Dreadful name, and definitely half baked.
Stick in some text in English and have it translated into a bunch of different languages and then generate audio of that same text in WAV format.
141–150 of 708 posts
Dreadful name, and definitely half baked.
Stick in some text in English and have it translated into a bunch of different languages and then generate audio of that same text in WAV format.
After recognizing the pretty steep learning curve many users face with Anki, including myself, I started working on munkle.it as a more intuitive alternative. The goal is to create the app that streamlines the flashcard creation process, prioritizing user-friendly design and simplicity, making it easier for learners to dive straight into effective study sessions without the initial overwhelm.
waggledance.ai! - source code: https://github.com/agi-merge/waggle-dance - buggy demo available @ https://waggledance.ai - really neat (IMO) adaptation of Plan-Validate-Solve - similar to AutoGPT and its ilk but with different architecture decisions - seeking co-founders/contributors
I want loosely coupled, functional, primarily no dependencies, practical, typescript first (if something can't be expressed in safe ts, it's too dynamic to belong there), easy to use and hackable/extendable packages. If they don't do i/o they need to work from browser ie. from observable, they're all dual esm/cjs.
Have plans for incorporating pretty much all AI image gen. capabilities, but only has image gen, loras, controlnet, background removal and upscale atm. There's like something new in AI image gen coming up everyday and I still haven't gotten lora training, checkpoint training, image colorizing, AnimateDiff and so much more done yet.
https://github.com/tikimcfee/LookAtThat View, search, and analyze arbitrary source code (best support of Swift right now) in 3D and AR space. You open your phone or tablet, yeet hundreds of files into 3D space, and can start highlighting, moving, and tracing execution by literally walking around your code. The desktop app has similar features, and the standard 3D viewer is just as fun. I would love help - from anyone…
Maybe, scale the size of the code based on how long it takes to execute that section. Would make it a cool way to find underperforming code.
Feel free to ping me more thoughts!
This is a thread I've been desperately needing. I've been working on a command line tool project for MacOS that I've been hesitant to post. I've been waiting until the project is 1.0 before posting on HN. Here it is currently! https://github.com/demyinn00/BootMe BootMe is a workflow manager app. There are 6 buttons that can be configured. 4 of the 6 buttons can be configured to open tabs and links, launch apps, and s…
Suggest you highlight clearly at the top that this is a command line tool for MacOS.
Abound: https://abound.art It's a generative art platform/marketplace. You write an algo that generates art according to some number of parameters, upload it as a Docker image, and let other people tinker and generate things they like the aesthetics of. Basically remote code execution as a service, but I spent a lot of time on the sandboxing side of things and feel pretty good about it. The infra is tuned for "saving…
This is really cool! How did you solve the sandboxing?
- Turn the Docker image into a Firecracker VM, I stole the idea from Fly.io [1]. Add all the trimmings like jailer and stuff, don't give it any network interfaces
- Run our own shim as PID 0 in the VM, which sets up a bunch of things to make the environment hermetic (time set to 1970, etc), and does some stuff with eBPF to monitor usage by the child process, and also enforces 1 minute timeout
- Run the jobs on a parent VM that doesn't otherwise have any privileges
- Copy images pixel by pixel (for raster images) or remove all the shady parts of an SVG that we don't otherwise trust
- Other general defense-in-depth stuff, validating request/response sizes, minimal privileges on separate services, private networking throughout.