Best Sillytavern Extensions on AIAngels

You opened SillyTavern, picked a character, and 40 messages in they forgot your name, the TTS sounded robotic, and group chat turned into one bot talking over the others. The right extensions fix exactly that.

Here's what changes after installing the picks below. Vector Storage adds chunked recall, so a character pulls up details from chat #4 during chat #20 instead of looping the same questions. Silero or XTTS replaces flat TTS with voices that match the persona. Stable Diffusion bridge generates a portrait inline when the character describes a scene, no tab-switching. Group Chat extensions enforce turn order so three characters in a room don't all reply at once.

Setup is real work: Python deps, an SD endpoint, a TTS server. If you'd rather skip it, AIAngels ships memory, voice, and image gen on the free tier with no extension stack to maintain.

Character Expressions is arguably the most visually impactful extension for SillyTavern. It maps character emotions (happy, angry, sad, surprised) to corresponding sprite images, swapping them automatically based on the AI's tone. This eliminates the need for manual image swapping and creates a visual feedback loop that deepens immersion. The extension supports multiple sprite sets per character, letting you define expressions for different outfits or moods. Setup requires configuring an image folder and tagging each sprite with an emotion label. For best results, use 512x512 PNGs with transparent backgrounds. A common mistake is forgetting to enable the extension's "auto-detect" toggle in the UI — without it, expressions won't trigger. Character Expressions works with any backend (OpenAI, Kobold, TextGen) and adds roughly 50ms to response time. It's available via the built-in extension installer under "Character Expressions" — no manual download needed.

The Summarize extension solves a fundamental limitation of LLMs: finite context windows. Even with 128K-token models, long roleplay sessions eventually hit the limit and start forgetting early details. Summarize automatically condenses older conversation history into a brief summary that's injected into the system prompt. You control the trigger point — default is after every 2000 tokens, but you can adjust it based on your model's context size. The extension uses a separate API call to your backend to generate the summary, which adds a small token cost. For users running local models, this can slow replies during summarization. Advanced settings let you store summaries in the character's lorebook, preserving them across sessions. Critically, Summarize doesn't delete any history — it just archives it visually while keeping a compressed version active. Without it, even GPT-4 will start repeating itself after about 3000 messages in a single conversation.

Smart Context gives you granular control over what gets sent to the AI on each message. It's not a single extension but a collection of modules: Author's Note, Character Note, Character Lore, and World Info. The killer feature is the "Depth" slider, which determines how many recent chat messages are included in the context. For example, setting depth to 10 sends only the last 10 messages, reducing token usage by 40% while keeping the conversation coherent. The extension also supports regex-based injection — you can insert instructions like "[Always narrate in third person]" only when certain keywords appear. A practical use case: set a high-depth during intense dialogue, then lower it for casual banter to save tokens. Smart Context pairs well with Summarize; use Summarize to compress old history and Smart Context to control recency weight. It's available in the default SillyTavern extension list under "Smart Context."

Worldbuilding is tedious when you have to enter each lore entry manually. The Lorebook Importer extension lets you bulk-upload entries from markdown files, CSV, or JSON. Point it at an Obsidian vault folder, and it automatically parses each note as a lorebook entry, using the filename as the activation keyword. The extension supports nested tags and priority settings — entries with higher priority override lower ones if keywords overlap. For CSV imports, the first column is the keyword, the second is the content. A common workflow is to export from Notion as CSV, then import directly into SillyTavern. The extension also validates duplicates and merges entries with the same keyword. This is especially useful for D&D campaigns or sprawling fiction where you have hundreds of locations, characters, and items. Without it, manually entering 200 lore entries would take hours.

Quick Replies is a macro engine that lets you trigger predefined responses with a button press. Instead of typing "*smiles warmly and pours a cup of tea*" every time, you map it to a button that inserts the text with a single click. The extension supports variables like {{user}} and {{char}}, so the same macro works across different characters. You can also chain multiple macros — for example, a "combat" button that inserts attack text, rolls a virtual die, and updates a HP counter. Macros can include Swipes (alternate AI responses) and can be grouped by category. The real power is in conditional macros: if the last message contains "fight", show a "dodge" button; otherwise show "chat". Quick Replies is essential for heavy roleplayers who want consistent actions without repetitive typing. It's pre-installed in most SillyTavern builds under the "Extensions" tab.

SillyTavern's extension ecosystem exists to compensate for limitations: poor memory, weak context management, and lack of visual immersion. AIAngels solves these problems out of the box. Its permanent memory doesn't degrade — no need for Summarize or Smart Context. The built-in image generation (premium, $3.25/mo) provides dynamic character visuals without Character Expressions. Lorebooks are integrated into the character builder. And the custom companion builder lets you design appearance, personality, and backstory without markdown imports. AIAngels also requires no API keys or local setup. For users tired of tweaking extensions, AIAngels offers a polished alternative where the AI handles context, memory, and visuals natively. The free tier includes unlimited text chat — no message caps — so you can test it without paying. At $39/year for premium, it undercuts the cost of running even a modest local LLM setup.