tonyabracadabra

pls bring it back

We built this so you can glimpse the future of real-time video generation, where stories render as fast as you imagine them.

If Sora 2 is aiming for AI‑Tok, ScaryStories Live is the jump-scare cousin: real‑time POV horror from a photo + a sentence. No film school, no GPU farm—just “upload face, pick fear level, go.” It’s less cinema, more haunted mirror, and it ships in seconds. scarystories.live

At Muku.ai, we set out to create AI influencer UGC ads. In the process, we built a customized pipeline that achieved state-of-the-art lip sync quality—all on a very limited training budget of just a few hundred dollars.

In a direct comparison with SyncLabs, a $6.4M YC-backed company, Muku.ai not only matched but surpassed their quality—all at half the per-minute serving cost.

We’re thrilled with these results and are preparing to launch our SOTA lip sync API soon. Check out our comparison video and let us know your thoughts—feedback is always welcome!

Is json a good file format for UI?

Is it similar to flutter?

I think that basically stretched the limit of what local model can achieve today, which also makes their image API almost useless for any serious generative art developers.

The image generation is dalle 2.5 level and feels really greasy to me, beyond that I think the overall launch is pretty good! I also congratulate rabbit r1 for their timely release months before WWDC https://heymusic.ai/music/apple-intel-fEoSb

Ok! Made a song about! https://heymusic.ai/music/apple-intel-fEoSb Hope you guys enjoy it!

Yes these licenses are common, but the dual licensing with essentially the same effect as a single license seems fishy and suggests there might be something hidden. Just my two cents.

The VGG License 2.0 is pretty confusing. It offers two licenses: VGG License 1.0 for commercial use (with royalties) and Apache License 2.0 for personal use only. But if personal use doesn’t involve making money, why add the Apache License? The original VGG License 1.0 could cover both personal and non-commercial use without royalties. Disregarding whether the royalty fee makes sense, the dual licensing setup just seems unnecessary and complicates things.

haha so oddly fun : P

One fascinating example in C programming is "Duff's device." This clever technique unrolls a loop to copy data from one location to another with remarkable efficiency. It's notorious among C programmers for its unusual and complex structure, often leaving them astonished that it not only compiles but also conforms to the C language standards.

And what’s weirder? You can now enjoy it in music!

AudioLCM needs only 2 iterations to synthesize high-fidelity audios, while it maintains sample quality competitive with state-of-the-art models using hundreds of steps!

The methodological unification of AI studies across different modalities is unavoidable.

Your observation on non-optimal random generative mapping misses a crucial point, the key challenge lies in establishing a topological invariant mapping, which ensures that the intrinsic structures and properties of the input data are preserved in the output. This becomes even more complex when dealing with higher-order structures that span multiple modalities.

yea indeed music is very personal, I also found 99% of musics on Spotify equally bad if you just opened a new account so the recommendation system knows nothing about the distribution of people you represent, so i guess it is more of a math problem than the quality issue AI music faces today

interesting! can the similar measurement be applied to finding redundant code (like low entropy) with extra works?

yea indeed, but think ai music as a collective mind of human music creation, does that make you feel better?

The pleasure is all mine!

The lyrics explains it all

wild https://gpt4o.ai

Very cool, and what’s cooler is this rap about alphafold3 https://heymusic.ai/blog/news/alphafold-3