zaptrem

23 ms·

241.
by zaptrem 2y ago
We also got Half Life Alyx which IMO lived up to the 3 hype with the exception of concluding the original story.
242.
by zaptrem 2y ago
They should use BigVGAN instead of HiFiGAN, afaik it’s better.
243.
by zaptrem 2y ago
My whole job is big-model GenAI music stuff and I'm still unclear what counts as "generative." Are loops in Ableton generative? What about that GarageBand drummer thing?
244.
by zaptrem 2y ago
Likely relevant to source of the living brain tissue: “Brain biopsies on 'vulnerable' patients at Mount Sinai set off alarm bells at FDA, documents show” https://www.cea.fr/english/Pages/News/world-p
245.
by zaptrem 2y ago
When interest rates are lower, money is easier to come by so terms are usually more favorable to founders.
246.
by zaptrem 2y ago
The currently publicly available model uses the old STT -> LLM -> TTS pipeline, so unless you’ve got early access you were using the above.
247.
by zaptrem 2y ago
Looks like they have exactly the same production target now they had in July of last year: https://www.reuters.com/technology/apple-forced-make-cuts-vi... "The report said Chinese contract manufacturer Luxshare, A
248.
by zaptrem 2y ago
Wow. I was expecting an edit that would further mock Apple but this super simple change massively improved the ad IMO while maintaining all of the cool ideas the original creators had.
249.
by zaptrem 2y ago
I spent three semesters in college learning RL only to be massively disappointed in the end after discovering that the latest and greatest RL techniques can’t even beat a simple heuristic in Tetris.
250.
by zaptrem 2y ago
Where I’m from we just call this hazing.
251.
by zaptrem 2y ago
You could also test this more easily with current events.
252.
by zaptrem 2y ago
What are your thoughts on Transit app? Seems to cover the different systems pretty well IMO.
253.
by zaptrem 2y ago
Our goal is for people to continue being the ones making the music. You'll still come to our site to see XYZ's newest release, it will just be more of a collaboration with AI, similar to how artists collaborate with each other and
254.
by zaptrem 2y ago
The paper says the environment is "browser compatible" but I don't see any in-browser demos we can try. Do you know of one?
255.
by zaptrem 2y ago
Tribes 2 would make an awesome modern game for players used to big 32v32 base-centered combat FPSes.
256.
by zaptrem 3y ago
Copilot’s prompt contains k-shot question answering examples. This is one of them.
257.
by zaptrem 3y ago
Iirc Tesla has a giant internal validation sets of edge-case-related disengagements they test against using a simulator whenever they’re gonna deploy a new version.
258.
by zaptrem 3y ago
That's no longer AI alone, you gave it the needed touch of humanity! That touch will take many different forms for different people.
259.
by zaptrem 3y ago
TikTok can tolerate a lot more active skipping than Spotify can before they annoy their users. We’d love to solve this. How would you? Maybe we could let users write why they didn’t like the song in natural language since we understand that
260.
by zaptrem 3y ago
I completely agree on the editing aspect. However if you want to generate five stem tracks, then all five tracks must have the full bandwidth of your auto encoder. Accordingly each inference or training staff would take much more compute fo
261.
by zaptrem 3y ago
Hacker News Song: https://sonauto.ai/songs/PEiO2sLZukHkpJLnHtmZ
262.
by zaptrem 3y ago
No, it's React (Native Web) Navigation... mistakes were made haha.
263.
by zaptrem 3y ago
Awesome to hear this resonates with you! If you join our Discord server I'll ping @everyone when improvements are ready.
264.
by zaptrem 3y ago
Never == "There will never be tears in my eyes as an AI sings ChatGPT-generated lyrics about the cycle of poverty a woman is stuck in ( https://en.wikipedia.org/wiki/Fast_Car ) because I know all of those experience
265.
by zaptrem 3y ago
More strength = force rhythm more. If you crank it to max it will probably result in just a drum line, so I prefer 3-4.
266.
by zaptrem 3y ago
Our variations feature will be nothing like Suno's (which just generates another song using the same prompt/lyrics). Since we use a diffusion model, we can actually restart the generation process from an early timestep (e.g., with
267.
by zaptrem 3y ago
The vast majority of our time was spent figuring out the model architecture and large-scale distributed training, and step 2 (starting now) is scaling everything up. Prompt understanding and audio quality will get significantly better once
268.
by zaptrem 3y ago
Agreed. My thoughts on this are here; https://news.ycombinator.com/item?id=39992817#39994616 Also, our model specifically excels at songs from the era before overproduction. Try asking for a Johnny Cash or Ella Fitzgerald-s
269.
by zaptrem 3y ago
For us it had nothing to do with collecting user data, adding what you mentioned would have just required another few hours of dev time haha. You’re right that it’s not hard to solve, we just wanted to focus on the rest of the app since the
270.
by zaptrem 3y ago
Our variations feature coming very soon is exactly this! Rhythm Control is an early version of this.

More ›