Locally Uncensored

Generate anything — text, images, video. Locally. Uncensored.

No cloud. No data collection. No API keys. Auto-detects 12 local backends. Your AI, your rules.

The only desktop app that runs AI chat, image, and video generation — locally, one click, no cloud.

Download · Features · Quick Start · Why This App? · Roadmap

Screenshots

Chat with Personas	Image / Video Generation

Model Manager	Create View with Parameters

v2.5.3 — Current Release

A maintenance and features patch on top of v2.5.0. Your chats, memories, and gallery carry over automatically, and auto-update prompts on next launch.

New: edit your custom personas, a civitai.red mirror toggle for regions where civitai.com is blocked, and an external TTS engine option — point LU at any OpenAI-compatible endpoint like Kokoro.

Fixed: LM Studio and Lemonade are reachable across your LAN, the Windows 11 "agent error" mid-chat, the app always showing a window after a bad state, multi-LoRA in the agent, FLUX on current ComfyUI, a clearer hint for the Windows "paging file too small" error, and a batch of smaller issues.

v2.5.0 — The big one

The biggest update since launch. v2.5.0 isn't about new app categories — chat, image & video generation, the coding agent and the model manager were already here — it's a deep upgrade to all of them, plus several genuinely new tools. Everything still runs 100% on your own machine, and your chats, models, and settings carry over automatically.

Auto-update prompts on next launch.

🤖 A much smarter Coding Agent

The coding agent got a full overhaul: a plan-first Architect mode, whole-codebase awareness (Repo-Map, Aider-style PageRank), review-before-apply diffs, a read-only Code-Review mode, a test-runner loop, background jobs that survive a restart, typed Git/GitHub tools, multi-repo workspaces, and per-project .lurules. Iteration caps raised to 200 / 400.

🪶 Small-Model Mode (new)

One switch tunes the app so compact 3B–8B local models follow instructions and call tools far more reliably (off by default). Plus new sub-4GB uncensored tool-callers and weight-class filters in Discover so you find models that fit your machine.

🎚️ Real context control (new)

Set the context window per model and watch true token usage. Before v2.5.0, LU never forwarded num_ctx to Ollama, so it silently capped at 2048 and clipped long chats + RAG — now it honours your setting. Plus one-click load/unload of any local model from memory.

🎙️ Voice that actually works

Talk-to-type with live, on-device transcription (nothing is ever sent automatically) and read-aloud in a natural neural voice, both set up with one-click installers. The mic was effectively dead before this release; now it's solid.

🎨 Fresh design & everyday UX

A calmer theme with floating panels and a one-click light/dark toggle, your own profile picture, a per-message action bar (copy / regenerate / edit / read-aloud), an editable memory brain icon, and unlimited chat history — chats & memories moved to a local database, so long, image-heavy threads no longer get cut off.

🖼️ Create-tab upgrades

On top of the existing image/video generation: a LoRA picker, VAE override and CLIP-skip, a model picker right next to Generate, a Text-to-Video / Image-to-Video switch on the main screen, generated images shown inline in chat, and a rebuilt flicker-free media viewer.

📦 Model Manager & hardware

A refreshed layout with weight-class filters, resumable multi-part (sharded) downloads with a size confirmation, per-row LM Studio load/unload, a fixed INSTALLED badge, and a new GPU picker (Settings → Hardware) to pin which card runs your models on multi-GPU rigs.

🧰 New tools & reliability

Import your chat history — bring ChatGPT / Claude / Gemini exports into your local knowledge base.
Troubleshoot panel — per-backend health (Ollama / ComfyUI / LM Studio) plus system facts.
Reliability — your chats & settings now survive restarts and updates (a major fix), no more stuck "ComfyUI starting…", a phantom GPU entry is gone, credentials are redacted from logs, and console.log is stripped from production builds.
Plus reporter-driven fixes (context window, download routing + badges, Hermes 3 pulls, backend persistence).

☁️ Optional cloud waitlist

A small, dismissible badge to be notified if a hosted version ever arrives — it only sends an email address you choose to type in. Everything else stays on your device.

Stability

Frontend test suite: 2,725 tests green. Rust test suite + production build green. No breaking changes — settings auto-default and your data migrates forward.

For previous release notes (v2.4.9, v2.4.8, v2.4.6, v2.4.5), see CHANGELOG.md.

Thanks to the community reporters and contributors whose feedback shaped this release: cinemazverev, vanja-san, THobbs23, kj103x, Aldrich Ironhart, leonsk29, BobbyT, MikeS++, and many more in Discord and GitHub.

v2.4.9 — superseded

Five bug fixes + two leonsk29 feature requests on top of v2.4.8. All user-reported (GH levoy1 / kj103x Discord / nightmare13740 Discord / leonsk29 Discord + GH).

What's fixed

ComfyUI Desktop App is detected and accepted in onboarding (Bug U — levoy1 GH #47). Pre-v2.4.9 both auto-detect and manual path entry rejected the %LOCALAPPDATA%\Programs\ComfyUI binary folder with "Invalid path — main.py not found". v2.4.9 walks a probe list when given a folder with ComfyUI.exe until it finds the actual Working Directory with main.py.
Chats with attached RAG documents persist across NSIS auto-update / WebView2 data reset (Bug V/a — kj103x Discord 2026-05-23). Separate 30 s / beforeunload IndexedDB backup → %APPDATA%\Locally Uncensored\rag_chunks_backup.json restores chunks on cold start and on warm starts where IndexedDB is empty.
LU-spawned Ollama is killed when you quit LU (Bug V/b — same kj103x report). ollama serve no longer lingers as a ~200 MB orphan after a tray-menu Quit. User-managed Ollama running pre-launch is still left alone.
Benchmark per-model display shows the latest session's tok/s, not a drifting historical average (Bug W — nightmare13740 Discord 2026-05-23/24). New getLatestSpeed groups runs into sessions via a 10 s timestamp gap; leaderboard keeps the cross-model average.
Thinking + Agent toggles light up for community-uncensored Gemma 4 builds (Bug X — leonsk29 Discord 2026-05-24). normalizeFamily rewritten as a greedy strip + suffix/quant peel + family-dash collapse /g, plus word-boundary contains-check.

What's new

Onboarding suggests nomic-embed-text for Document Chat / RAG (GH #45 — leonsk29). Auto-detects existing embedding models or offers a one-click 274 MB pull.
VRAM filter for text models in Discover (GH #46 — leonsk29). Lightweight ≤10 GB / Mid-Range 10–16 GB / High-End >16 GB chips on the text tab.

v2.4.8 — superseded

Hotfix: 8 fixes on top of v2.4.6. All user-reported. v2.4.7 was tagged but not separately released; its six fixes ship together with v2.4.8.

What's fixed

Text models in Discover keep their INSTALLED badge after restart (Bug S — leonsk29 GH #43). Pre-v2.4.8 the badge only lit up if the download finished in the current session, so a model installed yesterday looked uninstalled today. v2.4.8 also matches against the provider model list (Ollama tags directly, plus GGUF downloads via hf.co/<repo>:<quant> references), so whatever Ollama / LM Studio actually have on disk is what the Discover grid shows.
canPull:false text models now get a clickable HuggingFace link (Bug T — leonsk29 GH #44). Qwen 3.6 27B Samantha and GLM 5.1 754B MoE both have HF pages but no GGUF on day-one, so they show up as Available rather than downloadable. Before this they had no UI to open the HF page from inside LU. v2.4.8 adds the external-link button next to the Available badge.
LM Studio server-off banner is dismiss-able and matches the dropdown chrome (Bug Q UX polish). The v2.4.7 banner that surfaces "Start LM Studio Server" in the model picker had amber styling that clashed with the rest of the dropdown. v2.4.8 switches to neutral white/gray and adds an X to dismiss. Dismiss sticks across dropdown re-opens within the same LU session but is not persisted, so the hint resurfaces on the next launch in case the user forgot to start the server.
Benchmark tok/s now matches actual chat throughput (Bug M — nightmare13740 Discord). Pre-v2.4.7 the benchmark counted time-to-first-token + stream init in the tok/s denominator, so any local model looked slower in the Benchmark tab than in real chat. On nightmare13740's RTX 4070 Laptop 8 GB + gemma4:e4b: ollama CLI 30 tok/s, manual chat 23–25 tok/s, pre-fix benchmark 12 tok/s. v2.4.8 starts the clock at first-token-received and (for Ollama) prefers the server-reported eval_count / eval_duration from the final stream chunk, with a wall-clock fallback for OpenAI-compat providers when streaming gets buffered. End result: benchmark matches CLI / chat baselines on every provider.
Windows ComfyUI install probes git --version before clone (Bug N — juliandiggins-stack GH #40). If a WSL / Linux-mounted git is first on PATH, the previous code let the clone start and die mid-flight with cryptic stderr. New tri-state probe (Native / NonNative / Missing) either proceeds silently, surfaces a clear "install Git for Windows" hint when missing, or logs a soft warning when a non-native git might still work.
Anthropic custom-proxy provider no longer double-prefixes /v1 (Bug O — 0yagizz Discord). Users pointing the Anthropic provider at a proxy (claude-relay-server, LiteLLM, opencode-zen) whose docs end the baseUrl with /v1 got a silent 404 on …/v1/v1/messages. v2.4.8 collapses the duplicate.
Image / video generation timeouts now configurable in Settings (Bug P — ake0n_official Discord). CPU-only / iGPU users finished sampling mid-run when the previous hard-coded 20-min image cap hit. v2.4.8 surfaces both image and video timeouts as numeric inputs (defaults stay 20 min / 60 min, range 1–480 min).
Chat model picker surfaces "Start LM Studio Server" inline when LM Studio is installed but its server is off (Bug Q — wakeywakeynow GH #41). Symptom was "i am running lm studio but i can't choose any models i have installed!!!!" — root cause: LM Studio's HTTP server doesn't auto-start with the app, so the user has models on disk but the model picker silently dropped LM Studio. v2.4.8 detects the installed-but-off state and renders an inline banner in the picker with a working Start Server button; clicking it kicks off the LM Studio server via the existing start_lmstudio_server Tauri command and re-populates the model list without restarting LU.
Custom ComfyUI save nodes' outputs now surface in LU's gallery (Bug R — silentrunningcaUSA GH Discussion #6). Pre-v2.4.8 LU only scraped images / gifs / videos from the history payload, missing every community workflow that uses a non-canonical save node (audio, custom metadata, CivitAI flows). v2.4.8 generic extractor accepts any keyed array of file-shaped objects.

Stability

vitest: 2306 tests green.
cargo test --release: 100 passed + 1 ignored (incl. live git_probe host probe).
tsc --noEmit: clean. cargo check: clean (pre-existing dead-code warnings only).
No breaking changes, settings auto-migrate with safe defaults.

Heads-up

v2.4.8 is a Windows + Linux release; macOS is not part of this build. #bug-reports / #help-* / GitHub will be monitored daily for regression reports.

Still investigating: OpenRouter half of 0yagizz's report (needs F12 console output to repro).

For older releases, see CHANGELOG.md.

Why Locally Uncensored?

Feature	Locally Uncensored	Open WebUI	LM Studio	SillyTavern
AI Chat	Yes	Yes	Yes	Yes
Coding Agent (Codex)	Yes	No	No	No
28 Agent Tools + MCP	Yes	No	No	No
Plug & Play Setup	12 Backends	No	Built-in	No
Multi-Provider (20+ Presets)	Yes	Yes	Yes	No
A/B Model Compare	Yes	No	No	No
Local Benchmark	Yes	No	No	No
Image Generation	Yes	No	No	No
Image-to-Image	Yes	No	No	No
Image-to-Video	Yes	No	No	No
Video Generation	Yes	No	No	No
File Upload + Vision	Yes	Yes	Yes	No
Thinking Mode	Yes	No	No	No
Granular Permissions	7 Categories	No	No	No
Uncensored by Default	Yes	No	No	Partial
Memory System	Yes	Plugin	No	No
Agent Workflows	Yes	No	No	No
Document Chat (RAG)	Yes	Yes	No	No
Voice (STT + TTS)	Yes	Partial	No	No
Remote Access (Phone)	Yes	No	No	No
Plugins (Caveman + Personas)	Yes	No	No	Yes
Auto-Update	Yes	No	Yes	No
Open Source	AGPL-3.0	MIT	No	AGPL
No Docker	Yes	No	Yes	Yes

Features

Core

Plug & Play Setup — First-launch wizard auto-detects 12 local backends. Nothing installed? One-click in-app Ollama download and install with progress bar. ComfyUI one-click install with step-by-step progress. Configurable ComfyUI port and path in Settings. Zero config needed.
Uncensored AI Chat — Abliterated models with zero restrictions. Streaming + thinking display.
Multi-Provider — 20+ presets. Local: Ollama, LM Studio, vLLM, KoboldCpp, llama.cpp, LocalAI, Jan, TabbyAPI, GPT4All, Aphrodite, SGLang, TGI. Cloud: OpenAI, Anthropic, OpenRouter, Groq, Together, DeepSeek, Mistral. Switch per conversation.
Codex Coding Agent — Live streaming between tool calls, continue capability, AUTONOMY CONTRACT. File tree, folder picker, up to 200 iterations / 400 tool calls. v2.5.0 adds Architect/Editor split, Repo-Map (Aider PageRank), Multi-File Stage-and-Approve, Test-Driven Loop, typed git + gh tools, Code-Review mode, long-running background shell tasks, per-repo .lurules, pr_resume, project_init, parallel sub-agents.
Agent Mode — 28 tools + MCP: web search/fetch, file I/O, shell, code execution, screenshots, system info, time, typed git/gh, background shell tasks, run_tests, project_init. Parallel execution, sub-agents, budget system, multi-repo workspace switching.
Remote Access — Access your AI from your phone via LAN or Cloudflare Tunnel. Full mobile web app with Agent Mode, Codex, plugins, file attach.
Image Generation — FLUX 2 Klein, FLUX.1 (schnell/dev), Z-Image Turbo/Base, Juggernaut XL, RealVisXL, DreamShaper XL via ComfyUI. Full parameter control, no content filter.
Image-to-Image — Upload a source image, adjust denoise strength, transform with any image model.
Video Generation — Wan 2.1, HunyuanVideo 1.5, LTX 2.3, AnimateDiff Lightning, CogVideoX, FramePack F1 on your GPU.
Image-to-Video — FramePack F1 (6 GB VRAM), CogVideoX 5B, SVD-XT. Upload an image, get video.

Intelligence

Thinking Mode — Provider-agnostic. See the AI's reasoning before the answer. Toggle from chat input.
File Upload + Vision — Drag & drop, paste, clip button. Vision models analyze images.
Granular Permissions — 7 tool categories, 3 permission levels, per-conversation overrides.
Smart Tool Selection — Reduces tool definitions per request by ~80%. JSON repair for local LLMs.
Memory System — Persistent across conversations. Auto-extraction. Export/import.
Agent Workflows — Multi-step chains. 3 built-in (Research, Summarize URL, Code Review). Visual builder.

Productivity

Model A/B Compare — Same prompt, two models, side by side. Parallel streaming.
Local Benchmark — One-click benchmark any model. Tokens/sec leaderboard.
Document Chat (RAG) — Upload PDFs, DOCX, TXT. Hybrid search with source citations.
Voice Chat — Push-to-talk STT + sentence-level TTS streaming.
20+ Personas — Pre-built characters. Switch without prompt engineering.
Chat Export — Markdown or JSON. Token counter. Keyboard shortcuts.

Customization

Plugins Dropdown — Caveman Mode (Off/Lite/Full/Ultra for terse responses) + 20+ Personas in one menu. Per-chat. Works in Chat, Agent, Codex.
Auto-Update — Signed NSIS installer. In-app download with progress bar. User-controlled restart (no forced updates). Settings survive updates.

Polish

Standalone Desktop App — Tauri v2 Rust backend. Download .exe, run it.
Model Load/Unload — iOS-style toggle in header. Load into VRAM, unload when done.
AE-Style Header — Clean typography navigation. Models, Settings, Downloads at a glance.
Privacy First — Zero tracking, all API calls proxied locally. ComfyUI process auto-killed on app close.

Tech Stack

Desktop: Tauri v2 (Rust backend, standalone .exe)
Frontend: React 19, TypeScript, Tailwind CSS 4, Framer Motion
State: Zustand with localStorage persistence
AI Backend: 20+ providers (Ollama, LM Studio, vLLM, KoboldCpp, llama.cpp, LocalAI, Jan, OpenAI, Anthropic, OpenRouter, Groq, and more), ComfyUI, faster-whisper
Build: Vite 8 (dev), Tauri CLI (production)

Download

Windows

Download the installer from Releases:

.exe — NSIS installer (recommended)
.msi — Windows Installer

Other platforms: The source code builds on Linux and macOS via npm run tauri build, but only Windows is officially tested and supported.

Plug & Play: Just install and launch. The setup wizard auto-detects all 12 supported local backends (Ollama, LM Studio, vLLM, KoboldCpp, llama.cpp, LocalAI, Jan, GPT4All, text-generation-webui, TabbyAPI, Aphrodite, SGLang). Nothing installed yet? The wizard shows one-click install links for every backend.

Antivirus warning? Some engines (ESET, Avast, Microsoft SmartScreen) flag the installer as suspicious — this is a false positive caused by heuristics on unsigned NSIS installers that download other binaries. The installer is built by GitHub Actions from public source on master (.github/workflows/release.yml). The auto-update channel is signed against a public minisign key. Full context, verification steps, and one-click vendor submission links: see SECURITY.md.

Quick Start

New to Locally Uncensored? Read the Getting Started Guide with screenshots for every step.

From Source

git clone https://github.com/PurpleDoubleD/locally-uncensored.git
cd locally-uncensored
npm install
npm run dev

For Contributors — Dev-Mode Setup

⚠️ Just want to use the app? Grab the installer from Releases (the .exe or .msi in the Download section above). That gives you the full Tauri desktop app with auto-update. The commands below start LU in browser dev-mode — fewer features, Vite proxy noise, meant for contributing to the codebase.

git clone https://github.com/PurpleDoubleD/locally-uncensored.git
cd locally-uncensored
setup.bat   # Windows — installs Node, Git, Ollama, then npm run dev
# setup.sh  # macOS / Linux equivalent

Launches at http://localhost:5173 in your default browser.

Image & Video Generation

Open the Create tab. ComfyUI is auto-detected or one-click installed. Models download with one click. Workflow is set to Auto — just write a prompt and hit Generate.

Recommended Models

Text (any local backend)

Model	VRAM	Best For
Qwen 3.6 35B MoE	24 GB	Vision + agentic coding + thinking. Brand new.
GLM-4.7-Flash IQ2	12 GB	Strongest 30B class. Tool calling. 198K context.
Gemma 4 E4B	4 GB	Lightweight, fast, great for small GPUs.
Qwen 3.5 35B MoE	16 GB	Best agentic, 256K context. SWE-bench leader.
Gemma 4 31B	16 GB	Frontier dense model, native tools + vision.
Hermes 3 8B	6 GB	Agent Mode. Uncensored + tool calling.
DeepSeek R1 (8B-70B)	6-48 GB	Chain-of-thought reasoning.

Image (ComfyUI)

Model	VRAM	Notes
FLUX.1 Schnell / Dev	8-10 GB	Best text-to-image. Fast (schnell) or quality (dev).
FLUX 2 Klein 4B	8-10 GB	Next-gen, fastest FLUX model.
ERNIE-Image Turbo	24 GB	Baidu DiT, 8 steps, 1024x1024. New.
Z-Image Turbo	10-16 GB	Uncensored, 8-15 sec per image.
Juggernaut XL V9	6 GB	Best photorealistic SDXL.

Video (ComfyUI)

Model	VRAM	Notes
Wan 2.1 T2V 1.3B	8-10 GB	Fast entry point, 480p.
Wan 2.1 T2V 14B	12+ GB	High quality, 720p.
FramePack F1 (I2V)	6 GB	Image-to-video, revolutionary low VRAM.
AnimateDiff Lightning	6-8 GB	Ultra-fast 4-step animation.
HunyuanVideo 1.5	12+ GB	Excellent temporal consistency.

Roadmap

Plug & Play Setup (auto-detect 12 local backends, one-click install links)
Codex Coding Agent
MCP Tool Registry (13 tools)
Granular Permissions (7 categories)
File Upload + Vision
Thinking Mode (provider-agnostic)
Model Load/Unload from header
Multi-Provider (20+ presets)
Agent Mode + Workflows
Memory System
A/B Compare + Local Benchmark
RAG / Document Chat
Voice Chat (STT + TTS)
ComfyUI Plug & Play (auto-detect, one-click install)
20 Image + Video Model Bundles
Image-to-Image (I2I)
Image-to-Video (I2V) — FramePack, CogVideoX, SVD
Z-Image + FLUX 2 + ERNIE-Image support
Dynamic Workflow Builder (15 strategies)
VRAM-Aware Model Filtering
Think Mode in Chat Input
Remote Access (LAN + Cloudflare Tunnel)
Mobile Web App (Agent, Codex, Plugins, Thinking)
Codex Streaming + Continue + Autonomy Contract
Agent 13-Phase Rewrite (parallel, budget, sub-agents, MCP)
Auto-Update (signed NSIS installer)
Qwen 3.6 Day-0 Support
Plugins Dropdown (Caveman + Personas)
Codex Architect/Editor split + Repo-Map + Stage-and-Approve (v2.5.0)
Code-Review mode + Test-Driven Loop + typed git/gh tools (v2.5.0)
Multi-Repo Agent + .lurules + pr_resume + project_init (v2.5.0)
Voice Mode (Qwen Omni live voice)
Upscale + Inpainting

Build from Source

git clone https://github.com/PurpleDoubleD/locally-uncensored.git
cd locally-uncensored
npm install
npm run dev          # Development
npm run tauri build  # Production binary

Platform Support

Platform	Status	Download
Windows (10/11)	Fully tested	`.exe` / `.msi`
Linux / macOS	Build from source	`npm run tauri build`

Community

Join the Discord: https://locallyuncensored.com/discord. Ask questions, share what you built, or help others in our forum channels — chat / image gen / video gen / coding agent.

Contributing

Check out the Contributing Guide. See open issues or the Roadmap.

License

AGPL-3.0 License — see LICENSE.

Your data stays on your machine.

Website · Report Bug · Request Feature · Discussions

Locally Uncensored

About Locally Uncensored

Platforms

Languages

Links

Need Help Installing Locally Uncensored?