This is the first Scratch project that can download and play the audio of any YouTube video!!! Try it out: https://turbowarp.org/1049044865?clones=Infinity&limitless&fps=120 Potential bug: If it says "queued for synthesizing" for too long, restart the project. I highly recommend using TurboWarp for better quality. Run the project at 120 FPS and with infinite clones enabled (automatically done when using the above link). ➥ Notes: Can be used by multiple people simultaneously (no hard-coded limit). When using this project on TurboWarp, select the "US East" cloud server (selected by default). Audio marked as aged-restricted won't be rendered by this. Other people can't hear the audio you're hearing. (scroll down for Safety Notice) Getting this to work flawlessly and efficiently took longer than expected. The backend code will be published on GitHub soon. I think there's still room for optimization in terms of sound quality. ➥ Explanations: There's a python bot listening to TurboWarp's cloud variables and detecting requests (made with my python library scratchattach). When a request is received, the audio is downloaded from YouTube (using pytube), cropped and converted to WAV (using ffmpeg), analyzed using ARSS (https://arss.sourceforge.net/) and then the resulting data is sent to the project using cloud variables. Credit to @gilbert_given_189 for the base of the music renderer. While rendering music, the project plays 384 frequencies (ranging from 50 Hz zu 12.8 kHz, on a logarithmic scale) concurrently. ARSS provides the volumes for each of these frequencies, these are received using complex math (FFTM / Fast Fourier transforms on multipoles - this goes way beyond anything I ever learned at school, luckily ARSS does the analysis almost fully automatically). The volumes are updated 30 or 60 times per second (depending on the selected audio framerate). This creates the sound you hear in the project. The backend is hosted on a hosting platform provided by @justablock in a docker environment. The audio is only streamed because the analysis / synthesis process with ARSS is also streamed on the backend (the audio is cut in pieces that are analyzed seperately), this is to make sure you won't have to wait until the whole audio has been analyzed. The process is very ressource-intensive - to prevent the server from exceeding its ressource limits, some analysis processes will be queued when the server is under high load. The backend code will be published on GitHub soon. ➥ Safety notice: A Scratcher has raised concerns about whether this project is safe for Scratch. I take these concerns very seriously. This project doesn't render age-restricted videos, therefore it should be safe for Scratch. Scratch even has its own built-in video viewer ( Furthermore, after all the project will only render the videos that you tell it to render (if they are not marked as age-restricted). It is NOT possible to share videos through this and it's also NOT possible to see what videos other users were watching. This project has already gone through the Scratch team's moderation process and has been marked as NFE (the API allows me to see that). That means the Scratch team is okay with this project's existance, but considers it to be in a "grey zone" (or, according to the Scratch wiki: "Nfe projects are deemed inappropriate for the community, but not inappropriate enough to be censored or removed" More info on NFE projects: https://en.scratch-wiki.info/wiki/Project_Review_Status ➥ Credits: The project (the audio rendering and the creation of the freqs list) is based on this amazing Scratch project: - Dependencies used in the backend code: - ffmpeg - ARSS (https://arss.sourceforge.net/) - pytube for providing the free hosting for the backend server
