SillyTavern itself has no filter
It is a front end: it sends your character card and the chat to whichever API you connect. Refusals and toned-down replies come from the model behind that API, so choosing the model is the real decision.
SillyTavern · NSFW roleplay API
An uncensored SillyTavern API, with pictures and video
SillyTavern never filters your stories; the model behind it does. Connect it to SpicyAPI and you get 19 unrestricted chat models that take SillyTavern's default settings without errors, plus a free extension that turns a scene into a picture or a short clip, priced before it runs.
- Uncensored chat models
- 19
- Chat from
- $0.425per 1M output tokens
- Pictures from
- $0.006per image
- Video from
- $0.0112per second
The short answer
Three facts cover most of what people ask before they connect SillyTavern to an API.
19 unrestricted chat models
Their policy tier is unrestricted, so the model adds no refusal layer of its own, and each one has a published Freedom score from the same content tests. They connect through SillyTavern's own Custom (OpenAI-compatible) source.
The same key makes pictures and video
SillyTavern's built-in image generation cannot be pointed at an OpenAI-compatible picture API. The free SpicyAPI extension fills that gap with /spicy-img, /spicy-video and an Animate button on every picture in the chat.
Why SpicyAPI works well with SillyTavern
The things that usually go wrong with a custom endpoint, and what happens here instead.
No 400 errors from your sliders
SillyTavern sends temperature, top P and the frequency and presence penalties with every message. When a model does not use one of them, SpicyAPI skips it instead of rejecting the message and names it in the X-Spicy-Ignored-Params header; a value outside the model's range is set to its limit instead. Your presets keep working on every model.
Thinking stays out of the reply
Reasoning models send their thinking separately, so SillyTavern folds it into its collapsible thinking block and the reply itself stays in character.
Callable from the browser
The API answers cross-origin requests, so extensions that call it straight from your browser work without turning on SillyTavern's CORS proxy.
Freedom scores, not guesses
Every chat model carries a content policy tier and a Freedom score from the same standard tests, so you know how far a model goes before you spend anything.
Pay per token, in dollars
No subscription. You pay per token from a prepaid US dollar balance, each model page lists its input and output price, and a key can have a daily spending cap.
Results stay on your machine
The extension saves pictures and videos in SillyTavern's own image folder and hides them from the AI by default, so a text-only model is never sent a picture it cannot read.
Connect SillyTavern in three steps
Checked with SillyTavern 1.19.0. Steps one and two are all you need for chat; step three adds pictures and video.
- 1
Point SillyTavern at SpicyAPI
Create a key in your console, ideally a separate one with a daily spending cap. In SillyTavern, open API Connections, set API to Chat Completion and Chat Completion Source to Custom (OpenAI-compatible). Enter the endpoint below with no trailing slash, paste the key and press Connect.
Endpointhttps://api.spicyapi.ai/v1
- 2
Pick a chat model and start the scene
Type a model ID that ends in /chat into Enter a Model ID. The dropdown lists every SpicyAPI model, with the chat models first. Turn on Streaming, set Response (tokens) to 2000 or more, choose a character and write your first line.
- 3
Add the extension for pictures and video
Open Extensions, choose Install extension and paste the repository address below. Then open the SpicyAPI section of the Extensions panel, paste your key and press Check connection. Its Connect chat button can also do step one for you.
Extension repositoryhttps://github.com/spicyapi-ai/sillytavern-spicyapi
Settings that matter
Everything else can stay at SillyTavern's defaults.
| Setting | Value | Why |
|---|---|---|
| Streaming | On | Long replies appear as they are written, and a failed streamed request shows SpicyAPI's own error message instead of a bare Bad Request. |
| Response (tokens) | 2000 or more | Reasoning models count their thinking against this limit. At the default of 300 they can spend it all on thinking and return nothing. |
| Request model reasoning | On | Puts the model's thinking in a collapsed block whenever the model sends it. |
| Send inline media | Off for text-only models | When it is on, pictures in the chat travel with every later message, and a model that cannot read pictures rejects them. |
| Prompt Post-Processing | None | Switch to Semi-strict only if an error mentions message roles. |
The SpicyAPI extension: pictures and video from the chat
Free and open source under the MIT license. Pictures and videos are billed to your SpicyAPI balance; the prompt can come from you or from whatever chat model SillyTavern is connected to. If you switch it on, the chat model can also ask for a picture by itself, within a cooldown and a daily cap you set.
Animate any picture
Every picture in the chat gets a film button. It turns the picture into a short clip with your image-to-video model and adds the clip to the same message. The chat model can write the motion prompt, too.
Juno Vale on the starship bridge as the camera slowly pushes in and a violet nebula turns beyond the viewport
Seedance 1.5 Pro · 720p, 5 s, no audio · $0.1135
Priced before it runs
Run /spicy-img without a prompt and an editor opens with the model and the exact price while your chat model writes a prompt from the current scene. Edit it or write your own; nothing is charged until you press Generate, and a per-run ceiling refuses anything above your limit.
Z-Image Turbo · 16:9 · $0.01
Every model, with its price
The pickers list every picture and video model with its starting price, grouped by content policy: Uncensored, Softened and Filtered. Leave Automatic to use the cheapest uncensored model; settings marked $ change the price.
Prices as listed on 2026-10-07; the extension reads the live catalogue
Commands
| Command | What it does |
|---|---|
/spicy-img [prompt] | Makes a picture and posts it in the chat. Without a prompt, the chat model writes one from the scene. |
/spicy-video [prompt] | Makes a video from text. |
/spicy-video image=last [prompt] | Animates the latest picture in the chat. |
/spicy-task <task id> | Adds the result of an earlier run, for example after a page reload. |
/spicy-balance | Returns your balance in US dollars. |
A run that has been submitted keeps going even if you close the page, and it is billed if it succeeds. Failed runs are normally not charged.
Uncensored chat models for SillyTavern
Every chat model whose policy is unrestricted, live from our catalogue and sorted by Freedom score. Copy the ID into Enter a Model ID.
| Model | Model ID | Freedom | Inputper 1M tokens | Outputper 1M tokens |
|---|---|---|---|---|
| Grok 4.7 | xai/grok-4.7/chat | 100 | $2 | $6 |
| Grok 4.6 | xai/grok-4.6/chat | 100 | $2 | $6 |
| Grok 4.5 | xai/grok-4.5/chat | 100 | $2 | $6 |
| Grok 4.3 | xai/grok-4.3/chat | 98.9 | $1.25 | $2.50 |
| DeepSeek V4.1 Flash | deepseek/v4.1-flash/chat | 90.4 | $0.30 | $1.20 |
| Claude Opus 5.5 | anthropic/claude-opus-5.5/chat | 88.9 | $3.60 | $18 |
| DeepSeek V4 Flash | deepseek/v4-flash/chat | 88.9 | $0.44 | $1.32 |
| DeepSeek V4 Pro | deepseek/v4-pro/chat | 88.6 | $1.32 | $3.96 |
| Claude Sonnet 5.5 | anthropic/claude-sonnet-5.5/chat | 88.3 | $2 | $10 |
| Gemini 2.5 Flash | google/gemini-2.5-flash/chat | 84.4 | $0.285 | $2.375 |
| Claude Opus 4.6 | anthropic/claude-opus-4.6/chat | 79.4 | $5 | $25 |
| Claude Haiku 5.5 | anthropic/claude-haiku-5.5/chat | 77.8 | $0.10 | $0.50 |
| Claude Opus 4.5 | anthropic/claude-opus-4.5/chat | 75.6 | $4.50 | $22.50 |
| Kimi K3 | moonshot/kimi-k3/chat | 74.4 | $2.70 | $13.50 |
| GLM 5.2 | zai/glm-5.2/chat | 67.8 | $0.98 | $3.08 |
| Gemini 3.1 Pro Preview | google/gemini-3.1-pro-preview/chat | 67.2 | $1.90 | $11.40 |
| Claude Opus 5 | anthropic/claude-opus-5/chat | 65 | $4.50 | $22.50 |
| GLM 5.3 Flash | zai/glm-5.3-flash/chat | 64.4 | $0.1275 | $0.425 |
| Claude Sonnet 4.6 | anthropic/claude-sonnet-4.6/chat | 53.9 | $3 | $15 |
Unrestricted means the model adds no refusal layer of its own. The Freedom score is how far a model goes in our standard content tests. How the Freedom score is measured
Other roleplay apps
The same key and model IDs work in apps that take an OpenAI-compatible address. The docs list the exact fields for each one.
RisuAI
Address
https://api.spicyapi.ai/v1Choose the Custom API model and enter the address, key and model ID. On the website, requests go through RisuAI's server.
Setup stepsAgnai
Address
https://api.spicyapi.ai/v1Add an OpenAI-compatible provider, then pick the model in a preset from the list it fetches. Requests go through Agnai's server.
Setup stepsChub
Address
https://api.spicyapi.ai/v1Add the address as a reverse proxy in Secrets, then pick the model from the fetched list. Requests go straight from your browser.
Setup stepsKoboldAI Lite
Address
https://api.spicyapi.ai/v1Use its OpenAI connection, enter the key and fetch the model list. Requests go straight from your browser.
Setup stepsJanitor AI
Not verifiedAddress
https://api.spicyapi.ai/v1/chat/completionsJanitor needs the full path. Nobody on our side has run a real chat through it yet, and its page may block addresses other than OpenAI's, so treat it as something to try.
Setup steps
The ground rules
We do not add a content filter of our own. These are the limits that still apply.
Adults only
You must be 18 or older, and every character must be an adult, whatever the art style or backstory.
Real people need consent
Do not create sexual or intimate scenes, pictures or videos of a real person without their permission. Look-alikes of public figures count.
The law and our policy apply
Content that is illegal where you or the subject live stays off-limits, as does anything our Acceptable Use Policy forbids.
Models keep their own rules
Each model's policy tier is on its page. Filtered models may refuse and softened ones may tone a result down. A refused picture or video is not charged unless the model page says otherwise.
Your key and files
The extension keeps your key in SillyTavern's settings file in plain text, where other extensions can read it, so use a separate key with a daily cap. Our copies of generated files are deleted after up to 14 days.
Questions people ask
No. SillyTavern applies no content filter of its own; it sends your character card and chat to the API you connect. Refusals come from the model behind that API, so pick an unrestricted model if you do not want them.
One whose models do not refuse ordinary story requests and that accepts SillyTavern's settings. The table on this page lists every unrestricted SpicyAPI chat model with its Freedom score, and the LLM leaderboard compares their quality, speed and price.
Usually because the endpoint rejects a setting SillyTavern always sends, such as the frequency or presence penalty. SpicyAPI skips sampling settings a model does not use instead. A few settings are still rejected: Logit Bias and more than one swipe per generation. Stopping strings and out-of-range slider values are adjusted to what the model takes instead. Turn on Streaming to see the full error, which names the field.
Reasoning models count their thinking against Response (tokens). At SillyTavern's default of 300 they can spend the whole budget thinking, and SpicyAPI returns an error saying the output budget was spent on reasoning. Set it to 2000 or more.
Yes, with the SpicyAPI extension. /spicy-img makes a picture of the current scene, and the Animate button turns any picture in the chat into a short clip. With picture models that accept a reference, it can also use the character's avatar.
No. The extension downloads each result and saves it in SillyTavern's own image folder, so the chat only shows files from your machine.
RisuAI, Agnai, Chub and KoboldAI Lite take the same address and key; the docs list the fields for each. Janitor AI is not verified: it needs the full address, and its page may block it. The picture and video extension is for SillyTavern only.
Chat is billed per token and pictures and videos per run, from a prepaid US dollar balance, with no subscription. For this page, a picture cost $0.01 and a 5-second 720p clip cost $0.1135. Every picture and video is quoted before it runs.
We do not review content. Prompts are erased after 30 days and generated files after up to 14 days. SillyTavern keeps your chats on your own machine; some web apps, such as Agnai, send your key and chat through their own servers.
Give your next scene an uncensored model
A couple of minutes for chat and one more for pictures and video. You pay per use, in dollars, and every picture and clip is priced before it runs.