Free API for SillyTavern: Trial Credit and Setup

What free means here

If you are searching for a free api for sillytavern, the straight answer is that nobody can run a large language model for nothing indefinitely, and this service does not pretend otherwise. What you get is a free start. After you sign up with Google or with email and password, the API key page has a button that claims $0.50 of trial credit. It is one claim per person, it is valid for 7 days, and it needs no payment method.

That credit is real usage, not a demo that stops after three messages. SillyTavern talks to the API exactly as it would talk to any OpenAI-compatible backend, using the uncensored model, and every request draws from the balance at $0.15 per 1M input tokens and $0.50 per 1M output tokens. When the trial is gone you decide whether to continue. Top-ups start at $20, credit never expires, and there is no subscription to cancel.

The model is a single open-weight model run by an independent provider and tuned to answer without content refusals for lawful adult use. It is not an official API of any model vendor, and it is text only. SillyTavern's image and voice features need their own services; this endpoint does chat.

What $0.50 buys in a SillyTavern chat

SillyTavern resends a lot of text on every turn: the character description, your persona, any lorebook entries that fire, the system prompt and as much chat history as fits in your context setting. So the cost of a message is mostly the input side. Here are two worked examples at the real prices.

Scenario	Input tokens	Output tokens	Cost per message	Messages in $0.50
Lean chat, short history	6,000	250	$0.001025	about 487
Long chat, big lorebook	20,000	400	$0.0032	about 156

The math for the first row: 6,000 input tokens at $0.15 per 1M is $0.0009, and 250 output tokens at $0.50 per 1M is $0.000125. Add them and you get $0.001025. Divide $0.50 by that and you land at roughly 487 messages. For the second row, 20,000 input tokens cost $0.003 and 400 output tokens cost $0.0002, so $0.0032 per message and about 156 messages.

Scale it up and the numbers stay small. A thousand messages of the lean kind cost $1.025. A thousand of the heavy kind cost $3.20. The practical lesson is that history length drives your bill, not reply length. If a session gets expensive, shorten the context rather than the replies.

Failed or refused requests are not charged, so a connection error or a rejected request does not eat your balance. You can watch the balance in the X-Balance-USD response header if you test with curl, and the pricing page has the top-up tiers and bonuses.

Connect SillyTavern step by step

1. Get the key

Sign up on the API key page, claim the trial credit, and copy the key. Treat it like a password. If you ever paste it into a screenshot or a shared preset, replace it.

2. Test the endpoint outside SillyTavern

A thirty-second test saves a lot of guessing later. If this returns a reply, the key and URL are right:

Copycurl https://api.ttsapihub.com/v1/chat/completions \
  -H "Authorization: Bearer YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "uncensored",
    "messages": [{"role": "user", "content": "Reply with the word ready."}],
    "max_tokens": 10
  }'

3. Fill in the connection panel

Open the plug icon (API Connections) in SillyTavern. Set the API type to Chat Completion and the chat completion source to Custom (OpenAI-compatible). Menu names shift a little between releases, but the three fields are always the same:

• Custom Endpoint (Base URL): https://api.ttsapihub.com/v1
• Custom API Key: the key you copied
• Model ID: uncensored

Press Connect. A green status means SillyTavern could reach the endpoint. If you type the model ID by hand and the dropdown stays empty, that is fine; the field accepts the name directly. You can also check what the server lists with GET /v1/models.

Settings worth changing right away

SillyTavern's defaults are tuned for other backends, so a few changes pay off immediately.

Context size

The model has a 100,000-token window, shared between the prompt and the reply. You do not have to use it all. Many roleplay sessions run fine at 16,000 to 32,000 tokens of context, and the bill grows with every token you resend. Start at 16,000, and raise it only when the character forgets something you care about.

Max response length

This sets max_tokens. It is optional on the API side, but SillyTavern sends it, and it is your cap on rambling. 300 to 500 gives chat-length replies. For long-form story scenes, raise it.

Streaming

Turn streaming on. The API sends server-sent events, SillyTavern prints the reply as it arrives, and you can stop a bad generation early. One thing to remember: a streamed request keeps its concurrency slot until the stream ends, so do not leave a stalled generation running while you try another.

Sampling

Temperature, top_p, stop sequences, seed and the repetition penalties are passed through to the model. If the model feels repetitive, nudge the penalties up a little and change one value at a time. Changing five at once tells you nothing.

Limits you will notice in long sessions

A full key allows 8 concurrent requests and 300 requests per minute. A trial key allows 2 concurrent requests and 60 per minute. For one person in a single chat that is plenty. The place you can hit the ceiling is a group chat, where SillyTavern may generate several characters' replies in quick succession, or when you open two browser tabs and regenerate in both.

When you exceed a limit, the API answers with HTTP 429. The error code is rate_limited when you have sent too many requests per minute and concurrency when too many are in flight at once. On temporary overload you also get a 429 with a Retry-After: 5 header. In every case the fix is the same: wait a few seconds and press regenerate. Nothing is charged for the failed attempt.

The request body can be up to 8 MB, which is far beyond what a chat prompt needs, even with a 100,000-token context.

Troubleshooting a connection that will not work

Symptom	Likely cause	Fix
Status stays red after Connect	Base URL typed wrong, or the path appended twice	Use exactly the base URL above, ending in /v1, with no /chat/completions on the end
401 Unauthorized	Key missing, truncated or pasted with a space	Paste it again and check for leading or trailing spaces
Model not found	Wrong model name	Set the model ID to uncensored in lowercase
429 after a few swipes	Trial key concurrency of 2	Wait for running generations to finish, then retry
Reply cuts off mid-sentence	Max response length too low	Raise it, or ask for a continuation

If none of that explains it, run the curl test from step 2 again. When curl works and SillyTavern does not, the problem is in the SillyTavern fields. When curl fails too, check the key and your network. The quickstart docs list the request format in full.

Content, privacy and a hard limit

The model is tuned to stay in character for fiction and roleplay between adults, which is the reason many SillyTavern users look for a backend like this one. It is meant for lawful adult use. One rule has no exceptions: no sexual content involving minors, in any framing, and requests of that kind are not something the service exists to support.

On privacy, prompts are not used for training. That is a statement about this service; it does not cover what SillyTavern itself stores on your machine, so your chat logs and character files stay wherever you keep them.

When you are ready to move past the trial, the pricing page shows the top-up options. Any whole amount from $20 to $500 works, with +5% credit from $50, +10% from $100 and +15% from $200. Card (Visa, Mastercard, Apple Pay, Google Pay) and crypto (USDT on TRC20, USDC on Base) are both accepted.

More guides

• TTS API Base URL: What It Is and What to Enter
• TTS API Key in Python: Setup, Streaming, JSON Mode
• Elevenlabs API Checklist: Uncensored TTS Pipeline