanford.edu)), which is exactly what lorebooks achieve.

The quality of an RPG companion heavily depends on the model's parameters. SillyTavern exposes these in the 'AI Response Configuration' panel. For roleplay, start with temperature 0.9–1.1 to encourage creative and varied responses. Top P (nucleus sampling) at 0.9–1.0 helps maintain coherence. The repetition penalty should be around 1.1–1.2 to prevent loops. Context size is critical: use the maximum your model supports (e.g., 8k for GPT-4, 4k for many local models) to retain game state. Instruct mode (enabled by default) tells the model to follow system prompts—keep the system prompt concise, e.g., 'You are an RPG game master. Describe scenes, control NPCs, and respond to player actions.' For local models like Llama 3 or Mistral, lower temperature (0.7) may yield more logical responses. Experiment with the 'Character Bias' slider to amplify personality traits. If the companion becomes repetitive, increase repetition penalty or reduce top P. SillyTavern also supports custom 'Response Templates' that format the model's output—useful for adding dice roll syntax or action markers like '*swings sword*'.

SillyTavern allows multiple RPG companions to interact in group chats, enabling party dynamics or council meetings. To create a group, go to the 'Group' tab and add characters from your roster. Each character retains its own card and lorebook, and the AI alternates responses based on the conversation order. You can set a 'Narrator' character that describes scenes and controls non-party NPCs. For complex encounters, use the 'Reply as' feature to manually trigger a specific character's response. SillyTavern also supports 'Character Notes' that appear as reminders in the context—useful for tracking party HP, inventory, or quest objectives. Another advanced feature is 'Auto-summarize', which condenses past conversation into a summary entry, saving context tokens. However, for long campaigns, consider using a model with large context (e.g., GPT-4-32k or Claude 2.1) to avoid losing details. The platform's 'Extension' system adds dice rolling, web search, and image generation, making it a full RPG toolkit. Note that setup complexity increases with each feature—group chats require careful balancing of token limits.

While SillyTavern is powerful, its steep learning curve and requirement for external API keys or local hardware push some users toward simpler alternatives. Platforms like Character.AI offer ready-made RPG bots but lack custom lorebooks and deep parameter control. For those who want a more accessible AI companion with built-in memory and no setup, AIAngels provides a free tier with unlimited text chat, permanent memory, and a custom companion builder—no API keys required. AIAngels' premium plan ($2.99/mo annual) includes image generation and voice messages, making it a self-contained solution for RPG enthusiasts who prefer plug-and-play over configuration. However, advanced users who need full control over model parameters, lorebooks, and multi-character group chats will still prefer SillyTavern. The choice comes down to technical comfort: SillyTavern for tinkerers, AIAngels for those who just want to roleplay without managing infrastructure.