PromptPlay
Guide

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.

Edited by · Published

Short answer

Install NodeJS (latest LTS) and Git, clone SillyTavern's release branch, then run Start.bat on Windows or bash start.sh on Mac/Linux. It opens in your browser. Enter a persona name, click API Connections in the top bar and connect a model. AI Horde is free with the key 0000000000 at this review. Then open Character Management, pick Seraphina and press Enter.

What you need before installing

SillyTavern needs NodeJS 20 or higher, plus Git for the recommended install. The project's README makes these points:

  • Hardware: anything that runs NodeJS 20 is enough. Running a model on your own machine raises the bar to a 3000-series NVIDIA card with at least 6GB of VRAM.
  • What it is: a locally installed interface for text-generation LLMs, image generation engines and TTS voice models.
  • Hosting and tracking: the project offers no online or hosted services, doesn't programmatically track user data, and will always be free and open source.
  • Platforms: there are install guides for Windows, MacOS/Linux, Android (Termux) and Docker.

SillyTavern runs no model itself. Reviewers report that even once it's running locally, you can't chat until you connect a backend, either with an API key or a local inference server.

Pick the release branch. SillyTavern's docs recommend it for most users because it is the most stable and typically updates once a month. The staging branch has the latest features and updates several times a day, but the docs say it may break at any time and isn't for casual use.

Install with Git rather than a zip. The docs always recommend Git because updating is then just git pull, while zip installs need a tedious manual process.

Install SillyTavern on Windows

  1. Install NodeJS. The docs recommend the latest LTS.
  2. Install Git for Windows.
  3. Make a folder for SillyTavern outside Windows-controlled folders such as Program Files and System32.
  4. Open a Command Prompt in that folder.
  5. Clone the release branch.
  6. Double-click Start.bat. Don't run it as admin.

NodeJS installs the requirements, the server starts, and SillyTavern opens in your browser. Windows 7 can't run it because it can't run NodeJS 20. The docs also describe a GitHub Desktop route, and cloning the repository doesn't require a GitHub account.

Install SillyTavern on Mac or Linux

  1. Install git and nodeJS.
  2. Open a Terminal and clone the repo.
  3. Run cd SillyTavern.
  4. Run bash start.sh.

Mac users have a second route: install Homebrew, then git, then clone and run the SillyTavern Launcher.

Install SillyTavern on Android

SillyTavern runs natively on Android through Termux, but the docs say Termux installs are not officially supported and aren't guaranteed to work. Install Termux from F-Droid (the docs' recommendation) or GitHub releases, never Google Play, whose version is no longer maintained.

32-bit Android needs an external dependency that npm can't install, so a 64-bit phone is the easier path. Once SillyTavern is installed, optional aliases give you shortcuts:

  • st starts SillyTavern.
  • stup updates it.
  • pkgup updates Termux packages.

Use SillyTavern on an iPhone, iPad or another device

iPhones and iPads can't run the whole app. Instead, run SillyTavern on a computer on your home Wi-Fi and open it in the mobile browser. By default the server only accepts connections from its own machine, so change that first:

  1. Open config.yaml in the SillyTavern root folder, not /SillyTavern/default/config.yaml.
  2. Set listen to true.
  3. Configure at least one access control method. With remote listening on, the server won't start without one.
  4. If you whitelist addresses, include 127.0.0.1, or you can't connect from the host machine.
  5. On Windows 11, enable the Private Network profile under Settings > Network and Internet > Ethernet. Without it, remote devices can't connect even with firewall rules.
  6. In the phone's browser, type the address with the plain http prefix, not https.

Don't port-forward to expose the server to the internet. For access away from home, the docs point to a VPN or a tunnel such as Cloudflare Zero Trust, ngrok or Tailscale.

Set your persona

The onboarding screen asks for a persona name. That name is how you appear in the chat.

Connect a model in API Connections

AI Horde is the free way to test your install. SillyTavern's docs describe it as a community-driven AI service with a variety of models.

  1. Click API Connections in the top bar.
  2. Enter an API key. At this review 0000000000 works, or you can get a free key from AI Horde.
  3. Select a few AI models from the top of the list. You can change them later.
  4. Close API Connections.

Horde works out of the box on volunteers' GPUs, so the worker you land on decides your wait times, settings and available models.

To connect OpenAI instead:

  1. Click API Connections in the top bar.
  2. Under API, select Chat Completion (OpenAI).
  3. Under Chat Completion Source, select OpenAI.
  4. Paste your API key.
  5. Click Connect.
  6. Check that it says Valid.

SillyTavern uses GPT-4 Turbo with OpenAI by default. You can choose other models, and the docs warn you to learn the pricing before you do.

Your key is saved in secrets.json in the user data directory (/data/default-user/secrets.json by default) and is hidden in the interface after you save and refresh. Never share it: whoever has your OpenAI secret key can use GPT at your expense.

Which backend to pick

Backend Cost at this review Notes from the docs
AI Horde Free Volunteer workers set wait times, settings and models
OpenRouter Pay-per-token credits, plus free models with limited daily requests No enforced moderation unless the LLM vendor requires it
Google AI Studio Free tier with rate limits (Gemini Flash) May require billing information
DeepSeek Credit purchase, $2 minimum Fairly cheap for the quality
OpenAI, Claude Prepaid credits, charged per prompt Cloud content filtering applies
Local model Free Complex setup the SillyTavern team doesn't support; models are 5-50GB each

The docs frame the choice as a trade-off:

  • Cloud APIs are smarter than most local models, but all filter content to varying degrees and most require payment. The FAQ adds that paid web-service AIs log everything on their servers.
  • Local models have no content filter, offer total privacy and can run without internet.

Use Horde to confirm the install works. Pick a cloud API if you want quality, or a local model if privacy matters and you have the GPU.

Start your first chat

  1. Click Character Management at the far right of the top bar.
  2. Select an existing character, such as Seraphina.
  3. Type in the chat box at the bottom.
  4. Press Enter or click Send.

Once the chat is running, these controls do most of the work:

  • Swipe generates a different response. The Right arrow key does the same.
  • Edit changes any message. The Up arrow key edits the last one.
  • Options, next to the chat bar, holds author's notes and chat file management.
  • The ••• message menu has Checkpoint, which creates a story checkpoint, and Branch, which starts an alternate story path.

SillyTavern keeps your chat history, so you can resume a chat, start a new one or review old ones. If replies come back short, send an empty message to make the bot keep writing, or raise Response Length.

Create or import a character

  1. Click Character Management.
  2. Click Create New Character.
  3. Fill in Character Name.
  4. Fill in Description.
  5. Fill in First Message.
  6. Click Create Character.

Choose the name carefully: you can edit everything about a character after creating it except the name.

Reviewers report that a character card is a PNG image with personality, dialogue style, scenario and behavior rules embedded as metadata. If a card throws an invalid PNG card error, the docs give two usual causes:

  • The embedded definitions were stripped. Use the raw PNG exactly as posted.
  • The file is really a WEBP. Try renaming it to .webp.

SillyTavern wasn't designed for huge libraries, and slowdown becomes noticeable past 1000 characters.

Set SillyTavern up for roleplay

The FAQ fixes the most common complaint, the AI writing your character's actions, with an Author's Note phrase such as "You are never allowed to dictate actions or speech for {{user}}."

Three more features matter for roleplay:

  • World Info (lorebooks): reviewers report it injects persistent character, location and context snippets into the prompt whenever their keywords appear in the conversation.
  • Visual Novel mode: characters appear as sprites, as in visual novels like Doki Doki Literature Club.
  • Presets: a June 2026 user post describes a preset as telling the AI how to roleplay in a frontend like SillyTavern.

Update SillyTavern

  1. Open a terminal in the SillyTavern folder.
  2. Run git pull.
  3. Start SillyTavern again.

On Windows, try UpdateAndStart.bat first. Two errors come up after updates:

  • "Cannot find module" on startup: a new npm package is required. Run npm install in the SillyTavern folder. Start.bat and start.sh do this automatically.
  • Changed system files block git pull: move or delete the file, or run git pull --rebase --autostash.

Reset settings and back up your data

  1. Open User Settings.
  2. Go to Account.
  3. Click Download Backup to export all user data, or save a Settings Snapshot.
  4. In the Danger Zone, click Reset Settings to restore factory defaults, or Reset Everything for a complete account wipe and factory reset.

In default standalone mode, config.yaml and the data directory live in the server directory, and all data stays in the installation path. The Clean-Up tool permanently deletes files and can't be undone, so back up first if you're unsure.

Fix a slow or laggy interface

In User settings, enable No Blur Effect (Fast UI) and Reduced motion, and make sure hardware acceleration is on. For input lag, check browser extensions first, which the docs name as the most common cause.

What new users run into

One September 2026 review calls SillyTavern one of the least beginner-friendly AI tools right now, though worth it if you survive the setup. Users report similar friction:

  • A returning user found every beginner video tutorial obsolete.
  • Another got "weird" output and wondered whether they had picked the wrong APIs or set up characters poorly.
  • One got a refusal message from Claude, which fits the docs' warning that cloud APIs filter content.
  • The June 2026 preset post says the hobby is getting more expensive.
Reviews

Apps this guide covers

  • SillyTavern review35Home-cooked

    Free, open-source roleplay frontend: you bring the model, it runs the characters

Receipts

Sources

The source text behind the claims above, and the day we read it.

  1. Receipt 1Official site

    SillyTavern's docs say you can use AI Horde for free; it is a community-driven AI service with a variety of models.

    "You can use AI Horde for free. AI Horde is a community-driven AI service that provides access to a variety of AI models."

    docs.sillytavern.app, retrieved

  2. Receipt 2Official site

    On the onboarding screen, you enter a name for your persona, which is used in the chat.

    "In SillyTavern's onboarding screen, enter a name for your persona. This name will be used in the chat."

    docs.sillytavern.app, retrieved

  3. Receipt 3Official site

    For AI Horde, click API Connections in the top bar and enter an API key; 0000000000 works for now, or get a free key from AI Horde.

    "Enter an API key for AI Horde. You can use 0000000000 for now, or get a free key from AI Horde."

    docs.sillytavern.app, retrieved

  4. Receipt 4Official site

    Select a few AI models from the top of the list (changeable later), close API Connections, type in the chat box at the bottom and press Enter.

    "Select some AI models to use. Just choose a few from the top. You can always change them later. Close the API Connections window. Enter a message in the chat box at the bottom and press Enter."

    docs.sillytavern.app, retrieved

  5. Receipt 5Official site

    For OpenAI: under API select Chat Completion (OpenAI), under Chat Completion Source select OpenAI, paste the key, click Connect and confirm it says Valid.

    "Under API, select Chat Completion (OpenAI). Under Chat Completion Source, select OpenAI. Paste the API key you saved in the previous step. Click the Connect button. Confirm it says Valid."

    docs.sillytavern.app, retrieved

  6. Receipt 6Official site

    By default SillyTavern uses GPT-4 Turbo with OpenAI; other models can be chosen, but the docs warn to learn the pricing.

    "By default, SillyTavern will use GPT-4 Turbo. You can choose a different model, but educate yourself on the pricing."

    docs.sillytavern.app, retrieved

  7. Receipt 7Official site

    The docs warn never to share your OpenAI secret key, since whoever has it can use GPT at your expense.

    "DO NOT SHARE THIS KEY. WHOEVER HAS IT CAN USE YOUR ACCOUNT TO USE GPT AT YOUR EXPENSE."

    docs.sillytavern.app, retrieved

  8. Receipt 8Official site

    To test, click Character Management at the far right of the top bar and select an existing character such as Seraphina, then type and press Enter or click Send.

    "In SillyTavern's top bar, click Character Management at the far right. Select an existing character such as Seraphina."

    docs.sillytavern.app, retrieved

  9. Receipt 9Official site

    The release branch is recommended for most users, is the most stable, and is typically updated once a month.

    "release - &##x1F31F; Recommended for most users. This is the most stable and recommended branch, updated only when major releases are pushed. It's suitable for the majority of users. Typically updated once a month."

    docs.sillytavern.app, retrieved

  10. Receipt 10Official site

    The staging branch is not recommended for casual use: it has the latest features but may break at any time and updates several times daily.

    "staging - ⚠️ Not recommended for casual use. This branch has the latest features, but be cautious as it may break at any time. Only for power users and enthusiasts. Updates several times daily."

    docs.sillytavern.app, retrieved

  11. Receipt 11Official site

    In default standalone mode, config.yaml and the data directory live in the server directory; all data stays in the installation path.

    "Standalone mode (default) - uses the config.yaml file and data directory in the server directory. All data will be constrained to the installation path."

    docs.sillytavern.app, retrieved

  12. Receipt 12Official site

    The Windows guide says not to install into Windows-controlled folders (Program Files, System32), not to run Start.bat as admin, and that Windows 7 can't run it because it can't run NodeJS 20.

    "DO NOT INSTALL INTO ANY WINDOWS CONTROLLED FOLDER (Program Files, System32, etc). DO NOT RUN START.BAT WITH ADMIN PERMISSIONS INSTALLATION ON WINDOWS 7 IS IMPOSSIBLE AS IT CANNOT RUN NODEJS 20"

    docs.sillytavern.app, retrieved

  13. Receipt 13Official site

    The Git install on Windows: install NodeJS (latest LTS recommended) and Git for Windows, then open a Command Prompt in your folder and clone the release branch.

    "for Release Branch: git clone https://github.com/SillyTavern/SillyTavern -b release"

    docs.sillytavern.app, retrieved

  14. Receipt 14Official site

    After cloning, double-click Start.bat; NodeJS installs requirements, the server starts and SillyTavern opens in your browser.

    "Once everything is cloned, double-click Start.bat to make NodeJS install its requirements. The server will then start, and SillyTavern will pop up in your browser."

    docs.sillytavern.app, retrieved

  15. Receipt 15Official site

    There's also a GitHub Desktop route, and you don't need a GitHub account to clone the repository.

    "After installing GitHub Desktop, click on Clone a repository from the internet.... (Note: You do NOT need to create a GitHub account for this step)"

    docs.sillytavern.app, retrieved

  16. Receipt 16Reviewer

    The GitHub README describes SillyTavern as a locally installed interface for interacting with text-generation LLMs, image generation engines and TTS voice models.

    "SillyTavern (or ST for short) is a locally installed user interface that allows you to interact with text generation LLMs, image generation engines, and TTS voice models."

    github.com, retrieved

  17. Receipt 17Reviewer

    The README says the project offers no online or hosted services, doesn't programmatically track user data, and will always be free and open source.

    "We do not provide any online or hosted services, nor programmatically track any user data. SillyTavern is a passion project brought to you by a dedicated community of LLM enthusiasts, and will always be free and open sourced."

    github.com, retrieved

  18. Receipt 18Reviewer

    Hardware requirements are minimal (anything that runs NodeJS 20 or higher); for local LLM inference the project recommends a 3000-series NVIDIA card with at least 6GB of VRAM.

    "The hardware requirements are minimal: 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"

    github.com, retrieved

  19. Receipt 19Reviewer

    The README lists install guides for Windows, MacOS/Linux, Android (Termux) and Docker.

    "Android (Termux) Installation Guide"

    github.com, retrieved

  20. Receipt 20Reviewer

    SillyTavern supports many LLM APIs (KoboldAI/CPP, Horde, NovelAI, Ooba, Tabby, OpenAI, OpenRouter, Claude, Mistral and more), a mobile-friendly layout, Visual Novel Mode, image generation, TTS and WorldInfo lorebooks.

    "SillyTavern provides a single unified interface for many LLM APIs (KoboldAI/CPP, Horde, NovelAI, Ooba, Tabby, OpenAI, OpenRouter, Claude, Mistral and more), a mobile-friendly layout, Visual Novel Mode"

    github.com, retrieved

  21. Receipt 21Official site

    In chat, the Swipe button generates a different response and the Edit button edits any message; Right arrow swipes and Up arrow edits the last message.

    "You can also use the Right arrow key to swipe, and the Up arrow key to edit the last message in the chat."

    docs.sillytavern.app, retrieved

  22. Receipt 22Official site

    The Options button next to the chat bar holds more chat options such as author's notes and chat file management.

    "Chat options : Click the Options button next to the chat bar for more chat options like author's notes and chat file management."

    docs.sillytavern.app, retrieved

  23. Receipt 23Official site

    The message actions (•••) menu includes Checkpoint (create a story checkpoint) and Branch (start an alternate story path).

    "Checkpoint : Create story checkpoint"

    docs.sillytavern.app, retrieved

  24. Receipt 24Official site

    SillyTavern runs natively on Android through Termux.

    "SillyTavern can be run natively on Android devices using Termux."

    docs.sillytavern.app, retrieved

  25. Receipt 25Official site

    Don't install Termux from Google Play, since that version is no longer maintained; use F-Droid (recommended) or GitHub releases.

    "Avoid installing Termux from the Google Play Store, that version is no longer maintained. Instead, use F-Droid (recommended) or GitHub releases to get the latest version."

    docs.sillytavern.app, retrieved

  26. Receipt 26Official site

    32-bit Android needs an external dependency that npm can't install.

    "32-bit Android requires an external dependency that can't be installed with npm."

    docs.sillytavern.app, retrieved

  27. Receipt 27Official site

    Optional aliases give shortcuts: st starts SillyTavern, stup updates it, pkgup updates Termux packages.

    "st to start SillyTavern stup to update SillyTavern pkgup to update Termux packages"

    docs.sillytavern.app, retrieved

  28. Receipt 28Reviewer

    MakeUseOf (Sept 2026) reports SillyTavern is one of the least beginner-friendly AI tools right now, though worth it if you survive the setup.

    "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

  29. Receipt 29Reviewer

    Reviewers report a character card is a PNG image with personality, dialogue style, scenario and behavior rules embedded as metadata.

    "A character card is literally a PNG image with the personality description, dialogue style, scenario setup, and behavior rules embedded as metadata inside the file"

    www.makeuseof.com, retrieved

  30. Receipt 30Reviewer

    Reviewers report World Info (lorebooks) injects persistent character, location and context snippets into the prompt when keywords appear in the conversation.

    "a system for defining persistent characters, locations, and context snippets that automatically get injected into the prompt when specific keywords appear in the conversation"

    www.makeuseof.com, retrieved

  31. Receipt 31Reviewer

    Reviewers report that even once it runs locally you can't chat until you connect a model backend, with an API key or a local inference server.

    "Even if you get it running locally, you still won't be chatting with anything. You then need to separately connect a model backend, paste in an API key or point it at a local inference server"

    www.makeuseof.com, retrieved

  32. Receipt 32Official site

    The FAQ says SillyTavern handles communication with AI models, lets you create character cards and import other people's characters, and keeps chat history so you can resume, start new chats or review old ones.

    "It will keep your chat history with a character, allowing you to resume at any time, start a new chat, review old chats, etc."

    docs.sillytavern.app, retrieved

  33. Receipt 33Official site

    The FAQ says paid web-service AIs cost money, log everything on their servers, and are often censored; self-hosted models offer total privacy and can run without internet.

    "They cost money to use. Everything is logged on their server. Privacy concerns. They are often censored and will refuse to chat with you about certain subjects."

    docs.sillytavern.app, retrieved

  34. Receipt 34Official site

    iPhones and iPads can't run the whole app, but you can run it on another computer on home Wi-Fi and open it in a mobile browser.

    "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."

    docs.sillytavern.app, retrieved

  35. Receipt 35Official site

    Termux installs on Android are not officially supported and aren't guaranteed to work.

    "(NOTE: Termux installations are not officially supported, and we can't guarantee it will work.)"

    docs.sillytavern.app, retrieved

  36. Receipt 36Official site

    An invalid PNG card error usually means the embedded definitions were stripped (use the raw PNG as posted) or the file is really a WEBP; try renaming it to .webp.

    "The PNG file was actually a WEBP file with a.png filename. You can try renaming the card to.webp before importing, or look for a proper PNG version of the image."

    docs.sillytavern.app, retrieved

  37. Receipt 37Official site

    To make a character: click Character Management, then Create New Character, fill in Character Name, Description and First Message, and click Create Character.

    "Click the Character Management button Click Create New Character Under Character Name, give a name, like Amanda"

    docs.sillytavern.app, retrieved

  38. Receipt 38Official site

    You can edit everything about a character after creating it except the name.

    "NOTE: you can go back and edit any of this information once the character is created, except the name."

    docs.sillytavern.app, retrieved

  39. Receipt 39Official site

    API keys are saved in secrets.json in the user data directory (default /data/default-user/secrets.json) and are hidden in the interface after saving and refreshing.

    "SillyTavern saves your API keys to a secrets.json file in the user data directory ( /data/default-user/secrets.json is the default path)."

    docs.sillytavern.app, retrieved

  40. Receipt 40Official site

    Slowdown becomes noticeable with more than 1000 characters; SillyTavern wasn't designed for huge character libraries.

    "Evidential data suggests that the performance degradation starts to become noticeable when you have more than 1000 characters."

    docs.sillytavern.app, retrieved

  41. Receipt 41Official site

    For a slow or jittery UI, enable No Blur Effect (Fast UI) in User settings, enable Reduced motion, and use hardware acceleration; input lag is most often caused by browser extensions.

    "Try enabling the No Blur Effect (Fast UI) mode on the User settings panel."

    docs.sillytavern.app, retrieved

  42. Receipt 42Official site

    To make the bot keep writing, send an empty message (type nothing, click Send); raising Response Length also helps with short replies.

    "you can send an empty user message by typing nothing into the Input Bar and clicking Send. This will force the bot to continue the story."

    docs.sillytavern.app, retrieved

  43. Receipt 43Official site

    To stop the AI from writing your character's actions, the FAQ suggests Author's Note phrases like "You are never allowed to dictate actions or speech for {{user}}."

    "You are never allowed to dictate actions or speech for {{user}}."

    docs.sillytavern.app, retrieved

  44. Receipt 44Official site

    Local APIs are free with no content filter, but setup can be complex, the SillyTavern team doesn't support it, and models are 5-50GB each.

    "They are free to use and have no content filter. Installation process can be complex ( The SillyTavern dev team does not provide support for this )."

    docs.sillytavern.app, retrieved

  45. Receipt 45Official site

    Cloud LLM APIs are smarter than most local models but all filter content to varying degrees, and most require payment.

    "However, they all have content filtering of varying degrees, and most require payment"

    docs.sillytavern.app, retrieved

  46. Receipt 46Official site

    AI Horde works out of the box using volunteers' GPUs, but you're at the mercy of the worker for wait times, settings and available models.

    "At the mercy of the Worker in terms of generation wait times, AI settings, and available models"

    docs.sillytavern.app, retrieved

  47. Receipt 47Official site

    OpenRouter has a pay-per-token credit system plus free models with limited daily requests, and no enforced moderation unless the LLM vendor requires it.

    "Pay-per-token credit system, as well as free models with limited daily requests No enforced moderation, unless required by the LLM vendor"

    docs.sillytavern.app, retrieved

  48. Receipt 48Official site

    DeepSeek requires a credit purchase ($2 minimum), though its models are fairly cheap for their quality.

    "Requires a payment for credits ($2 minimum), but the models are fairly cheap for their quality"

    docs.sillytavern.app, retrieved

  49. Receipt 49Official site

    Google AI Studio has a free tier with rate limits (Gemini Flash) and may require billing information.

    "Has a free tier with rate limits (Gemini Flash), may require billing information"

    docs.sillytavern.app, retrieved

  50. Receipt 50Official site

    OpenAI and Claude both require prepaying for credits and charge per prompt.

    "Requires prepayment for credits and charges per prompt"

    docs.sillytavern.app, retrieved

  51. Receipt 51Forum

    A Reddit user returning to SillyTavern said all the beginner YouTube tutorials are obsolete and no one had an easy-to-follow tutorial.

    "All the beginner tutorials on youtube are obsolete now, and everyone here seems to already be too deep noone has an easy to follow tutorial."

    www.reddit.com, retrieved

  52. Receipt 52Forum

    A Reddit user said their setup output was "weird" and wondered whether they had picked the wrong APIs or set characters up poorly.

    "Like I sometimes think I didn't use the right APIs or I just set the characters up like very underwhelming. Is it really so complicated or do I just miss the right informations ?"

    www.reddit.com, retrieved

  53. Receipt 53Forum

    A Reddit user who tried Claude got a refusal message when generating text.

    "I will not engage with or generate that type of content."

    www.reddit.com, retrieved

  54. Receipt 54Forum

    A June 2026 Reddit preset post describes a preset as telling the AI how to roleplay in a frontend like SillyTavern, and says the hobby is getting more expensive.

    "Developed to save money on cache in a climate where this hobby is getting more expensive."

    www.reddit.com, retrieved

  55. Receipt 55Official site

    On Mac/Linux you install git and nodeJS, clone the repo in a Terminal, cd SillyTavern, and run bash start.sh.

    "cd SillyTavern to navigate into the install folder. Run the start.sh script with one of these commands: bash start.sh"

    docs.sillytavern.app, retrieved

  56. Receipt 56Official site

    Mac users can also install Homebrew, then git, then clone and run the SillyTavern Launcher.

    "Download Sillytavern Launcher with: git clone https://github.com/SillyTavern/SillyTavern-Launcher.git"

    docs.sillytavern.app, retrieved

  57. Receipt 57Official site

    By default the server only accepts connections from its own machine; set listen to true in config.yaml to let other devices such as phones connect.

    "By default, the ST server only accepts connections from the machine that it's running on (localhost). To allow it to listen for connections from other devices, set the listen option in config.yaml to true."

    docs.sillytavern.app, retrieved

  58. Receipt 58Official site

    Once remote listening is on, you must configure at least one access control method or the server won't start.

    "After enabling remote connection listening, you must configure at least one access control method. Otherwise, the server will not start."

    docs.sillytavern.app, retrieved

  59. Receipt 59Official site

    The docs say not to port-forward to expose the server to the internet; use a VPN or a tunnel like Cloudflare Zero Trust, ngrok or Tailscale.

    "You should not use port forwarding to expose your ST server to the internet. Instead, use a VPN or a tunneling service like Cloudflare Zero Trust, ngrok, or Tailscale."

    docs.sillytavern.app, retrieved

  60. Receipt 60Official site

    Edit the config.yaml in the SillyTavern root folder, not /SillyTavern/default/config.yaml; when whitelisting, include 127.0.0.1 or you can't connect from the host machine.

    "Ensure 127.0.0.1 is included, or you will be unable to connect from the host machine."

    docs.sillytavern.app, retrieved

  61. Receipt 61Official site

    On Windows 11, enable the Private Network profile under Settings > Network and Internet > Ethernet, or remote devices can't connect even with firewall rules.

    "On Windows 11, enable the Private Network profile type in Settings > Network and Internet > Ethernet. This is VERY important for Windows 11"

    docs.sillytavern.app, retrieved

  62. Receipt 62Official site

    When connecting from the remote device's browser, use http:// not https://.

    "Use http:// NOT https://"

    docs.sillytavern.app, retrieved

  63. Receipt 63Official site

    In User Settings > Account, the Danger Zone has Reset Settings (restore all settings to factory defaults) and Reset Everything (complete account wipe and factory reset).

    "Reset Settings : Restore all settings to factory defaults Reset Everything : Complete account wipe and factory reset"

    docs.sillytavern.app, retrieved

  64. Receipt 64Official site

    The Account dialog also offers Settings Snapshots to back up and restore user settings, and Download Backup to export all user data.

    "Settings Snapshots : Create, manage, and restore backups of your user settings Download Backup : Export a complete backup of all your user data"

    docs.sillytavern.app, retrieved

  65. Receipt 65Official site

    Visual Novel mode lets you chat with characters shown as sprites, as in visual novels like Doki Doki Literature Club.

    "Chat to characters with sprites, like in visual novels such as Doki Doki Literature Club and other famous VN games."

    docs.sillytavern.app, retrieved

  66. Receipt 66Official site

    The Clean-Up tool permanently deletes files and can't be undone, so make a backup first if unsure.

    "The Clean-up tool will permanently delete files. This action cannot be undone!"

    docs.sillytavern.app, retrieved

  67. Receipt 67Official site

    To update on Linux/Termux/Mac, run git pull in the SillyTavern folder and start it again; on Windows, try UpdateAndStart.bat first.

    "First try using the UpdateAndStart.bat which is located in your SillyTavern installation base folder."

    docs.sillytavern.app, retrieved

  68. Receipt 68Official site

    The docs always recommend installing with git, since updating is just git pull; zip installs need a tedious manual process.

    "When you have installed via git clone, all you have to do to update is type git pull"

    docs.sillytavern.app, retrieved

  69. Receipt 69Official site

    A "Cannot find module" error on startup means a new npm package is required; run npm install in the SillyTavern folder (Start.bat and start.sh do it automatically).

    "This means that SillyTavern added a new npm package requirement. Run npm install in the SillyTavern directory to fix this."

    docs.sillytavern.app, retrieved

  70. Receipt 70Official site

    If changed system files block git pull, try moving or deleting the file, or use git pull --rebase --autostash.

    "Another solution is using git pull --rebase --autostash"

    docs.sillytavern.app, retrieved

FAQ

Questions people ask

What do people use SillyTavern for?

People use SillyTavern to chat and roleplay with AI characters through a model they connect themselves. SillyTavern's FAQ says it handles communication with AI models, lets you create character cards and import other people's, and keeps chat history so you can resume a chat, start a new one or review old ones. The project's README adds image generation, TTS voices, Visual Novel Mode and WorldInfo lorebooks. It also lists many supported APIs, including KoboldAI/CPP, Horde, NovelAI, Ooba, Tabby, OpenAI, OpenRouter, Claude and Mistral.

How do I reset SillyTavern settings?

Open User Settings, go to Account, and use the Danger Zone. Reset Settings restores all settings to factory defaults, and Reset Everything is a complete account wipe and factory reset. The same Account dialog has Settings Snapshots for backing up and restoring settings, and Download Backup for exporting all user data, so take one of those first.

Can I use SillyTavern on my phone?

On Android, yes: SillyTavern runs natively through Termux, though the docs say Termux installs aren't officially supported or guaranteed to work. Get Termux from F-Droid or its GitHub releases, not Google Play, whose version is no longer maintained. iPhones and iPads can't run the whole app. Instead, run SillyTavern on a computer on your home Wi-Fi with listen set to true in config.yaml, then open it in the phone's browser.

How do I set up SillyTavern for roleplay?

Start by connecting a model in API Connections and creating or importing a character, then adjust a few settings: - To stop the AI from writing your character's actions, the FAQ suggests an Author's Note such as "You are never allowed to dictate actions or speech for {{user}}." - To make the bot keep writing, send an empty message. - For short replies, raise Response Length. - For persistent lore, reviewers report that World Info lorebooks inject character, location and context snippets into the prompt when keywords come up.

Is SillyTavern safe?

SillyTavern itself keeps everything on your machine. The project's README says it offers no online or hosted services and doesn't programmatically track user data, and the docs say all data stays in the installation path in default standalone mode. The exposure comes from what you connect to it: - The FAQ says paid web-service AIs log everything on their servers. - Anyone holding your OpenAI key can use GPT at your expense. - The docs say never to port-forward the server to the internet.

Slop check: passedThis page went through our slop counter before publishing. How we check our own copy