Yep - up to 256 waveforms.
Fathom Synth Development Thread
-
- KVRAF
- 1579 posts since 8 Jan, 2003 from Edinburgh
John Braner
http://johnbraner.bandcamp.com
http://www.soundclick.com/johnbraner
and all the major streaming/download sites.
http://johnbraner.bandcamp.com
http://www.soundclick.com/johnbraner
and all the major streaming/download sites.
-
- KVRian
- 1303 posts since 21 Nov, 2018
Yep I think sending review copies to people like PluginGuru and Sadowick on youtube would go a long way to getting the synth some wider exposure
-
- KVRian
- 1367 posts since 21 Dec, 2013 from USA
The Unfinished, Triple Spiral, Soren (Luftrum), Sampleconstruct, etc. 
-
- KVRAF
- 3735 posts since 17 Sep, 2016
Windows 10 and too many plugins
-
- KVRAF
- 2094 posts since 22 Sep, 2016
Hi Everett, regarding WT. I think the 16 slot limit is to limited.
Tone2 Icarus for instance has a a mode where you can play a single slot one shot selected by key which allows to put drumkits into wavetables, for instance slot 0 is a kick, slot 1 a snare...If you use the method "every second is skipped at import" then you gonna lose content. You have to keep in mind, that WT are not required to mimik a linear fade thru modulations, neigbour slots could be drastically different. Thus leaving out content and try to interpolate the gaps is only good fo some WTs.
Icarus has some nice resampling options as well, where you can import an arbitrary sample and for instance "robotize" it. The importer will fill more than the 16 slots as well. If you leave stuff out, you gonna lose content.
I would suggest to rethink the 16 slots limit...
Don't know if you remember my OSC track but for the word "cybernetic" I used 4 patches a 16 slots...which was kind of a nice exercise but overall tedious...
Tone2 Icarus for instance has a a mode where you can play a single slot one shot selected by key which allows to put drumkits into wavetables, for instance slot 0 is a kick, slot 1 a snare...If you use the method "every second is skipped at import" then you gonna lose content. You have to keep in mind, that WT are not required to mimik a linear fade thru modulations, neigbour slots could be drastically different. Thus leaving out content and try to interpolate the gaps is only good fo some WTs.
Icarus has some nice resampling options as well, where you can import an arbitrary sample and for instance "robotize" it. The importer will fill more than the 16 slots as well. If you leave stuff out, you gonna lose content.
I would suggest to rethink the 16 slots limit...
Don't know if you remember my OSC track but for the word "cybernetic" I used 4 patches a 16 slots...which was kind of a nice exercise but overall tedious...
-
- KVRAF
- Topic Starter
- 1579 posts since 25 Mar, 2017
I'm with you. In Fathom X the number of waves per wave table will be settable by the user up to 1024 waves per table !
Changing the number will auto-extrapolate everything smoothly.
The only challenge is that this will drastically increase the buffer calculation times, but this will be mitigated to some extent by GPU processing which is 1024 threads in parallel (I already have this running in my Fathom X build) and also if the user selects a high number of waves the basic work flow will change to using a progress bar when all the buffers need to be recalculated for any reason. As long as the user can visually see what is going on I think people will acclimate to it.
Also, I forgot to mention there is a massive change coming out later this week in version 2.3.2.
I am adding an option to perform all buffer calculations in the UI thread and then doing a fast hot-swap of the offline buffers into the run-time buffers immediately after each sample block is finished. I've been testing it and this completely eliminates all audio glitches when you move envelope points, waveform points and all wave draw mouse clicks and drags. This is probably the most significant change in Fathom's processing in the last two years, with the exception of AVX, since it makes all editing operations audibly as smooth as silk, no glitches.
It is a setting so the user can enable it or disable it depending on how the host handles the relative priority of the UI thread and VST processing thread. But in most cases turning it on works beautifully and drastically change the users editing experience for their eardrums.
This architecture change will enable the massive increase in wave table positions since all the buffer calculation can be done in the UI thread which has oodles of time available and then hot swapped seamlessly into the processor between sample blocks when mouse movements are complete.
I'm almost done so this will be released this weekend before the AVX release which will be soon after.
Changing the number will auto-extrapolate everything smoothly.
The only challenge is that this will drastically increase the buffer calculation times, but this will be mitigated to some extent by GPU processing which is 1024 threads in parallel (I already have this running in my Fathom X build) and also if the user selects a high number of waves the basic work flow will change to using a progress bar when all the buffers need to be recalculated for any reason. As long as the user can visually see what is going on I think people will acclimate to it.
Also, I forgot to mention there is a massive change coming out later this week in version 2.3.2.
I am adding an option to perform all buffer calculations in the UI thread and then doing a fast hot-swap of the offline buffers into the run-time buffers immediately after each sample block is finished. I've been testing it and this completely eliminates all audio glitches when you move envelope points, waveform points and all wave draw mouse clicks and drags. This is probably the most significant change in Fathom's processing in the last two years, with the exception of AVX, since it makes all editing operations audibly as smooth as silk, no glitches.
It is a setting so the user can enable it or disable it depending on how the host handles the relative priority of the UI thread and VST processing thread. But in most cases turning it on works beautifully and drastically change the users editing experience for their eardrums.
This architecture change will enable the massive increase in wave table positions since all the buffer calculation can be done in the UI thread which has oodles of time available and then hot swapped seamlessly into the processor between sample blocks when mouse movements are complete.
I'm almost done so this will be released this weekend before the AVX release which will be soon after.
-
- KVRist
- 64 posts since 8 Mar, 2019
I'm glad you mentioned that. Being a latecomer, I missed that one. Just listened, and OMG That was Epic Peter!] Peter:H [ wrote: ↑Wed Jun 26, 2019 6:46 pm Don't know if you remember my OSC track but for the word "cybernetic" I used 4 patches a 16 slots...
Ambient Sound Design & Soundtrack Composition - http://ambietica.lyonsdenmultimedia.com/
-
- KVRAF
- 2094 posts since 22 Sep, 2016
Many thanks for your feedback AmbieticA! I think your not a latecomer...I think your a right in time comer...I think "fun" is only about to start with fathom... and that does not mean to say it's not hilariously capable already by now...but with sampler, more WT capabilities and better performance it's going to be a "weapon"...And this in combination with a profiscient patch/sound designer like you...perfekt time!AmbieticA wrote: ↑Wed Jun 26, 2019 7:56 pmI'm glad you mentioned that. Being a latecomer, I missed that one. OMG That was Epic Peter!] Peter:H [ wrote: ↑Wed Jun 26, 2019 6:46 pm Don't know if you remember my OSC track but for the word "cybernetic" I used 4 patches a 16 slots...
-
- KVRian
- 896 posts since 8 Aug, 2011
Hello! I bought Fathom at intro an upgraded twice since. Now I'm trying to figure out if it's worth upgrading now to get the 'free' updates.
The Pro version is not out yet if I get this straight ?
The Pro version is not out yet if I get this straight ?
MXLinux21, 16 Gig RAM, Intel i7 Quad 3.9, Reaper 6.42, Behringer 204HD or Win7 Steinberg MR816x
-
- KVRAF
- Topic Starter
- 1579 posts since 25 Mar, 2017
If you already bought Fathom Pro you already get all the free upgrades.
Just go here:
https://www.fathomsynth.com/updates
It does not matter how long ago you bought Fathom Pro, you always get free upgrades. The purchase upgrades for $5 are completely optional and are only there since some users requested a way to contribute regularly to feature development. It's fine to download the free upgrade.
The Intel AVX version is not out yet which is what we have been discussing, Fathom Pro has been out now for just over two years.
Just go here:
https://www.fathomsynth.com/updates
It does not matter how long ago you bought Fathom Pro, you always get free upgrades. The purchase upgrades for $5 are completely optional and are only there since some users requested a way to contribute regularly to feature development. It's fine to download the free upgrade.
The Intel AVX version is not out yet which is what we have been discussing, Fathom Pro has been out now for just over two years.
-
- KVRAF
- 3735 posts since 17 Sep, 2016
Based on the posts here, only AVX2 will be in the initial release, with AVX support coded in a later release. I have an Intel Ivy Bridge CPU, and that is AVX only, so I will be waiting.
And my understanding is that AVX and AVX2 will only be used for the detune voice processes in this scenario. So the performance gains will be focused on that.
And my understanding is that AVX and AVX2 will only be used for the detune voice processes in this scenario. So the performance gains will be focused on that.
Windows 10 and too many plugins
-
- KVRAF
- Topic Starter
- 1579 posts since 25 Mar, 2017
Yes, AVX2 will be the first release and AVX1 will be the very next release,
coming out absolutely as soon as possible after that, maybe a week or two at the most.
Yes, currently AVX is only used for detune, however this is huge factor since the complexity of detune was 99% of the reason for Fathom's CPU challenges. The detune loop was actually the central processing loop for all oscillators and this is the code that has now been completely replaced with Intel AVX assembly language instructions. So for all intense and purposes Fathom's oscillator processor is now an AVX processor. The AVX code also spread to some other critical areas such as waveform buffer access and buffer interpolation.
Basically if you create a preset with all mono oscillators with no detune and then crank up the detune voices on every oscillator there will be no change in CPU load since it is parallel.
My testing so far has shown that this will bring Fathom's CPU in the same domain as Omnisphere and Sylenth, and in some cases even a little faster since my previous tests showed a 4X lag behind them and the new AVX is a 5X multiply.
Depending on the response in sales I will be replacing more code with AVX such as the filter and effects processors.
coming out absolutely as soon as possible after that, maybe a week or two at the most.
Yes, currently AVX is only used for detune, however this is huge factor since the complexity of detune was 99% of the reason for Fathom's CPU challenges. The detune loop was actually the central processing loop for all oscillators and this is the code that has now been completely replaced with Intel AVX assembly language instructions. So for all intense and purposes Fathom's oscillator processor is now an AVX processor. The AVX code also spread to some other critical areas such as waveform buffer access and buffer interpolation.
Basically if you create a preset with all mono oscillators with no detune and then crank up the detune voices on every oscillator there will be no change in CPU load since it is parallel.
My testing so far has shown that this will bring Fathom's CPU in the same domain as Omnisphere and Sylenth, and in some cases even a little faster since my previous tests showed a 4X lag behind them and the new AVX is a 5X multiply.
Depending on the response in sales I will be replacing more code with AVX such as the filter and effects processors.
-
- KVRist
- 399 posts since 25 Aug, 2018
According to the forum link RPH posted, at least as of last December, Serum assumes 2048 samples per waveform in a wavetable. However it would be trivial to parse the clm chunk to get the actual samples per waveform in case this changes in the future. Then getting the number of waveforms in a wavetable would be a simple matter of dividing the data size by the samples per waveform.jbraner wrote: ↑Wed Jun 26, 2019 8:27 amSerum can use 512, 1024 or 2048 samples per waveform. I'm not sure how it "knows" the right length to use - but you can create a "FolderInfo.txt" file (in the wavetable folder) to force it to use a particular size.but from what I can tell, all the wavetables I found in the collection had 2048 samples per waveform.
It's in the Serum manual.
