Imbutus

The ultimate uncensored AI — pentesting (vibe hacking), plus image, video & voice generation.

Run powerful uncensored LLMs and generate uncensored media — all on dedicated GPU infrastructure. Pay per use, no subscriptions.

More than a model — an agentic API

  • CVE and exploit database refreshed every 6 hours — answers always reference current, real vulnerability data
  • Built-in knowledge of connected Kali Linux VPS environments and their tooling
  • Rent a Kali Linux VPS directly — run the agent from a real attack surface, not your local machine
  • New integrations and knowledge sources added over time — request features via the support ticket system

Powerful LLM models

Abliterated, uncensored language models served via an OpenAI/Anthropic-compatible API — each running on its own dedicated GPU:

  • huihui-ai/Huihui-Qwen3.5-9B-Claude-4.6-Opus-abliteratedcheap — Most affordable yet capable model. Great for saving costs and running autonomous agents.
  • huihui-ai/Huihui-Qwythos-9B-Claude-Mythos-5-1M-abliteratedcheap — Upgraded 9B — newest Claude distillation (Mythos-5), up to 1M-token context, image input, abliterated. Affordable and capable.
  • huihui-ai/Huihui-Qwen3.6-35B-A3B-Claude-4.7-Opus-abliteratedgeneral — Middle ground — smart and capable without breaking the bank.
  • huihui-ai/Huihui-Qwen3-Coder-Next-abliteratedcoding — Most powerful coding-specialized model. Use for programming and code generation.

Uncensored media generation

Uncensored image, video, and voice bundles through a ComfyUI visual workflow interface — each on its own dedicated GPU, run them at the same time:

Video

  • Sulphur 2Decensored

    Generates video clips from a text prompt, image, audio, or video with LTX Director 2.0. An uncensored version of LTX 2.3.

    An uncensored build of LTX 2.3 — its safety filters were removed, so it generates freely with no refusals.

  • SCAIL-2Natively uncensored

    Animates or replaces a character in a video — give it a reference image plus a driving motion video and it transfers the motion (people auto-masked, no manual rigging). Apache-2.0, built on Wan2.1 14B.

    The Wan2.1 base ships with no content filter — it generates any content out of the box, nothing was removed.

Image

  • FLUX.2 klein 9B (True V3, uncensored)Decensored

    Fast 9B FLUX.2 klein, uncensored — the "True V3" aesthetic fine-tune with an abliterated text encoder. Text-to-image, multi-reference editing, and RefControl depth/structure control, light enough for a single 24GB GPU.

    Its safety-tuned base is freed by the 'True V3' fine-tune plus an abliterated text encoder — generates freely with no refusals.

  • Qwen-Image-2512Fully uncensored

    Generates images from text with high prompt fidelity and strong text-in-image rendering.

    Trained on unrestricted data and run with an abliterated text encoder — no refusals and no quality loss on any content.

  • Qwen-Image-Edit-2511Fully uncensored

    Edits existing images by prompt — background swaps, object add/remove, restyling (up to 3 input images).

    Trained on unrestricted data and run with an abliterated text encoder — no refusals and no quality loss on any content.

  • Ideogram 4Partly freed

    Text-to-image with strong typography — great for legible text & design. Visually place text and elements at exact positions on a canvas (regional layout control).

    The abliterated encoder removes prompt refusals, but NSFW was filtered from the model's training data, so results in that area are still inconsistent.

  • Boogu-Image TurboMostly freed

    Fast 4-step text-to-image with strong photorealism and bilingual (English/Chinese) text rendering.

    The abliterated encoder removes prompt refusals, but the model has soft safety and results in that area can still be inconsistent.

  • Boogu-Image EditMostly freed

    Instruction-based image editing — describe the change in text to insert, replace, or restyle objects in an image.

    The abliterated encoder removes prompt refusals, but the model has soft safety and results in that area can still be inconsistent.

  • Krea-2 TurboFully uncensored

    Fast, photorealistic text-to-image at up to 2K resolution, with 9 selectable style LoRAs for different looks.

    Natively NSFW-capable and run with an abliterated text encoder — no refusals and no quality loss on any content.

Voice

  • CosyVoice 3No content filter

    Text-to-speech and voice cloning / conversion, plus voice-to-SRT and SRT-to-voice. A reference voice sample defines the output voice.

    A text-to-speech model with no content restrictions — it speaks whatever text you provide.