Runs AI locally
69 projects tagged Runs AI locally, in the order of our ranking. Some replace a paid product; others are free tools people use alongside one.
Most often they replace ElevenLabs, Suno or Descript. 69 of 69 run on your own machine. The most starred is Ollama, with 186k stars on GitHub.
| Compare | Project | Runs on my machine | Allowed at work | Paid services | Hardware | Setup | Stars |
|---|---|---|---|---|---|---|---|
| OllamaLocal model runner that downloads and serves open LLMs (Gemma, Qwen, DeepSeek, gpt-oss and more) with a CLI and local API. | Yes | YesMIT | Optional only | No graphics card needed | Easy | 186k+820 this week | |
| ComfyUINode-graph app for running open image, video, 3D and audio models locally, with inpainting, outpainting, upscaling and a local API. | Yes | ConditionsGPL-3.0 | Optional only | GPU, about 4 GB | Easy | 145k+1,384 this week | |
| VoiceStudioLocal desktop app for voice cloning, voice design, video dubbing, dictation, transcription and audiobooks, built on OmniVoice and other engines. | Yes | NoAGPL-3.0-only | Optional only | No graphics card needed | Easy | 99k+9,787 this week | |
| vLLMOpen-source engine for serving LLMs at high throughput on your own GPUs, with an OpenAI-compatible API server; Apache-2.0. | Yes | YesApache-2.0 | None required | No graphics card needed | Medium | 96k+537 this week | |
| GPT-SoVITSFew-shot voice cloning and TTS WebUI; 5-second zero-shot or 1-minute fine-tune, five languages, MIT code and weights. | Yes | YesMIT | None required | No graphics card needed | Easy | 65k+808 this week | |
| whisper.cppDependency-free C/C++ port of OpenAI's Whisper speech recognition, running on CPU or GPU across desktop, mobile and the browser. | Yes | YesMIT | None required | No graphics card needed | Hard | 55k+179 this week | |
| HandyFree offline speech-to-text for Windows, macOS and Linux; press a shortcut, speak, and your words appear in any text field. | Yes | YesMIT | None required | No graphics card needed | Easy | 37k | |
| Fish SpeechFish Audio's multilingual TTS and voice cloning stack (Drama 3, S2-Pro) under a source-available community licence. | Yes | ConditionsFish Audio Community License 1.0, not OSI open source | Optional only | GPU, about 24 GB | Medium | 34k+136 this week | |
| MeetilyLocal AI meeting assistant that records mic and system audio, transcribes live with Whisper or Parakeet and summarises with an LLM. | Yes | YesMIT | Optional only | No graphics card needed | Easy | 32k+195 this week | |
| VoiceboxLocal desktop voice studio that clones voices, generates speech with 8 TTS engines, transcribes with Whisper and adds system-wide dictation. | Yes | ConditionsMIT | None required | No graphics card needed | Easy | 59k | |
| InvokeAILocal web UI for image generation with a unified canvas for inpainting and outpainting, node workflows and a model manager. | Yes | ConditionsApache-2.0 | Optional only | GPU, about 4 GB | Easy | 29k+92 this week | |
| faster-whisperReimplementation of OpenAI Whisper on CTranslate2, up to 4x faster with less memory, with int8 on CPU or GPU. | Yes | YesMIT | None required | No graphics card needed | Easy | 26k+234 this week | |
| UpscaylDesktop AI image upscaler for Windows, macOS and Linux, built on Real-ESRGAN models running through ncnn and Vulkan. | Yes | ConditionsAGPL-3.0 | None required | Needs a graphics card | Easy | 52k+539 this week | |
| WhisperOpenAI's general-purpose speech recognition model for multilingual transcription, translation to English and language ID. | Yes | YesMIT | None required | GPU, about 1 GB | Easy | 112k+462 this week | |
| WhisperXFast batched Whisper transcription with word-level timestamps via wav2vec2 alignment and pyannote speaker diarization. | Yes | ConditionsBSD-2-Clause | None required | No graphics card needed | Medium | 25k+169 this week | |
| FreeTokenApache-2.0 engine and desktop app for running very large open models on your own NVIDIA GPU, with OpenAI and Anthropic-compatible APIs. | Yes | YesApache-2.0 | None required | GPU, about 8 GB | Medium | 16k | |
| Ultimate Vocal RemoverDesktop app that splits songs into vocals and instrumental or other stems using MDX-Net, VR, Demucs and MDX23C models. | Yes | YesMIT | None required | GPU, about 6 GB | Easy | 27k+126 this week | |
| FluidVoiceOn-device voice dictation app for macOS that also records meetings (FluidMeet) with live captions, speaker labels and transcripts. | Yes | ConditionsGPL-3.0 | Optional only | See its page | Easy | 13k+203 this week | |
| YuE2Open song model that writes an editable melody-and-chord score from lyrics and style, then sings it as a full song with accompaniment. | Yes | ConditionsApache-2.0 | None required | GPU, about 24 GB | Medium | 13k+252 this week | |
| F5-TTSFlow-matching text-to-speech research model that clones a voice zero-shot from a reference audio clip. | Yes | NoMIT | None required | See its page | Medium | 16k+44 this week | |
| AutoClipDesktop app, Docker web UI and CLI/MCP that cut podcasts and interviews into platform-ready clips with captions, covers and post copy. | Yes | YesMIT | Optional only | See its page | Easy | 9.9k+170 this week | |
| ACE-Step 1.5Open-weight music model that writes and renders full songs with vocals locally, with covers, repainting, stem extraction and LoRA training. | Yes | YesMIT | None required | GPU, about 4 GB | Easy | 14k+248 this week | |
| SD.NextAll-in-one web UI for image and video generation across many models and hardware backends, with inpainting, upscaling and quantization. | Yes | ConditionsApache-2.0 | None required | No graphics card needed | Medium | 7.4k+5 this week | |
| chaiNNerNode-based image processing app built for AI upscaling, running PyTorch, NCNN, ONNX and TensorRT models on GPU or CPU. | Yes | ConditionsGPL-3.0 | None required | No graphics card needed | Easy | 6.1k+19 this week | |
| PiperFast, local neural text-to-speech engine with downloadable voices in 40+ languages, used by Home Assistant and NVDA. | Yes | ConditionsGPL-3.0 | None required | No graphics card needed | Easy | 6.1k+98 this week | |
| OpenScreenDesktop recorder and editor for product demos with webcam overlay, auto-zoom, on-device captions and a CLI; MIT, exports MP4 or GIF. | Yes | YesMIT | Optional only | No graphics card needed | Easy | 5.9k+930 this week | |
| Krita AI DiffusionKrita plugin for generative fill, live painting, upscaling and reference-guided generation, using a local or remote ComfyUI backend. | Yes | ConditionsGPL-3.0 | Optional only | GPU, about 6 GB | Easy | 11k+34 this week | |
| JasnaDesktop app and CLI that finds mosaic-censored areas in videos and images and restores them with AI models on your own NVIDIA or AMD GPU. | Yes | ConditionsAGPL-3.0 | Optional only | See its page | See details | 4.3k+542 this week | |
| HexLocal voice dictation for Mac, Windows and Linux; hold a key, speak, and it pastes. MIT, with Windows and Linux in beta. | Yes | ConditionsMIT | None required | See its page | Easy | 4.1k | |
| Final2xCross-platform desktop image super-resolution app with built-in models and support for loading your own. | Yes | ConditionsBSD-3-Clause | None required | No graphics card needed | Easy | 7.4k+7 this week | |
| ChatterboxResemble AI's family of open TTS models (Turbo, Nano, Multilingual) with zero-shot voice cloning from a reference clip. | Yes | YesMIT | Optional only | No graphics card needed | Easy | 28k+218 this week | |
| yovoiceLocal desktop audio studio for macOS and Windows that makes songs and instrumentals with ACE-Step 1.5, alongside text-to-speech and voice cloning. | Yes | ConditionsApache-2.0 | Optional only | GPU, about 12 GB | Easy | 2.0k+130 this week | |
| YuE2 StudioWindows desktop app for YuE2 that writes an editable score and sings it as a full song offline from 6 GB of VRAM, with covers and stems. | Yes | ConditionsMIT | Optional only | GPU, about 6 GB | Easy | 1.8k+418 this week | |
| Audio SeparatorCommand-line tool and Python package that splits audio into vocals, instrumental and other stems using UVR, Demucs and RoFormer models. | Yes | ConditionsMIT | None required | No graphics card needed | Easy | 1.5k+19 this week | |
| Stable Audio Open (stable-audio-tools)Stability AI's training and inference code for Stable Audio Open, which makes short instrumental music and sound clips from text. | Yes | ConditionsMIT | None required | See its page | Medium | 3.9k+4 this week | |
| RescriptTranscript-based audio and video editor; delete words in the transcript to cut the media. Runs on-device. Non-commercial licence. | Yes | NoPolyForm Noncommercial, not OSI open source | None required | No graphics card needed | Easy | 1.0k+70 this week | |
| Magenta RealTime 2Google's open-weights live music model that streams instrumental music in real time from text prompts on Apple Silicon Macs. | Yes | YesApache-2.0 | None required | Needs a graphics card | Easy | 1.9k+13 this week | |
| HeartMuLaApache-2.0 music language model (3B) that generates songs with vocals from lyrics and style tags, plus a codec and lyrics transcriber. | Yes | YesApache-2.0 | None required | See its page | Medium | 3.8k+9 this week | |
| QualityScalerWindows app that upscales and denoises photos and videos with ONNX models on any DirectX 12 GPU. | Yes | ConditionsMIT | Optional only | GPU, about 4 GB | Medium | 3.2k+15 this week | |
| ACE-Step StudioWindows desktop studio for ACE-Step 1.5 that makes full songs with vocals offline, with covers, repaints, stems, LoRA training and karaoke. | Yes | YesMIT | Optional only | GPU, about 4 GB | Easy | 487+39 this week | |
| Stable Diffusion web UI (AUTOMATIC1111)Gradio web UI for Stable Diffusion with txt2img, img2img, inpainting, outpainting and upscalers. Development has largely stalled. | Yes | ConditionsAGPL-3.0 | None required | GPU, about 4 GB | Medium | 165k+78 this week | |
| HotClipDesktop app that cuts long videos and livestream replays into 9:16 shorts with AI highlight picks, virality scores, auto-reframe and word-level captions. | Yes | ConditionsAGPL-3.0-only | Optional only | No graphics card needed | Easy | 413+32 this week | |
| VideoHighlighterLocal desktop app that scores the strongest moments in long footage from motion, audio, objects and transcript, then exports a highlight reel and clips. | Yes | ConditionsAGPL-3.0 | Optional only | No graphics card needed | Easy | 362+36 this week | |
| acestep.cppC++/GGML port of ACE-Step 1.5 with a local web UI and API; runs quantised song models on CPU, CUDA, Metal or Vulkan without Python. | Yes | YesMIT | None required | No graphics card needed | Hard | 452+8 this week | |
| KronosOpen-weights foundation model for financial candlestick (OHLCV) data that forecasts price series; MIT code and weights from 4M to 102M parameters. | Yes | YesMIT | None required | No graphics card needed | See details | 46k+2,155 this week | |
| ACE-Step UISuno-style web interface for a local ACE-Step 1.5 install, with a song library, covers, repaint, stem extraction and an audio editor. | Yes | ConditionsMIT | None required | GPU, about 4 GB | Medium | 5.2k+40 this week | |
| LexiconLocal-first desktop editor with LanguageTool proofreading plus offline AI rewrite, tone and summary tools from a bundled small LLM. | Yes | YesMIT | None required | No graphics card needed | Easy | 222+1 this week | |
| AutoShorts (gameplay)Python tool that picks moments from long videos (gameplay first), crops to 9:16 and adds styled captions and optional AI voiceovers. NVIDIA GPU only. | Yes | YesApache-2.0 | Optional only | Needs a graphics card | Hard | 206+1 this week | |
| Clips KittyWindows app that turns long streams and videos into scored 9:16 clips with speaker tracking, burned-in captions and optional publishing, using a local LLM. | Yes | ConditionsAGPL-3.0-or-later | Optional only | No graphics card needed | Easy | 197+42 this week | |
| HOT-Step CPPPortable local music studio running ACE-Step 1.5, MiniMax-Music3 and YuE2 in one C++ engine, with covers, repaint, stems and adapter training. | Yes | ConditionsMIT | Optional only | No graphics card needed | Easy | 190+5 this week | |
| podcliLocal AI podcast clipper with speaker face tracking, burned-in captions, multicam editing, web studio, MCP server and YouTube publishing. | Yes | ConditionsAGPL-3.0 | Optional only | See its page | Easy | 81 | |
| SendRecSelf-hosted async video messaging; record screen and camera in the browser, share links, comments, transcripts, AI summaries and viewer analytics. | Yes | ConditionsAGPL-3.0 | None required | No graphics card needed | Medium | 65+1 this week | |
| AudioCraft (MusicGen)Meta's research library for MusicGen, a text-to-music model for short instrumental clips, optionally guided by a hummed or played melody. | Yes | NoMIT | None required | Needs a graphics card | Medium | 24k+35 this week | |
| waifu2x-ncnn-vulkanPortable command-line waifu2x upscaler and denoiser that runs on Intel, AMD, Nvidia and Apple Silicon GPUs via Vulkan. | Yes | YesMIT | None required | No graphics card needed | Easy | 3.5k+4 this week | |
| FooocusOffline, Midjourney-style SDXL image generator with minimal settings. Now in limited LTS with bug fixes only. | Yes | ConditionsGPL-3.0 | None required | GPU, about 4 GB | Easy | 54k+103 this week | |
| Coqui TTSPython TTS library with many models (XTTS-v2, VITS, YourTTS, Bark, Tortoise); original repo unmaintained, continued by idiap fork. | Yes | ConditionsMPL-2.0 | None required | No graphics card needed | Easy | 46k+32 this week | |
| OpenVoiceMyShell's instant voice cloning model with tone-colour transfer, style control and cross-lingual cloning; MIT code and weights. | Yes | YesMIT | None required | No graphics card needed | Medium | 38k+44 this week | |
| Real-ESRGANReference AI super-resolution model for real-world photos and anime, with a Python script and portable Vulkan executables. | Yes | YesBSD-3-Clause | None required | See its page | Easy | 37k+93 this week | |
| Dia1.6B-parameter English text-to-dialogue model from Nari Labs with two-speaker tags, non-verbal sounds and audio-prompt voice cloning. | Yes | YesApache-2.0 | None required | GPU, about 4.4 GB | Easy | 19k+2 this week | |
| CodeFormerFace restoration model from NTU S-Lab with a fidelity slider, video input and face colorization; non-commercial licence. | Yes | NoNTU S-Lab License 1.0, not OSI open source | None required | See its page | Medium | 18k+10 this week | |
| Sesame CSMSesame's 1B Conversational Speech Model that generates speech from text plus optional conversational audio context. | Yes | ConditionsApache-2.0 | None required | See its page | Medium | 15k+5 this week | |
| Orpheus TTSLlama-based 3B speech LLM from Canopy Labs for expressive, emotive, low-latency streaming text-to-speech. | Yes | ConditionsApache-2.0 | Optional only | No graphics card needed | Medium | 6.4k+13 this week | |
| GFPGANTencent ARC face restoration model that repairs blurry, low-res faces in old photos, with optional Real-ESRGAN background upscaling. | Yes | ConditionsApache-2.0 with conditions | None required | See its page | Medium | 38k | |
| SeedVR2 Video Upscaler (ComfyUI)ComfyUI nodes and standalone CLI for ByteDance's SeedVR2, a one-step diffusion upscaler for images and video. | Yes | YesApache-2.0 | None required | GPU, about 8 GB | Medium | 3.0k+13 this week | |
| KokoroLightweight 82M-parameter open-weight text-to-speech model with a pip-installable Python inference library. | Yes | YesApache-2.0 | None required | See its page | Easy | 9.9k+186 this week | |
| MeloTTSMultilingual text-to-speech library from MyShell.ai and MIT covering English accents, Spanish, French, Chinese, Japanese and Korean. | Yes | YesMIT | None required | No graphics card needed | Medium | 7.7k+13 this week | |
| StyleTTS 2Research text-to-speech model using style diffusion and adversarial training, with pretrained LJSpeech and LibriTTS checkpoints. | Yes | ConditionsMIT | None required | No graphics card needed | Medium | 6.4k+5 this week | |
| DiffRhythmDiffusion model that generates full-length songs (up to 4m45s) from timed lyrics and a text or audio style reference. | Yes | ConditionsApache-2.0 | None required | GPU, about 8 GB | Medium | 2.4k+3 this week | |
| colibriRuns huge mixture-of-experts AI models on an ordinary PC by reading parts of the model from disk; pure C, CPU-first, with a chat dashboard and local API. | Yes | ConditionsApache-2.0 | None required | No graphics card needed | Medium | Not known |
Tick up to three to compare. “Allowed at work” includes projects with conditions, marked amber. Unknown is never counted as a yes. Stars show attention, not quality.
Know a project we’re missing? Suggest it.