SillyTavern review
Free, open-source roleplay frontend: you bring the model, it runs the characters
Edited by Troy ยท Published
SillyTavern is a free, open-source frontend for roleplaying with AI characters on a model you choose, local or cloud. Pick it if you want control over characters, memory and filters and will put in the setup. Skip it if you want an app that works the moment you open it.
- Best for
- Tinkerers who want full control over AI roleplay
- Pricing
- SillyTavern is free and open-source at this review, and the project takes no donations. The model you connect costs extra: API tokens for a cloud service, or your own hardware for a local model.
- Platforms
- Windows, Mac, Linux
Overall: Home-cooked
0 to 100, lower is better. The overall score is the average of the axes below.
35/100
Prose
Does it write like a person? Not scored, since this axis doesn't apply to this app.
Memory
Microwaved
50
Long-term recall depends on summaries the docs tell you to check and on add-on extensions, and long chats eventually overload the context window.
Asks: Does it remember you?
Characters
Microwaved
45
Every character is a card of prompts running on whatever model you connect, so a voice is only as distinct as the card and the model make it.
Asks: Does everyone sound the same?
Story
Does anything happen? Not scored, since this axis doesn't apply to this app.
Pricing
Artisanal
10
SillyTavern is free and open-source and takes no money, so the only bill is for the model you connect.
Asks: Does it charge you for caring?
Art
Does the art look like a person made it? No art to judge.
Decent
0 to 10, higher is better.
6/10
Rewarding once it runs, but setup stands between you and the first scene.
Key facts
| Type | Character chat, AI sandboxes |
|---|---|
| Platforms | Windows, Mac, Linux |
| Pricing at this review | SillyTavern is free and open-source at this review, and the project takes no donations. The model you connect costs extra: API tokens for a cloud service, or your own hardware for a local model. |
| Age rating | Not stated |
| Content policy | Little or no filtering |
| Who writes the characters | Made by other users |
| Memory | Chats are saved locally. Long-term recall comes from summaries, lorebooks and extensions that you set up and check yourself. |
| Best for | Tinkerers who want full control over AI roleplay |
Strengths and watch-outs
- Free and open-source. The project says it will stay that way and does not take donations.
- Connects to local and cloud backends. The docs FAQ says a self-hosted model keeps everything on your PC.
- Character cards, World Info lorebooks, group chats, user personas and built-in document RAG come included.
- SillyTavern adds no filter of its own, so the content rules are set by the model you pick.
- Reviewers report tens of thousands of community character cards that drop straight into a local install.
- Setup is the price of entry. Reviewers call it one of the least beginner-friendly AI tools, and one Reddit beginner spent a month configuring it without running a single session.
- Long-term memory is your job. The docs warn that summaries can lose details or hallucinate, and very long chats overload the context window.
- The model costs extra. Cloud APIs bill per token, and a local model needs a capable graphics card.
- There is no way to run it on iPhone or iPad. On Android it runs only through a route the project does not officially support.
- Paid web-service models log everything on the provider's server, and the docs warn the SillyTavern server is not secure enough to put on the public internet.
Ranked in
- #4 in The best AI chatbots for roleplay
Best for: Tinkerers who want to pick the model and set the rules
Guides for SillyTavern
- AI roleplay guide: how to keep long roleplays from falling apart
How to run AI roleplay that lasts: short character cards, keyword lorebooks, session summaries and fresh chats, and why long chats forget and cost more.
- How to Use SillyTavern: Install It, Connect a Model and Start Chatting
How to use SillyTavern from SillyTavern's own docs: install on Windows, Mac, Linux or Android, connect a model in API Connections, and load a character.
What it is
SillyTavern is a free, open-source interface for chatting and roleplaying with AI characters. The AI itself comes from an LLM backend you connect, either a local model on your own graphics card or a cloud service. SillyTavern supports many of both, and one backend gives instant out-of-the-box chatting. It runs on anything that can run NodeJS 20 or higher.
Everything in SillyTavern is built around character cards, collections of prompts that set the model's behavior. The docs say a card is required for persistent conversations. Other features include World Info lorebooks, group chats, user personas and built-in RAG for documents. The project started in February 2023 and now has over 300 contributors and 3 years of independent development. It is a community project that does not accept donations.
Reviewers report tens of thousands of downloadable character cards on community sites. They also describe a Visual Novel Mode with character sprites, expression changes and scene backgrounds, plus inline image generation through local image tools. The latest release added backend support for a batch of new cloud models.
What it is like to use
Setup decides whether SillyTavern works for you. Users on Reddit are split on how hard it is. One beginner spent a month configuring it, still hadn't run a single roleplay session, and felt burnt out. Others say a beginner using a cloud API service can be chatting within a few hours. One says the current version is easier than ever to install and, once learned, gives far better roleplay quality than the chatbot site they left. A newcomer was told it is more like building your own roleplay platform than visiting a chatbot website: "that's the appeal and the wall." Reviewers call SillyTavern worth the effort if you survive the setup, and one of the least beginner-friendly AI tools.
On long chats, you do the remembering. The built-in Summarize extension writes chat summaries that act as a sort of long-term memory, but the docs warn they may lose details or hallucinate and should be checked by hand. A vector storage extension can retrieve context from beyond the context window. For long, complex stories, Reddit users lean on third-party memory extensions. A community preset author puts the breaking point at about 400 messages: the context window overloads and the AI starts hallucinating, or you pay to send 120k tokens with every message.
SillyTavern applies no filter of its own. The docs FAQ says paid web-service models are often censored and will refuse certain subjects, so your choice of model sets the rules. Performance slows noticeably past about 1000 characters, because SillyTavern was not designed for huge libraries.
Pricing at this review
SillyTavern costs nothing at this review. The project says it will always be free and open-source, and it takes no donations. The bill comes from the model. Reviewers report you pay only for compute or API tokens, typically a few dollars a month for moderate cloud use, and token costs climb as long chats send more context with each message.
Running a model locally swaps the token bill for hardware. The project recommends a 3000-series NVIDIA card with at least 6GB of VRAM, while a Reddit user says enjoyable roleplay probably needs at least 16GB.
Who should skip it
Skip SillyTavern if you want to open an app and start chatting, because nothing happens until you connect a backend and learn the settings. Skip it if you only own an iPhone or iPad, since those can reach SillyTavern only in a browser from another computer on your home Wi-Fi.
Skip it if you want memory that holds without upkeep, or a library well past 1000 characters. If privacy is the reason you are here, use a self-hosted model: paid web-service models log everything on the provider's server, and the docs warn the SillyTavern server is not safe to expose to the public internet.
Receipts
Every claim on this page, with the source text we copied and the day we read it. Last checked against the live app on October 12, 2026.
- Receipt 1Official site
SillyTavern is a community passion project that says it will always be free and open-source.
"SillyTavern is a passion project brought to you by a dedicated community of LLM enthusiasts and will always be free and open-sourced."
docs.sillytavern.app, retrieved
- Receipt 2Official site
SillyTavern started in February 2023 as a fork of TavernAI 1.2.8 and now has over 300 contributors and 3 years of independent development.
"Beginning in February 2023 as a fork of TavernAI 1.2.8, SillyTavern now has over 300 contributors and 3 years of independent development under its belt"
docs.sillytavern.app, retrieved
- Receipt 3Official site
SillyTavern is only an interface; users need an LLM backend. AI Horde gives instant out-of-the-box chatting, and many local and cloud backends are supported.
"Since SillyTavern is only an interface, you will need access to an LLM backend to provide inference. You can use AI Horde for instant out-of-the-box chatting."
docs.sillytavern.app, retrieved
- Receipt 4Official site
It runs on anything that can run NodeJS 20 or higher; for local LLM inference the project recommends a 3000-series NVIDIA card with at least 6GB VRAM.
"it will run on anything that can run NodeJS 20 or higher. If you intend to do LLM inference on your local machine, we recommend a 3000-series NVIDIA graphics card with at least 6GB of VRAM."
docs.sillytavern.app, retrieved
- Receipt 5Official site
SillyTavern is built around character cards, collections of prompts that set the LLM's behavior and are required for persistent conversations.
"A character card is a collection of prompts that set the behavior of the LLM and is required to have persistent conversations in SillyTavern."
docs.sillytavern.app, retrieved
- Receipt 6Official site
The project does not accept monetary donations for itself.
"We do not accept monetary donations for the project itself."
docs.sillytavern.app, retrieved
- Receipt 7Official site
The built-in Summarize extension generates chat summaries that can act as a sort of long-term memory, but the docs warn summaries may lose details or hallucinate and should be checked and corrected manually.
"Since the summaries are generated by language models, the outputs may lose some important details or contain hallucinations, so you're always advised to keep track of the summary state and correct it manually if needed."
docs.sillytavern.app, retrieved
- Receipt 8Official site
Features include World Info (lorebooks), group chats, user personas and built-in RAG for documents.
"World Info support : create rich lore or save tokens on your character card"
docs.sillytavern.app, retrieved
- Receipt 9Reviewer
The docs FAQ says self-hosted models give total privacy, with everything staying on your PC, while paid web services log everything on their servers.
"Total privacy. Everything you write stays on your own PC."
github.com, retrieved
- Receipt 10Reviewer
With paid web-service models, everything is logged on the provider's server, which the docs flag as a privacy concern.
"Everything is logged on their server. Privacy concerns."
github.com, retrieved
- Receipt 11Reviewer
The docs FAQ says paid web-service models are often censored and will refuse certain subjects; the filter comes from the model provider.
"They are often censored and will refuse to chat with you about certain subjects."
github.com, retrieved
- Receipt 12Reviewer
iPhones and iPads can't run SillyTavern but can reach it in a browser from another computer on home Wi-Fi; Android can run it via Termux, which is not officially supported.
"iPhones and iPads are not capable of running the whole SillyTavern app, but since it's just a web interface, you can run it on another computer on your home Wi-Fi, and then access it in your mobile browser."
github.com, retrieved
- Receipt 13Reviewer
SillyTavern wasn't designed for huge character libraries; slowdown becomes noticeable past about 1000 characters.
"Evidential data suggests that the performance degradation starts to become noticeable when you have more than 1000 characters."
github.com, retrieved
- Receipt 14Reviewer
MakeUseOf calls SillyTavern a free, open-source frontend that's worth the effort if you survive the setup, but one of the least beginner-friendly AI tools.
"But that same customizability also makes it one of the least beginner-friendly tools in the AI space at the moment. If you survive the setup, it's an app that's worth the effort"
www.makeuseof.com, retrieved
- Receipt 15Reviewer
Reviewers report a vector storage extension enables long-term memory retrieval via embeddings, pulling past context beyond the context window.
"There's even a vector storage extension for heavier use cases that allows long-term memory retrieval using embeddings"
www.makeuseof.com, retrieved
- Receipt 16Reviewer
Reviewers report a Visual Novel Mode with character sprites, expression changes and scene backgrounds, plus inline Stable Diffusion/FLUX image generation via Automatic1111 and ComfyUI.
"SillyTavern has a Visual Novel Mode that turns the chat into an actual visual novel experience, complete with character sprites, emotional expression changes, and scene backgrounds."
www.makeuseof.com, retrieved
- Receipt 17Reviewer
Reviewers report community sites offer tens of thousands of downloadable character cards that drop into a local install.
"There are sites offering tens of thousands of downloadable character cards that you can drag and drop straight into your local install."
www.makeuseof.com, retrieved
- Receipt 18Forum
A Reddit user said it took a month to set up SillyTavern as a beginner and they still hadn't done a single RP session, feeling burnt out from configuring.
"I still haven't done a single actual RP session yet, I'm feeling a bit burnt out from all the configuring"
www.reddit.com, retrieved
- Receipt 19Forum
Other Reddit users counter that with a service like OpenRouter a beginner can be chatting within a few hours, needing little more than an API key.
"There's barely any configuration you need to do to use ST aside from selecting a service and pasting in the API key."
www.reddit.com, retrieved
- Receipt 20Reviewer
The latest GitHub release added backend support for new Claude models (Claude Fable 5 and 5.1, Opus 4.8 and 5, Sonnet 5) and GPT-5.6 / GPT-6 Astra.
"Anthropic: Added support for Claude Fable 5 and 5.1, Claude Opus 4.8 and 5, Claude Sonnet 5."
github.com, retrieved
- Receipt 21Reviewer
Reviewers report SillyTavern applies no filter of its own; whatever filter exists comes entirely from the local or cloud model you connect.
"It is a chat interface only, with no hosted model of its own, so whatever filter exists comes entirely from whichever local or cloud model you point it at."
tech-insider.org, retrieved
- Receipt 22Reviewer
Reviewers report SillyTavern has no age gate because it is self-hosted software.
"No age gate; self-hosted software"
tech-insider.org, retrieved
- Receipt 23Reviewer
Reviewers report there is no subscription; users pay only for compute or API tokens, with cloud API use typically a few dollars a month for moderate use.
"Cloud API: pay-per-token to OpenRouter, OpenAI or Anthropic, typically a few dollars a month for moderate use No subscription; you pay only for compute or API tokens you use"
tech-insider.org, retrieved
- Receipt 24Official site
Each user account's folder holds character files, conversation history and settings on the local install.
"Created for each user account, this folder contains user-specific data such as character files, conversation history, and settings."
docs.sillytavern.app, retrieved
- Receipt 25Official site
The docs warn the SillyTavern server is not secure enough to expose to the public internet.
"Despite following many security best practices, the SillyTavern server is not secure enough for exposure to the public internet."
docs.sillytavern.app, retrieved
- Receipt 26Forum
A Reddit user migrating from Character.AI was told ST is a frontend, not a chatbot website, more like building your own RP platform: that's the appeal and the wall.
"ST is a frontend, not a chatbot website. You're not just making an account and picking bots"
www.reddit.com, retrieved
- Receipt 27Forum
A Reddit user says the current version is easier than ever to install and, once learned, more comfortable than Character.AI with far better RP quality.
"Current version is easier than ever to install, and once you figure out how to switch models and create prompts, it's even more comfortable to use than CAI"
www.reddit.com, retrieved
- Receipt 28Forum
A Reddit user says running a model locally well enough to enjoy roleplay probably needs at least 16GB VRAM.
"If you want to run locally you are probably going to need at least 16gb Vram to run something sophisticated enough to enjoy the roleplay"
www.reddit.com, retrieved
- Receipt 29Forum
Reddit users lean on third-party memory extensions (Memory Books, qvink, Summaryception, openvault, tunnelvision) for long, complex stories.
"Memory Books using the comprehensive synopsis prompt combined with qvink are my mainstays but I've been looking around for alternatives."
www.reddit.com, retrieved
- Receipt 30Forum
A community preset author describes the long-chat problem: 400 messages in, the context window overloads and the AI starts hallucinating, or you pay to send 120k tokens per message.
"you're 400 messages deep in an RP, your context window is overloaded, and the AI starts hallucinating because it's drowning in old text"
www.reddit.com, retrieved
Questions people ask
Is SillyTavern completely private?
SillyTavern is fully private only with a self-hosted model. In that case the docs FAQ says everything stays on your PC. With paid web-service models, everything is logged on the provider's server, which the docs flag as a privacy concern. Your character files, conversation history and settings sit in each user's folder on your install, and the docs warn that the SillyTavern server is not secure enough to expose to the public internet.
What is SillyTavern AI?
SillyTavern is a free, open-source interface for chatting and roleplaying with AI characters. It contains no AI model of its own, so you connect it to a local or cloud LLM backend. Each conversation runs on a character card, a set of prompts that shapes how the model behaves.
How do I get started with SillyTavern?
Install SillyTavern on any machine that runs NodeJS 20 or higher, then connect a backend. The docs say one supported backend gives instant out-of-the-box chatting. Reddit users say a beginner using a cloud API service can be chatting within a few hours with little more than an API key. For local models, the project recommends a 3000-series NVIDIA card with at least 6GB of VRAM.
Is SillyTavern free?
Yes. At this review SillyTavern is free and open-source, the project says it always will be, and it takes no donations. Reviewers report that you pay only for compute or API tokens. Moderate cloud API use typically runs a few dollars a month.
Does SillyTavern have a content filter?
SillyTavern has no filter of its own, so any filter comes from the model you connect. The docs FAQ says paid web-service models are often censored and will refuse certain subjects. Reviewers report there is no age gate, because SillyTavern is self-hosted software.
Can I use SillyTavern on an iPhone or Android phone?
An iPhone or iPad cannot run SillyTavern itself, but it can open SillyTavern in a browser when it runs on another computer on your home Wi-Fi. Android can run it through a terminal app, a setup the project does not officially support.
More like this
Slop check: passedThis page went through our slop counter before publishing. How we check our own copy