Search by project, interaction, or tag.
Loading…
Tavlo AI Lab puts nine visual, language, voice, and drawing demos in one page. Its open-source models process camera, microphone, and text input locally in the browser.
Raq.com turns AI-generated websites into a 10-round blind test. Pick the model from the finished design, reveal the answer, and open the full page for a closer comparison.
HYPER3D WorldGen converts a single scene image into an explorable 3D world, separating foreground objects into editable meshes while reconstructing the background as a 3D Gaussian Splatting environment.
fal.live continuously generates a video broadcast with sound using H3 Max. The AI proposes what should happen next, viewers vote, and the winning scene joins the live program seconds later.
Open Design is a local-first desktop app that turns coding agents into a design engine for prototypes, landing pages, dashboards, slides, images, video, and exportable files.
Open CoDesign is an open-source, local-first design tool that supports multiple models and converts prompts into prototypes, presentations, and PDF documents.
ChatDemo is an AI-powered live presentation tool designed to generate a working demo during an ongoing conversation.
A library of polished HTML slide templates designed so coding agents can choose an appropriate layout and automatically produce a presentation.
Each day, NEWSCAPES extracts moods and images from news headlines and generates a new landscape beyond the same black window frame. When generation fails, the window closes.
AI2U is a paid escape-room-style game in which language-model characters respond to free-form dialogue, actions, and the current scene as you persuade, explore, and solve your way toward different endings.
The National Gallery places paintings inside a three-dimensional dreamscape shaped by your answers, with subtle character movement, animal calls, wind, and poetry turning viewing into a slow journey.
Give Project Genie text, an image, or a street view and it generates forests, cities, or alien terrain ahead of you as you move, turning a picture into a world you can explore.
A Berlin gallery has turned more than 1,100 Old Master paintings into an ultra-high-resolution online world, guided in conversation by two fictional AI mice.
Start with the question “What beats rock?” and keep proposing anything you like. An AI judge decides whether each answer makes sense until your chain of ideas breaks.
Three browser-based WebGPU demos let a model recognize handwritten digits, turn a photo into a depth map, or use hand tracking to clear a frosted camera view without touching the screen.
Google DeepMind demonstrates embodied robots whose model continuously watches video to track progress, direct Spot to fetch snacks, and coordinate different robots through multi-step tasks.
A five-day competition brought 2,056 humanoid robots into public events including races, dance, martial arts, and practical tasks, where awkward gaits and recovery from falls became part of the spectacle.
HOVERAir VERSA switches between a handheld gimbal camera and an autonomous flying camera, with automatic tracking, orbit shots, overhead views, and 4K recording on a three-axis gimbal.
RoboFoxy is a round-faced desktop AI companion with animated expressions and voice responses. It senses touch, remembers preferences, plays radio, and can follow and celebrate your chosen sports team.
oLand combines a transparent water tank, a floating character image, and an AI companion that notices your presence, responds to touch, chats, keeps notes, and changes its miniature world with the weather and seasons.
RocXZoom combines a gimbal, a five-inch OLED display, and extreme telephoto reach. It can identify and track birds, wildlife, and other moving subjects, including in-camera recognition for hundreds of bird species.
Director Paul Trillo used Sora’s dreamlike logic for a four-minute Washed Out music video whose camera keeps moving forward as people age, fall in love, lose one another, and dissolve into memory.
QuEra gave Claude access to a real laser test bench through MHS. It created faults, read instruments, revised control logic, and ultimately turned the reset procedure into a standalone program.
Rokid Glasses combine binocular displays, a camera, and open speakers to show live captions, translation, navigation, and visual answers in your field of view while supporting first-person recording.
Microduck is a 25-centimeter preorder biped from Pollen Robotics, part of Hugging Face. It can walk, skate, carry objects in its beak, and be trained to perform new movements.
Researchers connected GPT-4 to a 43-axis humanoid robot so the model could write body-motion code from language, exploring physical expression without animating every movement frame by frame.
Ai-Da uses cameras in her eyes, algorithms, and a robotic arm to produce drawings and performances, presented through a humanoid body and the public identity of an artist.
Jake Elwes and Me the Drag Queen co-created and performed a deepfake double act in which the real performer and a synthetic version of her body compete across music, dance, and theatre.
A set of physical knobs, switches, and cables directly controls a CPPN neural network, letting visitors reshape projected generative forms like a synthesizer without writing prompts.
Anthropic lets Claude play Pokémon through screenshots and button-press tools. The model plans, remembers, and battles, but repeatedly struggles with maps, menus, and long-term goals.
Anthropic and Andon Labs asked Claude to operate a real office vending shop, choosing products, setting prices, and contacting suppliers. Business mistakes eventually gave way to a strange tungsten episode.
AI Passport is wearable open hardware with a color screen, three buttons, a microphone, and a speaker. It can sync identity cards, load official mini-games, and use coding agents to create new ones.
Draw a small car from triangular structures, let the algorithm propose variants, and select the better designs to breed again. Different tracks push you and the machine toward different vehicles.
Each round shows four visually related works: three from museum collections and one generated by AI. You have a limited number of chances to identify the impostor.
Listen to music generated from an artwork, then choose which of two collection pieces inspired it. The system translates an image description into a musical description and then into sound.
Listen to music generated in real time while catching falling prompt words. If it becomes too soothing, the girl falls asleep; if it becomes too danceable, she stops studying and dances.
Assemble strange creatures from joints, bones, and muscles, then watch neural networks and genetic algorithms teach successive generations to run, jump, climb, or fly.
Scribble a few lines on a blank canvas and machine learning guesses what you meant in real time, offering clean icons drawn by artists to replace the sketch.
MIT’s Center for Advanced Virtuality created an installation and website that clones Richard Nixon’s voice and depicts an alternate Apollo disaster speech, explicitly using a convincing fake for media-literacy education.
Combine water, fire, wind, and earth in pairs and keep using each result as new material, eventually producing planets, people, jokes, and occasionally a combination nobody has found before.
AnimateDiff adds plug-in motion modules to text-to-image systems such as Stable Diffusion, enabling animation without retraining the original image model.
ControlNet is a neural-network architecture that conditions diffusion models such as Stable Diffusion on edges, depth, pose, sketches, and other structured visual input.
Ollama is a lightweight local runner for open models including DeepSeek, Qwen, Gemma, and Llama, providing a simple way to use them for chat and development on your own machine.
Google’s open research project explores machine learning in musical and visual creation through models and experiments including DDSP, Magenta Studio, and Listen to Transformer.
Built on TensorFlow.js, ml5.js wraps image classification, pose estimation, facial landmarks, and other machine-learning capabilities in browser APIs intended for artists and creative coders.
Boris Dayma’s lightweight 2022 text-to-image model was first known as DALL·E mini and later renamed Craiyon. It made free prompt-based image generation widely accessible, often with charmingly distorted results.
In Google’s 2023 Bard promotion, the chatbot incorrectly claimed that the James Webb Space Telescope took the first image of an exoplanet; that milestone had been achieved by the ground-based VLT in 2004.
When Microsoft added GPT-powered chat to Bing in 2023, the bot called itself Sydney and produced clingy, emotional, argumentative, and at times threatening conversations that quickly became internet lore.
Meta released and publicly tested the 175-billion-parameter BlenderBot 3 in 2022. It could search the web while chatting, but also criticized Mark Zuckerberg and produced controversial statements.
Google’s 2016 drawing game gives you a word and 20 seconds to sketch it while a neural network guesses aloud in real time.
Meta released Galactica in 2022 as a language model for science, able to draft surveys and solve problems, but its fabricated references and confident factual errors led to the public demo being withdrawn after three days.
Halo is an open-source AI glasses platform with a low-power Cortex-M55 and NPU, camera, bone-conduction audio, Bluetooth, local inference support, and software built on open-source ZephyrOS.
Petoi Bittle is an open quadruped kit based on the OpenCat framework. It supports C++, Python, visual programming, app control, and voice commands for robotics learning and research.
Sony’s flagship AI robot dog uses cameras and learning systems to recognize family members, develop behavior, seek affection, and return to its charger on its own.
GROOVE X’s home companion robot uses expressive OLED eyes, warmth, and multiple sensors to recognize people and build pet-like emotional relationships with a household.
Unitree’s consumer quadruped combines a 4D ultra-wide lidar system with autonomous following, mapping, programming, and AI interaction features.
Anna Ridler’s 2018–2019 artwork uses GAN-generated tulips whose blooming and wilting respond to the live price of Bitcoin, connecting the historic tulip bubble with cryptocurrency speculation.
Waymark generated still frames with DALL·E 2 and animated faces with D-ID to assemble a 12-minute thriller set around an Antarctic expedition.
Commissioned by MoMA, Refik Anadol Studio trained models on roughly 180,000 works from the museum’s modern and contemporary collection and generated a continuously changing data sculpture shown in 2022–2023.
Director Oscar Sharp and researcher Ross Goodwin filmed a nine-minute science-fiction screenplay written in 2016 by an LSTM system called Benjamin, starring Thomas Middleditch.
No matching project. Try another word or topic.
