fbnt
Joined 16 years ago
Joined 16 years ago
In any case, 1hr of iPhone/day is way too strict for a teenager nowadays. I was lucky to be the 3rd child, and my parents didn’t really paid attention to how I was spending my time. I overdosed on a bunch of things as a kid: TV, videogames, my computer, the internet when I got it (I’m in my late 30s), but then I got over them, except computers, which became my profession.
On the other hand, people my age that were only allowed one hour of Nintendo per day as kids, are still craving the next PlayStation like grown up junkies. Last time I played a videogame I was maybe.. 22.
It's a iPhone news aggregator that works in the opposite way of traditional news reading apps. It uses Twitter as a sort of enhanced RSS feed and ranks news stories based on how much they are shared in real-time, laying them down in a newspaper-like format specifically designed for mobile devices.
It's really interesting because it lets you find what's important for people rather then what matters to newsroom's editors, and it puts you in a whole different point of view.
The bare bone MVP is currently available in US, UK and Italy. I'll soon release a new version with categorization and custom topics.
I've been working with video encoding, subtitles and basically anything that is "layer things on top of videos" for many years now, and lately I started wondering if there was a more efficient way to do it without relying on the usual methods and libraries. Basically, every time you paint stuff over a video, you create the overlays with 'some' tool (something that generates cards as PNGs, or a headless browser for more complex animated designs.. or other techniques), and then use good old ffmpeg to merge the composition together, or eventually use ffmpeg's complex filtergraphs with all their quirks and limitations if you're desperate, but nobody does that anymore.
But what if ffmpeg could do proper composition, natively? That's probably a bit out of scope for them, so I set out to start rewriting ffmpeg's CLI in Rust with some extras.. because, why not. Of course, the codecs still comes from libav*, but everything above it is new: a planner that figure out the cheapest way to produce the output (frames with nothing drawn on them aren't re-encoded), a compositor with layers, keyframes, masks and transitions, and a layout engine for the part of CSS that is most commonly used: flexbox, gradients, shadows, clip paths, blend modes, and @keyframes played as written.
Now I have something that paints pretty animated graphics natively over videos very quickly, and without using playwright or chrominum. You can see some benchmarks here against two of the most common methods used to achieve this here: https://github.com/geneva-render/geneva/blob/main/docs/bench...
As you probably guessed, this was all written by Fable/Opus under my guidance. Just stating it plainly so you can decide whether to trust it or not. This is just the engine, and I have a couple of ideas on how to move it forward, but I'd like to hear some feedback before carrying on blindly, so feel free to reply here or get in touch.
Some rendered examples are in the repo and here: https://genevarender.com