miguelnegrao

Joined 1 year ago

Karma

Submissions

Member since

May 29

Recent submissions

Don't such effects already upsample and downsample as needed internally ? You don't need to waste cpu on the effects that don't need higher sampling rate, right ?

Comment · 1 month ago

What do you recommend for destructive editing which is open-source ?

Comment · 2 months ago

Wouldn't most linux users just use codex cli or codex vscode ? Does the app bring anything not available on codex ?

Comment · 2 months ago

If by solution you mean a proof and by testing you mean encoding it in lean and compiling it, the space of possible syntactically correct proofs which you can encode probably explodes in a way that is well beyond what any computer could try to brute-force. LLMs don't brute-force proofs, i believe their approach is quite similar to humans. I believe the same is essentially true for counter-examples of the type that have been found latelly, they are not found by search, but by using theory.

On the other hand even if the compute allocated by openai is esquivalent to day 10 human mathematicians, the machines can work 24h per day, that is already a lot more productive.

Comment · 1 year ago

For VR, since you have to put the helmet in your head, it is usually assumed users don't mind putting headphones as well, so most solutions are via headphone-based spatialization.

Comment · 1 year ago

Actually, the farther way the speakers are from the angles specified in the 7.1 format (see https://www.dolby.com/about/support/guide/speaker-setup-guid...) worse will be the localization accuracy. And if the the person is not sitting centered relative to the loudspeakers, but closer to one of the loudspeakers localization can completely collapse, and it will sound like the sound only comes from the closest loudspeaker.

In the case of gamers, they are usually centered relative to the loudspeakers, and usually the loudspeakers tend to be placed symmetrical to the computer screen, so the problem is not so bad.

For cinema viewers sitting in the cinema the problem is much worse, most of the audience is off center... That is why 7.1 has a center loudspeakers, the dialogue is sent directly there to make sure that at least the dialogue comes the right direction.

Comment · 1 year ago

Unreal Engine, the only engine I'm more familiar with, implements VBAP which is just amplitude panning when played through loudspeakers for panning of 3D moving sources. It also allows Ambisonics recordings for ambient sound which is then decoded into 7.1.

For headphone based spatialization (binaral synthesis) usually virtual Ambisonics fed into HRTF convolution is used, which is not amplitude based, specially height is encoded using spectral filtering.

So loudspeakrs -> mostly amplitude based, headphones not amplitude based.

Comment · 1 year ago

I believe Atmos as used in cinema rooms, is as far as I know amplitude based (VBAP probably), and it is impressive and immersive. Immersion depends more on the number and placement of loudspeakers. Some systems do use Ambisonics, which can encode time differences as well, at least from microphone recordings.

HRTF as used in binaural synthesis is for headphones only, not relevant here.

Comment · 1 year ago

Tihs is true, but a high density of loudspeakers allows the use of Wave Field Synthesis which recreates a full physical sound field, where all 3 cues can be used.

Comment · 1 year ago

I'm sorry, but this is not accurate at all. Using "only" two signals, humans are quite good at localizing sound sources in some directions:

Concerning absolute localization, in frontal position, peak accuracy is observed at 1÷2 degrees for localization in the horizontal plane and 3÷4 degrees for localization in the vertical plane (Makous and Middlebrooks, 1990; Grothe et al., 2010; Tabry et al., 2013).

from https://www.frontiersin.org/journals/psychology/articles/10....

Humans are quite good at estimating distance too, inside rooms.

Humans use 3 cues for localization, time differences, amplitude differences and spectral cues from outer ears, head, torso, etc. They also use slight head movements to disambiguate sources where the signal differences would be the same (front and back, for instance).

I do agree that humans would not perceive the location difference between two pixels next to each other.