Apples model to generate Gaussian splats from a single image. Takes about 30 seconds on an M1 Pro.
It falls apart once you move too much, but for a little side-wiggling or a second-eye view for VR, it's great. And looks a lot better than the old approach of depth map + vertex shaders that I use in https://github.com/combatwombat/tiefling. But ml-sharp has 2.6 GB weights, a bit too big to run in the browser.
An LLM step also works pretty well for diarization. You get a transcript with speaker-segmentation (with whisper and pyannote for example), SPEAKER_01 says at some point „Hi I’m Bob. And here’s Alice“, SPEAKER_02 says „Hi Bob“ and now the LLM can infer that SPEAKER_01 = Bob and SPEAKER_02 = Alice.
Very neat. I like the channel idea. There’s also https://random-video.com which shows one of 4.5 Billion YouTube videos randomly (or with a filter for view count, language, year).
Pretty fun to discover what’s out there, without being influenced by YouTubes algorithm. It’s based on the YT Archive project and some other sources, since YT sadly has no randomize function.
There seem to be a few fancy boat projects that turn out to just be renderings and NFT scams.
Like https://www.pangeosyacht.com/projects, an $8 Billion „Terayacht“ maybe launching in 2033. Until then they sell NFTs for a place in the virtual Unreal Engine version of this (also still to be built).
https://font2png.com to browse font-icons and export them as PNG, with background/foreground color. Usually you want SVG, but sometimes a PNG is better. I use it mostly to generate quick favicons. It was also fun to make it work completely client side with canvas.
https://github.com/combatwombat/rmdb imports the IMDb database (at least the limited .tsv files they provide) into MySQL so you can query it. List the highest rated horror movies of the 90s, genre distribution by year etc. I made that mostly to c̶h̶e̶a̶t̶ ̶o̶n̶ help with https://www.reddit.com/r/GuessTheMovie, with limited success. Still fun though to SQL-query over all movies ever made.
> Although terminal emulation has been largely supplanted by the Web for online access, Kermit software continues to play a role in other applications such as remote sensing and data collection, management and troubleshooting of networking and telecommunications equipment, back office work, cargo and inventory management, medical insurance claim submission, electronic funds transfer, and online filing of income tax returns. Kermit software is embedded in network routers and switches, in cell-phone towers, in medical diagnostic and monitoring equipment, even in cardiac pacemakers, not to mention the cash registers of quite a few big-name "big box" retailers. In 2002 Kermit flew on the International Space Station, and Kermit software is the communication method used by EM APEX ocean floats (left) supplying realtime data to hurricane researchers and trackers to this day (the hurricane project entered a new expanded phase in 2010 based on a new version of Embedded Kermit).