Pinokio
FeedAppsRemoteCreate
Download Pinokio
Sign inCreate accountJoin
Sign inCreate accountJoin

Apps

Announcements

View all

Official updates.

Pinned announcement from Pinokio
pinokio
@cocktailpeanut•1d ago
Pinokio Create: Turn Anything into a 1-Click Launcher, Automatically.Pinokio now talks to your AI agent to package anything into a 1-click launcher, publish to github, and list on pinokio, automatically. Just click the "Create" button at the top, select a template, enter what you want, and open in pinokio. Pinokio will hand off all the instructions needed to your own AI agent, and your agent will take it from there, taking care of everything automatically: Build a new app: Build a new app with 1-click launcher built-in, publish to github, and list to pinokio Publish your own app: Take your existing project, package it as a 1-click launcher, publish to github, and list to pinokio 1-Click Launcher for any existing project you want to try: Take any existing project you want to try out, package it as a 1-click launcher, publish to github and list on pinokio, all automatically.
Mumblebox 3D Asset Foundry
@mumblebox•1h ago
Foundry Preview 35 · The retro ray gun, better MCP tools, and an update fixSorry about all the commits and updates today. We kept making the same circular mistake. Update restarted Foundry, and reopening Foundry could start Update again. Preview 35 removes that step and checks for a newer release before stopping Foundry. I’ve been working on making Foundry easier to try, easier to connect to your agent, and easier to help improve. I made a 1930s-style ray gun to show what that looks like. You can load the same reference in Foundry, export the textured model, and take it into Blender. The Foundry GLB rendered under Blender studio lights. What’s fixed. Clicking Next in Single Image or Text could stop with setText is not defined. Thanks to @rgdesign for flagging it here and to @SibioJak for reproducing it on GitHub. We also fixed a Windows ZIP verification error caused by inconsistent line endings. The Preview 35 update keeps Foundry running when you already have the newest release. What you can try. Open Single Image or Text and choose Use sample image. First-use files shows which files the workflow expects to need and calls out anything it can’t estimate. If you hit an error, Copy details gives you a short report you can paste into an issue. The interface and workflow tips are available in all nine supported languages. The original image you can load with Use sample image. The exported ray gun spinning in Blender. What changed under the hood I’ve been giving the MCP and the companion SKILL.md a lot of attention. The tool descriptions now explain what inputs the agent needs and which calls read information, change a draft, or start a job. An agent can check first-use files before it queues work, watch the same progress you see in the wizard, and pick up the ZIP when it’s ready. On the Red Comet run I stopped the texturing pass through MCP, then retried it. Foundry picked up from the saved geometry and refinement. The same workflow exported a GLB with two embedded 4K maps, which I imported into Blender and lit for the turntable. The updated skill walks agents through choosing the right workflow, checking the result, and bringing it into Blender. Walkthrough, source image, settings, and GLB For the Windows ZIP fix, I compared all 639 files in the release ZIP with what Git produces from the release tag under both Windows line-ending settings. The new build matched byte for byte. Preview 35 should be fine for tonight. I’m sure I’ll find more bugs tomorrow. Thanks for your patience and for trying Foundry, sending in bugs, and sharing your results. More workflows are coming, and I’ll keep refining the recipes and defaults as we see what works on real assets. If you make something you like, I’d love to see it and hear which settings helped. Preview 35 on GitHub · Foundry on Pinokio
Phosphene
@bizarro•23h ago
Phosphene 5.0.24 - now in ten languagesPhosphene 5.0.24 is out. The Studio now speaks your language. Everything here is free for everyone. https://phosphene.video/ Ten languages: English, Spanish, German, French, Italian, Portuguese, Russian, Japanese, Korean and Chinese (Simplified). Phosphene follows your Mac's language by itself; to pick another, open Settings and choose Language. Counts and times read naturally in every language, and labels that were written in pieces now read as whole sentences. The Pro window says what Pro does on this Mac, shows both plans with their prices, the 14-day refund and how your key arrives. Mac app 5.0.24: the app's own screens come in the same ten languages. The Studio in Spanish. Press Update, then restart Phosphene. The Studio picks your Mac's language on the next start. Release notes: https://github.com/mrbizarro/Phosphene/releases/tag/v5.0.24
MiniMax H3 Studio
@mvrdvk•2d ago
How to try it: Open Motion → Face Swap, add a source clip and a portrait of the replacement face, then describe exactly which person to swap. Preview a short shot first and check that the face follows the original head movement and expressions.
Haru
@cocktailpeanut•2d ago
Haru: HER is here. Realtime voice chat with an AI friendRemember the movie HER? You talk to an AI like talking to a friend. It's not an assistant. It's your friend. You can just chat about anything, you can even interrupt while they're talking. Well, now you can actually experience the same thing with Haru. It's fully open source and open weight. It's completely free to use forever. More on X: https://x.com/cocktailpeanut/status/2121287442086994133 And here's what the UI looks like. When you start the chat (which can continue up to 60 minutes) and turn on the microphone, haru introduces himself/herself and you just start chatting. You can even interrupt while haru is speaking, just like chatting with a human. It is not perfect and has some quirks, but definitely worth trying out, really fascinating. System Requirements I've tested on NVIDIA GPU and Mac. Based on my experience you need: At least 12GB VRAM (but 16GB VRAM if you want actual smooth experience) NVIDIA GPU or a 12GB RAM Mac Supported Languages English Spanish Portuguese German French Chinese Japanese Russian Korean Examples Check out some examples: https://x.com/vinami/status/2121049617383231604?a=f1l6aAn7At7YIwYGWao&s=20 https://x.com/HigurashiAI/status/2121043937762889832?a=LxX12XkRp1HaexjZ_8Q&s=20
PotionUI
@oroborus•2d ago
PotionUI 0.2.0: Kits, API keys, Perfect loop videos and a rebuilt details viewPotionUI 0.2.0 is out. Click Update in Pinokio to get it. The Update button now updates the launcher itself too. This release is about reusing what you already have and building on top of PotionUI. Kits keep sets of pictures, clips, audio and text together so you can drop them into any prompt (in other apps this is also called "Actors" or "Cast", but I didn't want to limit it to video only), and API keys let scripts and tools use your instance with exactly the access you choose. What's new Kits: make a named set of pictures, clips, audio and text in the Library and cite one item in a prompt as @kits.name.item. Only the items you cite go to the model. Add items straight from History. One media picker: History, Library and Kits are tabs of the same picker everywhere, with Small, Medium and Large tiles and a Replace menu on any loaded item. Video modes, simplified: MiniMax-H3 Video is now Video Director, with Text to Video, Image to Video and First to Last Frame as single-clip modes. Old sessions still open. Perfect loop for MiniMax-H3: give it a picture and what should move, get back a clip whose end flows into its start. LTX-2.5 Ingredients: characters and props from your pictures stay consistent in a new video. MOSS-Image regional prompting: draw regions on a canvas and describe each one. MiniMax-H3-Image is a new model family, with text to image and edit presets that take LoRAs. API keys: create keys with their own permissions, choose which MCP tools each can use, and generate from a saved session with a script. There is a new API guide with examples. Preset mirrors: admins can rename a preset and make copies that reuse its files with their own name, cover, defaults and modes, plus fixed prompt text that regular users never see. Plugin catalogs: install plugins from a catalog right in Admin. Queue routing: jobs spread across the backends of an engine, and a busy or silent worker no longer fails them. Nicer to use A rebuilt generation details view: parameters in boxes you copy with a click, section counts, and Library items open in the same view with previous and next. Prompt references show as readable thumbnail chips. A button on the resolution picker swaps to the mirrored size. Quick delete: press Delete on the current generation, or delete a whole collection's entries. Tags on models in Admin. A six-step setup wizard that no longer fails because a test generation did, and picks up again after a restart. Community contributions Big thanks to M4-S4 for a large share of this release: Install plugins from a catalog in Admin, and a plugin can no longer end up enabled after a failed install Face detection, face vectors and image embeddings for plugins, plus more plugin page and preset tools, and test databases for plugin authors RTX Upscale can raise a video's frame rate (2x, 4x or a target rate), and RIFE accepts a target frame rate Wan video decoding fits in less memory, and Wan 2.2 5B gets image to video MiniMax-H3-Image edits work on cards with less memory, such as 12 GB The default 1080p SeedVR2 video upscale fits a 12 GB card Effects work again with ffmpeg 8 The Plugins docs now point to community catalogs Fixed The first generation after a start no longer fails on a half-loaded module. Wan 2.2 image to video, and long Wan clips, fit in memory where they ran out. Security: setting up the first account behind a proxy now needs the setup token, users can no longer switch their own MCP access back on, and hidden fixed prompt text can no longer leak. Before you update Back up your data first. The first start migrates your database and that cannot be undone. Your MCP tokens become API keys that work for MCP only, so create a new key if a script needs the HTTP API. A saved session on the retired Long video mode opens with an offer to convert it to Video Director. You may see harmless "plugin not found" lines for two plugins that no longer exist at startup. The old MCP token routes and the token in a socket URL still work this release but are deprecated, so move your scripts over. Windows is fully supported. Update Click Update in Pinokio, start PotionUI and let it finish its first-start migration. Then open Settings and look at API keys. Full release notes: https://github.com/PotionUI/PotionUI/blob/master/CHANGELOG.md Website: https://potionui.com Discord: https://discord.gg/avR4trp3b8
FeaturedPopularNewUpdated
Runs on:Any
System
AnymacOSWindowsLinux
GPU
AnyNVIDIAAMDApple
Show:Launchers
LaunchersPluginsAll projects
#tts64#image-generation50#video-generation42#music35#utility35#video28#próba27#3d26
VoxCPM 2
Tokenizer-free TTS for context-aware speech, voice cloning, and voice design. 2B params, 48kHz, 30 languages (Gradio UI).
PierrunoYT/VoxCPM-2-Pinokiov5.0updated 1mo ago
@pierrunoyt3 check-insNVIDIAAMDApple
PRX Pixel
Pixel-space PRX text-to-image pipeline (~7B params, Qwen3-VL text encoder, no VAE)
PierrunoYT/PRX-Pixel-Pinokiov5.0updated 1mo ago
@pierrunoyt1 check-inNVIDIAAMDApple
GLM-TTS
🎙️ Controllable & Emotion-Expressive Zero-shot TTS with Multi-Reward Reinforcement Learning. High-quality text-to-speech synthesis supporting zero-shot voice cloning and streaming inference with natural emotional expression.
PierrunoYT/GLM-TTS-Pinokiov3.7updated 1mo ago
@pierrunoyt0 check-insNVIDIAAMDApple
LFM2.5-450M-VL
LFM2.5-VL-450M (Liquid AI): compact vision–language model for image understanding. Gradio UI with image upload/URL or video (sampled frames), prompt, and generation sliders.
PierrunoYT/LFM2.5-450M-VL-Pinokiov5.0updated 1mo ago
@pierrunoyt0 check-insNVIDIAAMDApple
IndexTTS 2.5
Zero-shot multilingual voice cloning with cross-lingual synthesis, disentangled emotion control, pronunciation guidance, and speaking-speed control.
PierrunoYT/IndexTTS-2.5-Pinokiov5.0updated 1mo ago
@pierrunoyt0 check-insNVIDIAAMDApple
dots.tts-base
2B-parameter fully continuous, end-to-end autoregressive text-to-speech with zero-shot voice cloning. https://huggingface.co/rednote-hilab/dots.tts-base
PierrunoYT/dots.tts-Pinokiov5.0updated 1mo ago
@pierrunoyt1 check-inNVIDIAAMDApple
Transcribr
Bulk transcribe many YouTube videos, whole playlists, or your own uploaded audio/video files at once with faster-whisper. Outputs txt, srt, vtt, or json.
PierrunoYT/Transcribr-Pinokiov5.0updated 1mo ago
@pierrunoyt0 check-insNVIDIAAMDApple
VidLingo
YouTube to MP3, Cohere transcription, TranslateGemma translation, OmniVoice TTS. https://github.com/PierrunoYT/VidLingo-Pinokio
PierrunoYT/VidLingo-Pinokiov6.0.0updated 1mo ago
@pierrunoyt0 check-insNVIDIAAMDApple
Hy-MT2
Hy-MT2 multilingual translation — Gradio UI with 38 language and variant choices for Hy-MT2-1.8B, Hy-MT2-7B, and Hy-MT2-30B-A3B.
PierrunoYT/Tencent-HY-MT2-Pinokiov5.0updated 1mo ago
@pierrunoyt0 check-insNVIDIAAMDApple
LuxTTS 🎙️
High-quality rapid TTS voice cloning model (150x+ realtime) — 48kHz speech, voice cloning
PierrunoYT/LuxTTS-Pinokiov7.0updated 1mo ago
@pierrunoyt2 check-insNVIDIAAMDApple
PocketTTS
Lightweight CPU text-to-speech with preset voices and optional Hugging Face-authenticated voice cloning.
PierrunoYT/pocket-tts-pinokiov5.0updated 1mo ago
@pierrunoyt1 check-inNVIDIAAMDApple
Higgs Audio v3 TTS
Pinokio launcher for Higgs Audio v3 TTS with Gradio UI, SGLang-Omni backend, and automatic model download.
PierrunoYT/HiggsAudioV3-Pinokiov7.0updated 1mo ago
@pierrunoyt2 check-insNVIDIAAMDApple
Breeze TTS 2
BreezeBlue Breeze TTS 2 — open-weight bilingual (EN/ZH) text-to-speech with voice clone, voice design, voice direction, and inline vocal events. Requires an NVIDIA GPU with 12GB+ VRAM.
PierrunoYT/Breeze-TTS-2-Pinokiov5.0updated 1mo ago
@pierrunoyt0 check-insNVIDIAAMDApple
Audio Flamingo 3
NVIDIA's Audio Flamingo 3 - Large Audio-Language Model for speech, sound, and music understanding with Gradio web interface
PierrunoYT/Audio-Flamingo-3-Pinokiov7.0updated 1mo ago
@pierrunoyt0 check-insNVIDIAAMDApple
PersonaPlex
🗣️ PersonaPlex - NVIDIA's real-time speech-to-speech conversational AI model. Natural full-duplex conversations with customizable personas and voices.
PierrunoYT/PersonaPlex-Pinokiov5.0updated 1mo ago
@pierrunoyt3 check-insNVIDIAAMDApple
Supertonic 3
Lightning-Fast, On-Device, Multilingual TTS — Gradio, ONNX, 44.1kHz
PierrunoYT/Supertonic-3-Pinokiov5.0updated 1mo ago
@pierrunoyt1 check-inNVIDIAAMDApple
SmolLM3-3B Chatbot
A local AI chatbot powered by SmolLM3-3B with a Gradio web interface.
PierrunoYT/SmolLM3-3B-Pinokio-Gradiov5.0updated 1mo ago
@pierrunoyt0 check-insNVIDIAAMDApple
Cohere Transcribe
State-of-the-art open-source speech recognition model supporting 14 languages. 2B parameter ASR model from Cohere Labs.
PierrunoYT/cohere-transcribe-pinokiov5.0updated 1mo ago
@pierrunoyt1 check-inNVIDIAAMDApple
DramaBoxFeatured
Expressive TTS with voice cloning, prompt-driven speech synthesis built on LTX-2.3 by Resemble AI
PierrunoYT/DramaBox-TTS-Pinokiov5.0updated 1mo ago
#tts#voice-clone
@pierrunoyt9 check-insNVIDIAAMDApple
ChatterBox
AI-Powered Text-to-Speech with Voice Cloning using Chatterbox TTS and a Gradio interface. Includes Turbo, Multilingual (23+ languages), and Original models. Runs locally; CUDA GPU recommended, CPU supported. Windows, Mac, and Linux.
PierrunoYT/chatterbox-tts-pinokiov5.0updated 1mo ago
@pierrunoyt3 check-insNVIDIAAMDApple
Previous
1...232425...79
NextPage 24 of 79

Get project updates

Pulse

View all

The latest posts from across Pinokio.

Foundry Preview 35 · The retro ray gun, better MCP tools, and an update fix
@mumblebox · Mumblebox 3D Asset Foundry
Sorry about all the commits and updates today. We kept making the same circular mistake. Update resta...
is it free?
@preneshkl · MiniMax H3 Studio
Checked in to Maestro.
@gjphilander · Community
Checked in to FaceFusion 3.9.2.
@alexvichko1997 · Community
browser-use: AssertionError: SRE module mismatch
@alaye771 · browser-use1
App: browser-use (browser-use.git) Repo: https://github.com/pinokiofactory/browser-use.git Generated:...

Radar

View all

Projects getting attention.

Followed·just now
PotionUI
PotionUI: self-hosted AI image and video generation with a preset-driven interface, a native diffusers engine and ComfyUI support. Requires Windows or Linux with an NVIDIA GPU (driver 580 or newer, 8 GB VRAM and 16 GB RAM minimum). Other machines get a remote-only install.
Installable
Followed·2 min
YUE2 // GROOVE
YUE2 // GROOVE is the latest music studio built on the open-source Yue2 model and its inference stack. Generate high-quality full songs from style and lyrics with an editable score plan — powered by the latest YuE model — cover from audio with SheetSage2 and MERT2, refine and compare edits, and keep your works in a reusable, easy-to-manage library. You get high-quality creation with real creative control. Hardware: an NVIDIA GPU with 24 GB VRAM on Linux (YuE2's recommended setup, validated end-to-end on NVIDIA L4 hosts; a 16 GB memory budget runs everything the app can produce, and 12 GB runs the unquantized model at CFG 1.0 or for shorter songs) or an Apple Silicon Mac with 32 GB+ unified memory (where this app is developed and tested). Windows is best effort: install, launch, Cover and a full-length song verified on Windows 11 (RTX 2070, 8 GB); song generation there runs through the GGUF engine. Cards under 16 GB (and Windows) get the optional GGUF engine: the same model through yue2.cpp with an 8-bit backbone, 8.2 GB peak for a full song, measured indistinguishable from the reference rendering in a blind ABX, a different take for the same seed; the reference PyTorch configuration stays the default wherever it fits.
Installable
Followed·3 min
Comfyui
The most powerful and modular diffusion model GUI, api and backend with a graph/nodes interface. https://github.com/comfyanonymous/ComfyUI
Installable
Followed·4 min
Mumblebox 3D Asset Foundry
Turn an image or prompt into a textured 3D asset on your own GPU. Install and start Foundry in Pinokio, then try the included ray-gun reference in Single Image or Text and export a GLB you can edit in Blender. Foundry also offers guided meshes, rigged hair, Gaussian splats and exports for game development and 3D printing. Connect Codex, Claude Code or another compatible client to the included local MCP to start jobs, see progress and retrieve exports from the same workflows. Windows/NVIDIA preview. Available in English, Simplified Chinese, Latin American Spanish, French, Brazilian Portuguese, German, Japanese, Korean and Hindi. 用中文做出你的下一个 3D 素材! ¡Crea tu próximo asset en español! Créez votre prochain objet 3D en français ! Crie seu próximo asset em português! Erstelle dein nächstes 3D-Asset auf Deutsch! 日本語で次の3Dアセットを作ろう! 한국어로 다음 3D 에셋을 만들어 보세요! अपना अगला 3D एसेट हिंदी में बनाएँ!
Installable
Followed·6 min
Video Studio
Text-to-video, image-to-video and audio-driven music videos (LTX-Video / Wan 2.x via Diffusers)
Installable
Pinokio
PrivacyTerms
FeedAppsCreateRemote

Explore the community

#showcase
Minimax H3 Singularity 4 Step Generation Prompt: Use the provided image as the starting frame. Preserve the exact minimalist black-and-white Wojak meme drawing style, characters, and party setting. Static camera, no cuts, no zooms, no dialogue, no text. 0–3s: The man leans casually against the wall in the corner, legs crossed, one hand in his pocket, holding his drink. He has a faint, peaceful smile. Across the room, three extremely attractive adult women dancing at the party notice him. They exchange excited glances. 3–7s: The first woman confidently approaches him, flips her hair, gives h...
@cocktailpeanut·17d ago
#cool-project
Preview of my next project!
@theawakenone·4d ago
#guide
Pinokio Create: Turn Anything into a 1-Click Launcher, Automatically.
@cocktailpeanut·1d ago
#testers-wanted
For those of you brave enough to install... Thanks for trying it out. I am adding a bug report feature and fixing the workflows for objects/creatures/vehicles. I am also adding a coding agent driven headless Blender node. I will be looking to optimizes for low memory systems soon. There is alot to do on the Roadmap. Thanks for your interest.
@mumblebox·13d ago