SillyTavern vs Janitor AI: Which Fits You?
SillyTavern is software you run yourself and connect to a model of your choice. Janitor AI is a hosted website where you chat with characters in a browser. This article compares the two and shows where an API fits into each.

Updated
Key points
- SillyTavern is a frontend you install; the model behind it is your choice.
- Janitor AI is a hosted service. Check its current terms, limits and options on its own site.
- Our API is only a backend: an OpenAI-compatible endpoint with model
uncensored, no frontend included. - A 30,000-token roleplay message costs about $0.0047 at $0.15 / $0.50 per 1M tokens.
The short answer
In the SillyTavern vs Janitor AI question, the deciding factor is who runs what. SillyTavern is an open-source interface that you install on your own computer or server. It does not include a model, so you point it at a backend: a local model, or an API. Janitor AI is a website. You open it, pick or create a character, and chat, with the hosting handled for you.
Pick SillyTavern if you want control over prompts, lorebooks, sampling settings and where your data lives, and you do not mind a longer setup. Pick a hosted site if you want to start chatting in a minute and prefer not to maintain anything. Many people use both.
Either way, the language model is a separate decision. That is where an OpenAI-compatible API such as this one comes in.
Side-by-side comparison
| Question | SillyTavern | Janitor AI |
|---|---|---|
| What is it? | Self-hosted chat frontend | Hosted character chat website |
| Where does it run? | Your machine or your server | The provider's servers |
| Does it include a model? | No, you connect one | Offers its own options; check the site for current ones |
| Setup effort | Install Node, run the app, add an API connection | Create an account in the browser |
| Prompt and sampling control | Extensive, in the app | Check the site for what it exposes |
| Cost | The app is free; you pay for the backend you choose | Check their current pricing |
The cells that say "check" are deliberate. Hosted services change plans, limits and content rules, and a table here would go stale. Look at their current pages before you decide.
Where an API fits
A frontend and a backend are different layers. SillyTavern is the frontend. Our service is a backend: it takes a standard POST /v1/chat/completions request and returns text from a single open-weight model, run by an independent provider and tuned to answer without content refusals for lawful adult use. It has no characters, no gallery and no chat interface of its own.
That makes it a fit for SillyTavern and for any other client that accepts an OpenAI-compatible endpoint. If you use a hosted site that lets you enter a custom endpoint, the same base URL and key work there as well. Whether a given site offers that setting, and on which plans, is something to confirm on that site.
The model is not the official offering of any big lab, and it is text only. If you need images or voice, you will need other services alongside it.
Connect SillyTavern to the API
The steps below follow the current SillyTavern menu names. They can shift slightly between versions.
- Create an account and copy your key from the API key page.
- In SillyTavern, open the plug icon (API Connections).
- Set API to Chat Completion and the source to Custom (OpenAI-compatible).
- Enter
https://api.unrestrictedaimodels.com/v1as the custom endpoint (base URL). - Paste your key into the API key field.
- Type
uncensoredas the model ID. - Press Connect, then send a test message in any chat.
If it fails, test the key outside the app first:
curl https://api.unrestrictedaimodels.com/v1/models \
-H "Authorization: Bearer YOUR_KEY"
A list containing uncensored means the key and URL are right, and the problem is in the app settings. A /v1 typo or a trailing path is the most common cause.
Settings worth changing after you connect
The default values in a frontend are written for the average backend, so adjust a few of them once the connection works.
- Context size. Set it below the 100,000-token window, for example 32,000 or 64,000. SillyTavern trims older messages to fit, and a smaller number keeps each request cheaper.
- Response length. Cap the reply at a few hundred tokens for chat, and raise it for long scenes. Output tokens cost more than input tokens, so a long cap costs more if the model uses it.
- Temperature. Values around 0.8 to 1.0 suit fiction. Lower it if the character drifts or repeats facts wrongly.
- Streaming. Turn it on so words appear as they are written. The API sends tokens over SSE and reports usage at the end.
- Repetition penalties. The frequency and presence penalties are passed through. Raise them slightly if replies loop on the same phrases.
You can watch spending as you go: each response carries an X-Balance-USD header with your remaining credit. A full key allows 8 concurrent requests and 300 per minute, which is more than a single chat session will use.
What roleplay actually costs
Roleplay frontends send a lot of text on every message: the character card, the system prompt, lorebook entries and the recent chat. Say each request carries 30,000 tokens of prompt and the model writes 400 tokens.
- Input: 30,000 x $0.15 / 1,000,000 = $0.0045
- Output: 400 x $0.50 / 1,000,000 = $0.0002
- Total: $0.0047 per message, or $4.70 per 1,000 messages.
A $20 top-up would cover about 4,250 messages like that. Larger top-ups carry a bonus: +5% from $50, +10% from $100 and +15% from $200, so $100 gives $110 of credit. Credit does not expire and failed or refused requests are not charged. Details are on the pricing page. Payment is by crypto.
The context window is 100,000 tokens, so very long chats fit, but trimming old turns will keep your bill lower.
Choosing between them
Choose SillyTavern with an API backend when you want to own the setup, switch models without changing apps, and keep chat logs on your own disk. Expect to spend an evening learning its settings.
Choose a hosted site when convenience matters most and you accept its model options and rules. If you later want more control, you can move your character cards to SillyTavern, since both ecosystems commonly share cards.
This service is for lawful adult use, and sexual content involving minors is not allowed under any circumstances. If you want to test it with no commitment, new users can claim $0.50 of trial credit on the key page, valid for 7 days, no card needed. The documentation lists the full parameter set, including streaming and the temperature and penalty settings SillyTavern sends.
More guides
Questions and answers
Is SillyTavern better than Janitor AI?
They are different kinds of tools. SillyTavern gives you control and runs on your own machine, while Janitor AI is hosted and quicker to start. The better choice depends on how much setup you want.
Can I use this API with SillyTavern?
Yes. Choose Chat Completion, then Custom (OpenAI-compatible), and enter https://api.unrestrictedaimodels.com/v1, your key and the model ID uncensored.
Does the API include characters or a chat interface?
No. It is a text-only backend. You bring a frontend such as SillyTavern or write your own.
How much does a roleplay message cost?
With a 30,000-token prompt and a 400-token reply, about $0.0047 at $0.15 per 1M input and $0.50 per 1M output tokens.
Your key is one form away
Create an account, copy the key, change the base URL. That is the whole setup.
Get API key