Everything Uncens can do

01What is Uncens?

Uncens is an uncensored AI workspace: chat with models that don't lecture you, generate and edit images, build whole websites by describing them, send files and pictures, search the web - all under one dark roof. No app to install: everything runs in your browser at uncens.xyz.

02Quickstart

From zero to your first answer in under a minute:

1. 1Open uncens.xyz - the ask bar is right there. Type your question and hit send. No login needed to try it.
2. 2The bar flips into a full chat. Pick a model from the dropdown at the top - each shows its cost. Free models cost nothing while you explore.
3. 3Want history, premium models or the API? Sign up (email + password) - free accounts get daily credits to keep testing.
4. 4Attach files with the paperclip button (left of the composer): text files merge into your prompt, images become vision attachments.
5. 5Ask for a picture in plain words - “generate a red neon skull” - and the image model fires automatically, with a cost confirm for premium models.
6. 6Want a whole website? Ask for one and hit Start coding on the offer card that appears - the builder agent takes over (see Projects).

03FAQ

04Chat basics

The composer at the bottom is all you need: type, send. Each conversation keeps its own model choice; the sidebar lists your conversations and coding projects, lets you rename or delete them. Code in answers renders as proper widgets - filename header, line numbers, copy and download buttons.

05Choosing models

The model picker groups everything by class so you always know what you're getting:

06Files, images & vision

The paperclip button (left in the composer) uploads files. There is deliberately no "mode toggle" - what a file becomes depends on what it is:

07PDF editing

Attach a PDF and ask for structural changes in plain words - the assistant prepares an edit plan, shows it as a card, and the edited document downloads when you press Run & download. Everything executes in your browser against the stored original.

08Image generation & editing

There are no image "modes" - you ask in natural language. Saying something like “generate a picture of a red neon skull, cyberpunk style” routes your prompt to the image model automatically; premium image models show their cost for confirmation before running.

1. 1Type the request in normal words - “draw…”, “generate an image of…”, “make a picture of…”.
2. 2Confirm the cost if the model is premium (free-tier image models just run).
3. 3The picture appears inline, typically within 5-45 seconds. Regenerate for variations.

Editing: attach an image (paperclip or paste) and describe the change - “make the sky stormy red”, “put the character in a leather jacket”. Dedicated edit models handle it; edits bill the edit model's own price and failed generations or edits are always refunded automatically.

09Web search & research

Ask about something current - “what happened with X today”, “latest news on Y” - and the model can search the live web before answering, with sources baked into the reply. While it works, a live research panel shows each step: the query it ran, the pages it read, what it extracted.

10Voice input

Tap the microphone in the composer and speak - your words are transcribed privately, in your browser, via a local Whisper model. Nothing is uploaded anywhere. The first use offers a one-time model download; if your browser can't run it, the standard Web Speech engine is used as a fallback.

11Completion sounds

A small bell pill in the chat and project headers toggles completion sounds: a soft chime plays when an answer finishes, a build arpeggio when a project build completes, and a falling tone on errors. Handy for long generations - switch tabs and come back when you hear it. Sounds are on by default and the preference is remembered per browser.

12Ghost mode

Toggle ghost mode (the ghost icon next to the composer) and nothing about the conversation is ever stored: no conversation record, no messages. The chat lives only in your browser session. Rate limits still apply - ghost mode is privacy, not a free pass.

13Shortcuts & tips

14Build websites & apps

Projects turn a chat model into a website builder - describe what you want, watch it appear in the live preview, refine by talking, then download the files and host them anywhere. No build tools, no servers, no lock-in. Open the ⌗ coding tab in the chat sidebar (or the standalone Projects page), create a project and go.

1. 1Switch the sidebar to ⌗ coding and hit ＋ - or just ask for something buildable in a normal chat: “build me a landing page for X”. The assistant answers with an offer card - press Start coding and a project is created with the perfected prompt and the build starts automatically.
2. 2Describe what to build in the chat panel - “a dark landing page for a streetwear brand with a gallery and newsletter signup”. The agent plans, writes files one by one and streams progress into a step-by-step transcript.
3. 3Iterate in plain words - “make the header sticky”, “add a pricing section”, “fix that error”. Each turn rewrites only the files that change; the builder always sees the current state of every file.
4. 4Preview errors are fed back to the model automatically - if the generated code throws in the console, the next step includes the error so it can self-heal.
5. 5Edit any file yourself in the Code tab (save applies instantly), inspect the project's real database, then download the whole project as a ZIP for hosting anywhere - Netlify, GitHub Pages, any static host.

Projects are plain websites: HTML, CSS and JavaScript with libraries loaded from CDNs. They run fully client-side, so anything a browser can do works - canvas games, dashboards, API-driven tools (fetch), persistence via a real per-project SQLite database. The preview runs in a hardened sandbox inside your own browser; project code never executes on Uncens servers.

15The agent's toolbox

The builder is an agent with real tools, not a single prompt. Everything it does shows up as expandable cards in the transcript, so you can audit every step:

16Builder tips & limits

17Plans, tiers & the weekly bonus

Uncens has a free tier and paid tiers (Tier 1, Tier 2). Tiers unlock premium models, bigger monthly credit refills, unlimited use of many models, higher API rate limits and API access. Live prices and exactly what each tier includes are on the pricing page and manageable from your profile.

On top of the monthly refill, every paid tier gets a weekly plan bonus - an extra refillable pool of ◆ 50 (Tier 1) / ◆ 240 (Tier 2+) credits that lands every 7 days:

18Credits

Everything is priced in credits - ◆ 100 credits = $1. Chat models charge per million tokens by real usage (input and output separately), images and edits by the picture. Costs are only charged after a successful response; failures are refunded automatically.

19Gift codes

Redeem a gift code in your profile (wallet card → redeem) to instantly get its tier and/or credits. Codes stack: a tier code extends your plan pro-rata, credit codes add to your balance. One code, one redemption - after that it's spent.

20Privacy & data

21API overview

The Uncens API is an OpenAI-compatible HTTP API. Point any OpenAI SDK or tool at the base URL below, use one of your API keys as the Bearer token, and everything works - same models, same uncensored behaviour, same credit prices as the website. Agent-grade by design: function calling, vision input and streamed reasoning are first-class.

22Authentication

All requests authenticate with a Bearer API key. Create one in your profile → Settings → API keys - the key is shown exactly once, so copy it immediately. Keys start with uk_….

# always send the key as a Bearer token
Authorization: Bearer uk_xxxxxxxxxxxxxxxxxxxxxxxx

Lost a key or afraid it leaked? Revoke it in your profile and create a new one - revocation is effective immediately.

23API quickstart

The fastest path: create a key in your profile, then run the curl above with your key. Python and Node examples:

from openai import OpenAI

client = OpenAI(base_url="https://uncens.xyz/api/v1", api_key="uk_...")

resp = client.chat.completions.create(
    model="UncensRoleplay",
    messages=[{"role": "user", "content": "tell me about yourself"}],
)
print(resp.choices[0].message.content)

import OpenAI from "openai";

const client = new OpenAI({ baseURL: "https://uncens.xyz/api/v1", apiKey: "uk_..." });

const resp = await client.chat.completions.create({
  model: "UncensRoleplay",
  messages: [{ role: "user", content: "tell me about yourself" }],
});
console.log(resp.choices[0].message.content);

24Chat completions

POST /chat/completions - OpenAI-compatible. Sends up to the last 40 messages to the model and returns the reply. System messages are supported (content policies follow the website defaults).

{
  "id": "cmpl-...",
  "object": "chat.completion",
  "created": 1725000000,
  "model": "UncensRoleplay",
  "choices": [
    { "index": 0, "message": { "role": "assistant", "content": "..." }, "finish_reason": "stop" }
  ],
  "usage": { "prompt_tokens": 12, "completion_tokens": 87, "total_tokens": 99 }
}

25Vision (image input)

Vision-capable models accept OpenAI-style multipart content: text plus image_url parts with data: URLs (base64-encoded images). Send the picture inline with the question - no separate upload endpoint.

{
  "model": "UncensRoleplay",
  "messages": [{
    "role": "user",
    "content": [
      { "type": "text", "text": "what's in this picture?" },
      { "type": "image_url", "image_url": { "url": "data:image/png;base64,iVBORw0KGgo..." } }
    ]
  }]
}

26Function calling

Send OpenAI-format tools definitions and the model returns tool_calls you can execute and feed back - full agent loops work unchanged. Up to 64 tools per request; tool_choice is honored.

{
  "model": "UncensRoleplay",
  "messages": [{ "role": "user", "content": "weather in Berlin?" }],
  "tools": [{
    "type": "function",
    "function": {
      "name": "get_weather",
      "parameters": {
        "type": "object",
        "properties": { "city": { "type": "string" } },
        "required": ["city"]
      }
    }
  }],
  "tool_choice": "auto"
}

27Streaming & reasoning

Add "stream": true to receive the answer as SSE chunks - identical to OpenAI's streaming format, terminated by data: [DONE]. Thinking models additionally stream their reasoning in a delta.reasoning_content field, separate from the answer content.

curl https://uncens.xyz/api/v1/chat/completions \
  -H "Authorization: Bearer uk_..." \
  -H "Content-Type: application/json" \
  -d '{"model": "UncensRoleplay", "stream": true,
       "messages": [{"role": "user", "content": "count to five"}]}'

data: {"choices":[{"delta":{"reasoning_content":"let me count…"}}]}
data: {"choices":[{"delta":{"content":"1"}}]}
data: {"choices":[{"delta":{"content":", 2"}}]}
...
data: [DONE]

28Models & pricing

GET /models lists exactly what your key can use, including your live credit balance. The tables below are the live catalog - chat models bill per 1M tokens by actual usage, image models per picture.

Chat models - text, per 1k tokens

Image models - per generated picture

Image-edit models - per edit

curl https://uncens.xyz/api/v1/models -H "Authorization: Bearer uk_..."

29Image generation

POST /images/generations - OpenAI-compatible. Generates a picture from a text prompt with the named IMAGE model and returns base64 PNG bytes. Statelessness: nothing is stored on our side.

curl https://uncens.xyz/api/v1/images/generations \
  -H "Authorization: Bearer uk_..." \
  -H "Content-Type: application/json" \
  -d '{"model": "Uncens-MimiMax-H3-Text-to-Video", "prompt": "a red neon skull, cyberpunk"}'

{
  "created": 1725000000,
  "data": [{ "b64_json": "iVBORw0KGgo..." }],
  "credits_charged": 1.5
}

# decode and save
import base64
with open("out.png", "wb") as f:
    f.write(base64.b64decode(resp["data"][0]["b64_json"]))

30Image editing

POST /images/edits - edit an existing picture with natural language using a dedicated IMAGE_EDIT model. Give it any image and describe the change.

# base64 or https URL both work
curl https://uncens.xyz/api/v1/images/edits \
  -H "Authorization: Bearer uk_..." \
  -H "Content-Type: application/json" \
  -d '{"model": "Uncens-Image-Edit",
       "prompt": "make the sky stormy red",
       "image": "https://example.com/pic.png"}'

31Credits & billing

Everything is billed in credits - ◆ 100 credits equal $1. Chat models charge per 1M tokens (in and out separately) based on actual usage reported by the model, charged after the response completes - a failed request never costs anything. Image generation and edits charge their fixed per-image / per-edit price. On paid tiers the weekly bonus pool is consumed first, then the normal balance. Your balance and every model's exact price are in GET /models and your profile.

32Errors

{ "error": { "message": "...", "type": "invalid_api_key", "code": "invalid_api_key" } }

33Rate limits

34SDKs & tools

Any OpenAI-compatible client works - point it at the base URL with your key:

Base URL: https://uncens.xyz/api/v1
API key:  uk_... (profile → Settings → API keys)

Works with: OpenAI Python/Node SDK · Cursor · Continue · OpenWebUI ·
LiteLLM · langchain · aider · anything with a "custom OpenAI provider"