Vanilagy

Mediabunny has exceeded 1 million monthly npm downloads holy shit

Never expected this to happen this fast, but here we are. It's just a number at first, but it's kind of surreal when you picture what it represents. Thank you! ❤️

You can tell Claude Code is written in React because it's slow and buggy

Mediabunny v1.31.0 ships with ID3 metadata read/write support for ADTS (.aac) files, and has a more resilient MPEG-TS demuxer. 🚀

I live in crazy times where I get emails like this

When the debug videos are so good you forget you're working on Mediabunny and just start watching them

Mediabunny's now at ¾ of a million monthly downloads!!! (almost) 🎉 Time to treat myself to some malatang

Finally added this to Mediabunny: a warning if the library is bundled twice. This is almost always unintentional and leads to runtime errors, so it's good if this mistake is surfaced early.

Using a symbol on the global object for the detection.

Genuinely refreshing to be on Bluesky and get non-evangelist viewpoints on AI.

I mean they're media files like any other!

"I did but I found it's located in San Francisco"

Goated reason not to do something. (not being sarcastic)

Mediabunny has world-class seeking performance across all file formats.

On the left, VLC *trying* to seek through a .ts file, it just kinda gives up.

On the right is Mediabunny. It just works.

I also came across Polar in my search for the "best MOR", have you checked them out?

Also yeah, tbh, screw using USD as a default currency. It's time we move away from that ASAP. I'd prefer a completely neutral currency tbh, the Euro also isn't that.

I find that LLMs often struggle with writing *confident* code that fully understands all invariants at play.

This code will work, but the highlighted condition is unnecessary (always evaluates to false). I always try to catch these cases to make the code tighter.

Ryan Dahl?

A huge milestone: Mediabunny v1.29.0 now ships with read & write support for MPEG Transport Streams! 🎉 This format is commonly used in live broadcasting.

Quite proud of the implementation. It supports AVC, HEVC, AAC, MP3, O(log n) seeking, and it's very fast.

HLS is next!

Mediabunny reached 5k GitHub stars!

Does anybody know if V8 (or other JS JITs) use hidden classes for common tuple types? So let's say I use a lot of [number, string, boolean], will these get optimized to a tight 16-byte struct or will these remain the slow, unoptimized array type?

B-frames are truly a mess with how out-of-order they can be; here I need to read ahead by SIX packets just to correctly compute the duration of the packet with timestamp 10.1. Very annoying! I might use the bound given by max_num_reorder_frames in the future.

Lovely! Have you considered condensing 'process' and 'upload' into a single step? Mediabunny supports stream conversion! But, this does limit your choice of output format somewhat.

Also thanks to @graham.systems for sharing mediabunny with me — I switched Anisota’s video pipeline over to it and its been so much better than ffmpegwasm so far!

https://mediabunny.dev/

It is a tool with strengths and weaknesses.

Been using Claude Code all day, and it has been the single best antidote against the Claude Code hype.

TypeScript is trying to type-check my MPEG Transport Stream videos hahahahaha

API learning of the day: You can use TextDecoder.decode with { stream: true } to ingest incomplete UTF-8 data in chunks!

They really said "modern" 💀

You know your algo crazy when

Dude just wait until the final mission. Absolutely blew my mind

Oh, you have your own to-do app!

Took a bit after the annual Christmas download dip, but Mediabunny has now reached 600k monthly downloads! Thank you guys 🐰

I don't have a day job right now! Working on Mediabunny and related projects full time, or at least, as full time as I'm able to at the moment 😂