Abyss released
-
- KVRAF
- 3231 posts since 18 May, 2003 from Sweden
Abyss is an adventure in sound, whatever the underlying tech. I've still only started the discovery phase, but this is going to be fun! A keeper, for sure.
If it were easy, anybody could do it!
-
- KVRian
- 666 posts since 18 Sep, 2010
Tracktion Support followed up and restarted the 90-day trial. Great!
Really impressed with the wide sound palette (since we're into the artistic aspect ) of Abyss. This may well be a keeper.
-
- KVRAF
- 4218 posts since 15 Sep, 2010
Sounds -and looks- quite gorgeous and unique. Wow.
- KVRist
- 264 posts since 17 Oct, 2005
The sheer amount of inspiration derived from the finely crafted UI is something that other developers should take note of. Abyss is a game changer in that the interface completely gets out of the way. I have not had this much fun with any instrument in quite a long time!
Duality without regard to physicality
-
- KVRAF
- 4218 posts since 15 Sep, 2010
c0nsilience wrote: ↑Tue Jun 29, 2021 6:28 pm The sheer amount of inspiration derived from the finely crafted UI is something that other developers should take note of. Abyss is a game changer in that the interface completely gets out of the way. I have not had this much fun with any instrument in quite a long time!
- KVRAF
- 2838 posts since 8 Jun, 2018
when i am back in the studio, i shall do the demo... i am very curious...
Primoridal Music: sadà\exposadà - Indusrial & Expanding Your Mind Hurts: Sound Brut
- KVRAF
- 8826 posts since 6 Jan, 2017 from Outer Space
I would like to ask peter, about the technology behind this brilliant design. Others suspected that it uses samples, I don’t believe that. The size is too small for it…
Now my guesses: the user interface for the sound field reminds me on concatenative synthesis (IRCAM) or the AI sorting of sounds in Atlas/XO. I guess its a bunch of wavetables sorted by a neural network to find useful paths between two arbitrary points within that sound field. Though it sounds a lot like granular synthesis and neighboring points in the editor do not sound very similar… Maybe its something inbetween…
Now my guesses: the user interface for the sound field reminds me on concatenative synthesis (IRCAM) or the AI sorting of sounds in Atlas/XO. I guess its a bunch of wavetables sorted by a neural network to find useful paths between two arbitrary points within that sound field. Though it sounds a lot like granular synthesis and neighboring points in the editor do not sound very similar… Maybe its something inbetween…
- KVRist
- 90 posts since 11 Aug, 2015
Thanks for the interest! The topic is a bit complex ... hence the answer is a bit long:
Every single tone-colour is a spectral model. These models have been derived from natural samples. This has a couple of benefits: the models capture the fluctuations of natural sounds, hence they sound very organic and have a nice texture. At the same time they are not fixed samples, but can be modified. In the current version you can add / remove detail, add "musical" noise which is specifically tuned to the tonecolour, increase the fluctuations or make the sound static like an oscillator voice. This is just the tip of the iceberg - the models are quite flexible and other parameters may be added in later versions (if there is interest). The underlying spectral model used in Abyss is based on my own research and is in many aspects different from the usual "STFT" or harmonic+noise+residual approaches.
Careful listeners here in the forum suspected that the engine is based on samples. That is also true to some extent: the spectral model is quite CPU intensive, but I wanted Abyss to be less demanding on CPU. Hence when you click / select any tone-colour the spectral model is rendered into a multisample, so when playing the synth these multi-samples are played back. With more and more CPU available I will at some point extend the technology to work directly on the spectral models. This will allow realtime modulations of the spectral model. But you can achieve this with ease in Abyss already today: simply put multiple variations of tone colours on the gradient - there is no practical limit on how many tone-colours you place on the gradient. In practice I have rarely needed more than a handful tone colours to make even the smoothest transitions.
The tone-colours are sorted into the 2D "surf" area with help of an "artificial hearing model". Based on this model sounds that are similar are placed close to each other and also have similar tone colour. However, your inner representation of sounds is not two dimensional! Actually the way our ears and brains resolve and understand timbre is quite amazing and fascinating topic, and an active area of research. We know for sure that our inner representation of sound has at least 20 dimensions, probably more. So any projection onto two or three dimensions will necessarily place different tone colours close to each other. This is bad for scientific approach of remodelling sound, but for creative sound design I think it is even a feature: the 2D map _roughly_ places similar timbres in proximity, but not with mathematical precision, hence you have many pleasant surprises. Also "surfing" the map is so speedy that you can "browse" through dozens of sounds in matter of seconds, which is fun and effective at the same time.
I hope this demystifies the technology. There will be an "ask me anything" livestream on Wednesday 7th July - feel free to join and ask more questions!
Every single tone-colour is a spectral model. These models have been derived from natural samples. This has a couple of benefits: the models capture the fluctuations of natural sounds, hence they sound very organic and have a nice texture. At the same time they are not fixed samples, but can be modified. In the current version you can add / remove detail, add "musical" noise which is specifically tuned to the tonecolour, increase the fluctuations or make the sound static like an oscillator voice. This is just the tip of the iceberg - the models are quite flexible and other parameters may be added in later versions (if there is interest). The underlying spectral model used in Abyss is based on my own research and is in many aspects different from the usual "STFT" or harmonic+noise+residual approaches.
Careful listeners here in the forum suspected that the engine is based on samples. That is also true to some extent: the spectral model is quite CPU intensive, but I wanted Abyss to be less demanding on CPU. Hence when you click / select any tone-colour the spectral model is rendered into a multisample, so when playing the synth these multi-samples are played back. With more and more CPU available I will at some point extend the technology to work directly on the spectral models. This will allow realtime modulations of the spectral model. But you can achieve this with ease in Abyss already today: simply put multiple variations of tone colours on the gradient - there is no practical limit on how many tone-colours you place on the gradient. In practice I have rarely needed more than a handful tone colours to make even the smoothest transitions.
The tone-colours are sorted into the 2D "surf" area with help of an "artificial hearing model". Based on this model sounds that are similar are placed close to each other and also have similar tone colour. However, your inner representation of sounds is not two dimensional! Actually the way our ears and brains resolve and understand timbre is quite amazing and fascinating topic, and an active area of research. We know for sure that our inner representation of sound has at least 20 dimensions, probably more. So any projection onto two or three dimensions will necessarily place different tone colours close to each other. This is bad for scientific approach of remodelling sound, but for creative sound design I think it is even a feature: the 2D map _roughly_ places similar timbres in proximity, but not with mathematical precision, hence you have many pleasant surprises. Also "surfing" the map is so speedy that you can "browse" through dozens of sounds in matter of seconds, which is fun and effective at the same time.
I hope this demystifies the technology. There will be an "ask me anything" livestream on Wednesday 7th July - feel free to join and ask more questions!
-
Echoes in the Attic Echoes in the Attic https://www.kvraudio.com/forum/memberlist.php?mode=viewprofile&u=180417
- KVRAF
- 11053 posts since 12 May, 2008
Thanks for the explanation. Looking forward to seeing how it evolves!
-
- KVRAF
- 7020 posts since 26 Jul, 2018
yes, thanks for that explanation. Interesting read!
- KVRAF
- 35276 posts since 14 Sep, 2002 from In teh net
Sounds a bit like the old Plex and PPG Wavemapper.
- KVRAF
- 8826 posts since 6 Jan, 2017 from Outer Space
Wow, very interesting and inspiring. Would animate me to do some experiments in Max/MSP or the Grid…
- KVRAF
- 2838 posts since 8 Jun, 2018
indeed wow, and in real live, and with MPE, indeed made for MPE (CC114, but why not also 113 and 115... XYZ on a roli block, small detail, not important now: EDIT can be changed..).
how long with this intro pricing last? (the term STFT i didn't encountered before, even in conversations in developers... i learn..)
only working with it now, for 10 minutes... so give me some more hours, and i shall break... even if i can do with Biotek 2, a lot, a synth, next to Generate, i use next to Abyss, to dive into the deep.. o no.. and shall compare to other synths, in one project, it will stand out, i am sure...
i thought the time of interesting new niche synths was, or that there was strange pause, after Plasmonic, Generate. (i think i forget 1 or 2..)
yes, Sumu comes... but still...
nice to see, a synth that goes takes its own path.
how long with this intro pricing last? (the term STFT i didn't encountered before, even in conversations in developers... i learn..)
only working with it now, for 10 minutes... so give me some more hours, and i shall break... even if i can do with Biotek 2, a lot, a synth, next to Generate, i use next to Abyss, to dive into the deep.. o no.. and shall compare to other synths, in one project, it will stand out, i am sure...
i thought the time of interesting new niche synths was, or that there was strange pause, after Plasmonic, Generate. (i think i forget 1 or 2..)
yes, Sumu comes... but still...
nice to see, a synth that goes takes its own path.
Last edited by WasteLand on Thu Jul 01, 2021 3:03 pm, edited 1 time in total.
Primoridal Music: sadà\exposadà - Indusrial & Expanding Your Mind Hurts: Sound Brut
-
- KVRAF
- 1573 posts since 28 Jul, 2006
Wait, this has a 90 day free trial??