When's the last time your IDE made a sound?

When a new JavaScript feature comes out, I always think, "Who the hell needs that?" Async generators? WeakRef?? FinalizationRegistry??? Symbol.dispose???? And then I always end up finding a use for them a few years later. What's your favorite "niche" JavaScript feature?

Super excited to announce that I'll be speaking at Demuxed in London this October! 🚀 I'll share Mediabunny's origin story and some of the decisions that shaped its design. My first real conference talk, so I'm excited. Big thanks for the Demuxed team for giving me a slot! 💘

You guys are insane. 2.5 months after its release, Mediabunny has already hit more than 100k monthly downloads. Thank you! 🐰📈

Another big one: Mediabunny v1.17.0 now ships with full read & write support for .flac files! 🚀 Huge thanks to Jonny Burger for taking care of this one! ▶️

Neat small blog post I found today, for anybody trying to build code to run both in the browser and in Node. Learned about the "browser" package.json field! It allows you to remap imports if the code is bundled for a browser environment. zirkelc.dev/posts/bundli...

I feel like most developers nowadays don't even notice complexity. I was sent a repro for a Mediabunny bug in the form of a Next.js app. What could've been a 50-line HTML file is instead a 0.5 GB (!!) behemoth. It's wild to me that this doesn't make people shudder.

You know it's over when you start dreaming about audio codecs.

The world would be better with a little less robustness principle

Mediabunny v1.15.0 ships with cropping functionality in the Conversion API! 🚀 You can now convert videos and only keep a specific rectangular region of the original input video. It also works nicely in conjunction with the rotation and resizing options.

The world is full of signal and full of noise. Which do you choose to contribute?

An interesting phenomenon is how people (even those with no media background) are quick to write off Mediabunny as an FFmpeg wrapper - if not directly, then surely indirectly through browser APIs. I attribute this to FFmpeg's strong image. Happy to expand yalls world view, tho🫰

If you want to compress videos before uploading them, make sure to use Mediabunny's stream-upload feature, where the video is compressed and uploaded at the same time. If the network is slow, the compression will automatically adjust its pace as well. Super cool!

If writing software is building a tree of features, there are two approaches: - Depth-first: Build out every feature to 100% completion before moving on - Breadth-first: Work on all features at the same time, adding detail and polish later on Which approach do you practice more?

You can play with Mediabunny's new metadata tag extraction feature in the metadata extraction example on the website. Here's it extracting title, artist, cover art and more from a song!

My lovely girlfriend takes me out for all-you-can-eat sushi for every GitHub star milestone of Mediabunny. This one was for 1k stars. She said the next one happens at 10k. Wonder when that will be 🤔

Mediabunny v1.14.0 is out and it's a huge milestone for the project! 🚀 Mediabunny can now read and write metadata tags (like title, artist, album, cover art) for all formats, including ID3v1/v2, iTunes, QuickTime, Matroska, Vorbis, and RIFF-style metadata.

Spotify's "multiple devices, one state" approach is so utterly genius. Instead of a "you can't use Spotify on multiple devices" error, they turned it into a feature. Now I can just control the playback on my PC from my phone. Love it

Mediabunny's website now has API docs! 🚀 Complementing the guide, these additional 148 pages document every class, method, property and type in Mediabunny's public API. They're pretty sexy, so, go check them out! mediabunny.dev/api/

Ah yes

Mediabunny v1.13.0 adds ReadableStreamSource, offering a way to pipe append-only byte streams into Mediabunny. This enables many new use cases, such as real-time processing of MediaRecorder output while the recording is taking place. 🚀 See the docs: shorturl.at/Yo2KF

Who even needs <video>? I certainly don't 👀

The upcoming version of Mediabunny will ship with a built-in way to read directly from the file system, for server-side use. Works in Node, Bun and Deno! (interestingly, Deno comes out on top performance-wise)

Mediabunny has now already overtaken mp4-muxer, my previously most-starred repo which I released only 2 years ago. Thank so you much for all the support and kind words, it has really meant a lot. Super excited for where this goes over the next year! 🚀

Mediabunny can generate videos at ludicrous speeds. This 720p video is rendered & encoded at around 500 FPS, fully client-side:

Do you guys prefer API docs that live in a single, giant document (optimized for Ctrl+F), or do you want them to be organized into multiple pages?

Mediabunny v1.9.0 adds support for reading & writing ADTS (.aac) files! 🚀 It also fixes some nasty race conditions in the MP3 demuxer.

Used @vanilagy.bsky.social 's Mediabunny in supermosh.github.io for rendering. Perf gain for most devices (6s -> 0.3s for this video), but most importantly since I don't rely on the live <canvas> recording anymore, the output is deterministic & the same for all devices. Many ppl asked for this :)

I've heard Big Buck Bunny in the most cursed ways due to audio rendering bugs. This is the first bug that acutally sounds quite cool?? Besides the bird getting smacked at the end there

Mediabunny v1.8.0 adds better chunked encoding for long AudioBuffers, and now supports way more sample rates for @mediabunny/mp3-encoder! (8000, 11025, 12000, 16000, 22050, 24000 are the new ones)

I'd love an LLM-backed "no-outdated-comment" ESLint rule that revalidates comment validity for comments around the latest diff. Would be crazy. I'm like 99% good at catching outdated comments but somehow, one always slips through.

Nothing feels more like "shouting into the void" than posting on Bluesky. Literally - is anyone here? *echoes* H-hello?

Vectorized interleaved to planar conversion in Zig. Runs like 8x faster than the scalar version! This was super fun to write! Comptime computation of the shuffle mask, a bunch of inline switch and for - Zig makes me feel like I've collected all Infinity Stones of programming.

Things are happening in Web Audio 👀 Firefox is *finally* implementing suspend()/resume() on OfflineAudioContext; methods which are basically required for hack-free progressive audio rendering. Thank you @padenot! bugzilla.mozilla.org/show_bug.cgi...

Mediabunny v1.7.0 adds track-specific options to the Conversion API for more control over multi-track files. It also adds streamable (append-only) MP3 writing, better MP3 reading and a couple of fixes. Let's get it 🚀🚀🚀

Mom: We have TypeScript at home

Excited to release Mediabunny's first extension package: @mediabunny/mp3-encoder 🚀 This now gives Mediabunny the ability to encode MP3 directly in the browser, a feature that no browser supports. It uses a super-fast SIMD build of LAME that runs in a worker and is only 130 kB.

Here's a nice visualization of parallelism when Mediabunny converts an MP4 to an MP3: While the main thread parses the file and decodes the audio, a worker does the expensive MP3 encode. These run in lockstep, meaning a slow encoder will throttle the decoder on the main thread.

Just released Mediabunny v1.5.0! 🚀 You can now transform the frame rate of the video in the Conversion API, a requested feature. It also ships with a couple of fixes. Try it out!

I have a fucky method of debugging regressions in the media files produced by Mediabunny: I convert both the old (working) and new (broken) file into hex, bytes separated by newlines, and then pipe it through diffchecker.com 😂 It does the job! I even get the byte offsets

Recently implemented fast video seeking for a client using Mediabunny. Feels almost illegal how silky-smooth this is 🤤

Just released Mediabunny v1.3.0 (🚀), which includes a bunch of fixes and new features around determining if a frame is a key frame or a delta frame. For this, Mediabunny now pokes into a packet's encoded bitstream and, depending on the codec, figures out what frame type it actually is.

Mediabunny v1.2.0 adds support for reading and writing RF64 .wav files! 🚀 RF64 allows .wav files to exceed a length of 4 GiB.

This comment on my HN post reinforced a belief I already had: that it's important to allow direct usage of your libraries without going through a package manager. This is why I offer fully-built ESM & UMD artifacts of my libs for direct download. The choir has been preached to.

It was important to me that Mediabunny behaves consistently for all files, even the weird ones. For the weirdly-formed MediaRecorder .webms, Chrome just gives up and has no clue about the duration. Mediabunny just... displays the correct duration!

Mediabunny is general purpose enough that you can literally use it to build your own <video> element from scratch. Here's mine playing Big Buck Bunny off my disk! 100% pure TypeScript, no <video>, <audio>, WASM or anything. Makes it perfect for building high-performance yet high-precision editors.

I built a new library! It's called Mediabunny and is a TypeScript library for reading, writing, and converting video and audio files. It's written from scratch for the web and has zero dependencies. It supersedes my old library mp4-muxer and webm-muxer. Check it out! mediabunny.dev

Decided to also create an account here! Seems like a mostly nice place to be, although I am surprised by the NSFW stuff on my feed already 😂