What about the „less gain in captures“ findings by Youtubers who compared the capture to the real thing? I think Pete Thorn and some other guy mentioned this. Any thoughts on this?
Cheers, Todde
I caught that too. it seems like an easy compensation: either process the Capture a little hotter on the level control… or once you have your Capture just turn it up a bit. as Pete said “six of one a half dozen of the other.” I also think this Neural Capture algorithm will get tweaks i.e. a work in progress.
1 Like
Agreed- I’m betting that the tweaking is already done and in testing now. 
1 Like
I am doing a lot of captures (in the thousands) and the annoying thing about this issue is that it makes the comparisons with the reference amp harder. There was another larger thread on this issue: Quad Cortex Captures lower gain and lack of dynamic response
It was closed with this explanation: "Please email support@neuraldsp.com so they can provide some tips and help you troubleshoot what might be causing it. Often times when user’s aren’t happy with their captures it just takes some experimenting. Many community members find gain staging is one of the most important parts of the capturing process.
Send us an email and also discord can be a good place to connect with the community for capture tips. People share their methods there if you want to talk directly with other users too."
So the issue was formulated in a way like if it would be a user error/issue."
I was very surprised that this thread with 43 posts was closed because l can tell you this: the issue is not resolved and it is not a user error and i have confirmed this with 3!!! hardware units myself and i have had this issue with 30 amps. Some more noticeable some less.
I have contacted support and they were very responsive, which i admire, but the recommendations where off target again and again. It took a lot of e-mails and two video demonstrations in the end to get this to the developer level. This is when i received a proper answer which stopped me looking into possible user errors and let me focus on workarounds.
Here is the answer i got after the developer team was consulted:
"We just noticed that the Master of the amplifier in the video is almost all the way up. I guess similar settings are being used on the other amps you mentioned: Super Reverb, JTM45, AC30, Small Box 50, etc. It seems the amplifiers have a lot of power amp distortion.
Although it is true that Neural Capture is at the cutting-edge of black-box modeling, it is also true that we are still not done with it yet. There are a few key areas we want to improve with regards to accuracy under specific amp-settings/cases and even types of devices that can be Captured with reasonable accuracy (Fuzz pedals and even time-based effects such as reverbs).
We are working on all of the above and please rest assured that as soon as these improvements are ready to be deployed, you will have access to them via free-of-charge CorOS updates."
It is not an answer i wanted to hear, but it is a straight up proper one and i would have appreciated to get to know this sooner/easier.
I still feel like the results are very good and have learned to work around the issue by increasing the gain with a clean preamp after the capture out jack during capturing. That resolves the gain difference. For most amps about 4db of level increase does the trick.
There is still the last 5%-10% which the Cortex seems unable to capture, probably stuff happening in the power amp - slightly less dramatic low-end thump (which might be what some others perceive as less dynamic), a bit less mid frequencies with the amp cranked (which seem to build up in the power amp on high levels). Slightly less of that presence bite. But we are talking about nuances - it still beats out anything else out there easily.
So i am fine with that after receiving a straight answer. That other thread did not deserve to be closed - it needs a straight answer and proper tips for workaround(s). Hope this one does not get closed, as the issue is not resolved. Also i do not know if the issue can be solved, as it must be technically extremely difficult to measure a system only from one injection point - the input, without the possibility to inject test signals for example right before the power amp. Maybe a Cortex Pro version for the guys who know where to inject the signal, even if there is no fx loop? 
Meanwhile what i would wish for from Neural - as there is this issue - PLEASE make the gain knob for the capture available already in the comparison screen to the reference amp and PLEASE let the user “hard code” the gain setting in the capture (and while we are at it, same goes for the level control, as sometimes there are also unexpected level differences). This would be the easiest workaround for the time being. On your own device you can set the default knob settings, including gain for a capture, but they are lost if you share it.
Hope this helps anyone still wondering if they are in an “intel inside idiot outside” situation, as i was for quite some time.
4 Likes
Sorry about the previous deleted post (I wrote something far too long-winded & decided to remove it 
).
My experience is that the amount of gain present in a capture can go either way depending on the amp in question. It’s also massively affected by how you have the input level set on the QC. There’s often an issue of gain balance as well (match the gain in the low-end & there isn’t enough in the high-end, match the high-end & there’s too much in the low-end). The previous poster is bang on regarding the EQ differences too.
Having said all of that, one can get it pretty darn close after the fact by using one’s ears along with the controls provided.
1 Like
just increase the amp gain - the capture signal is always the same level, so if your input signal isn’t calibrated just right to match (which it might not be since different instruments have different levels, and the input level on the QC itself might be adjusted), then the gain will be different a bit. There is no refinement or gain matching after the capture, so you’ll just have to adjust it manually either on the amp ahead of time or the capture afterwards.
2 Likes
If you want to get the gain levels correct without the need for adjustment after the fact then it requires a bit of experimentation with the global Input level (found in the I/O window) & you should totally Ignore the ‘Instrument In’ control found in the capture window. I feel that most people set their levels a bit too low & that’s the main culprit of lower gain levels than the target device.
1 Like
I get your point, but what do you do, when the amp is on volume 10 (which in many cases sets the level of gain on the amp too). From my experience there is no effect on the capture how you set the instrument level - it is there to just give you a reference, nothing more. The IN level on the capture screen just sets the VOLUME of the capture. So on the capture you have no control over the signal strength going into the reference device. After the capture is done, you can just compare it to the reference. Again, no option to increase and hardwire the increased gain to the capture. You can do it only afterwards, set it as a default setting, but it is lost, when shared with others.
So there is no options inside the device to circumvent this, at least to my knowledge.
I talked about this with Neural support and tested it, but, unless there was a user error on my side, all it does is basically adjust the “floor” of the levels - it does not solve the issue that i end up with a capture that has about 4dB less signal on the input (at least perception-wise) compared to the reference device.
Yes, Instrument IN is just for reference - you can test how the amp setting behaves with more boost or less, etc., but there is no effect on the captures gain or level.
I had the opposite issue with drive pedals. My captures were all much gainier than the pedal when I was running it with identical settings in front of the QC. It turned out that I had my global input level set to +10dB so effectively the capture wasn’t gainier and the capture itself was fine but I was running a hotter signal into the capture block compared to my pure guitar signal. In your case it would translate into having the global input level set to -4dB when testing your capture in a preset. But I guess you ruled that out already, did you?
1 Like
To confirm, I’m not saying that the global input level alters the amount of gain baked into captures, I’m saying that the amount of gain that captures will produce when inserted into the grid will differ depending upon how hot your input signal is.
Think about it - the captures are designed to produce different levels of gain depending on the strength of the signal that they are seeing (just the like the devices that their based on & just like the inbuilt modelled devices as well). If this wasn’t the case then the gain level would remain identical no matter what guitar/pickups you used, there’d be no dynamics at all regardless of how you played, the volume knob on your guitar would be rendered pointless & the captures would be utterly useless quite frankly. This is why there’s a multitude of different sounds played at different levels during the capture process - it allows for the QC to analyse how the target device reacts to different signal levels at it’s input.
Try this:
Load up any amp block into the grid, leave the settings of said amp block as is, open the I/O window & adjust the input level up & down while playing. You will see that the gain/distortion level will change as you alter the input level. Another way to test this is to simply turn the volume knob on your guitar up & down while playing & see how that effects your gain/distortion levels (no need to even alter anything on the QC to see that the input level does most certainly make a whole load of difference).
Again, to be 100% crystal clear here, I’m not saying that ANY of the controls anywhere in the QC in any way alter the level of gain baked into the captures. What I am saying however is that for those of you who want to ensure that the gain level produced when using your captures with their gain control set at 0.0db is as close as possible to plugging straight into the target device with it’s settings kept the same as they were at the time of capture - the answer is to experiment with your global input level so that the signal hitting the input of the capture produces as close a match as possible to what what you get from the original device when you plug straight into it (the best way to do this is to use an A/B switcher so that you can switch back & forth quickly while comparing). Fair warning though, even if you do this you will likely still find that you need to make some adjustments somewhere as the capture process isn’t perfect & not all captures will perform equally. For example; with my global input level set as it currently is, I have one preset which has three captures that have identical gain levels as their original target devices when the gain controls in the capture blocks are set to 0.0db; however, I have one capture in that very same preset that needs a significant gain cut.
1 Like
Exactly! The global input level most certainly DOES affect how ALL of the virtual devices in the QC perform (captures included). I’ve personally experimented a great deal to get to the point where I’m happy with my input level & can confirm that the very same captures produce very different amounts of gain based upon these settings. It’s definitely worth putting in the effort if you’re consistently unhappy with the gain levels of your captures.
Thanks for the hint, but yes. Even did 2 factory resets, checked the global levels again, etc.
I wanted to add that your method would be useful only if you use an A/B pedal, but you noted that in your follow up post. Yes, then i understand what you mean - increase the test signal floor with the global level to where you feel it is closest to unity and then compare to reference via A/B pedal, as you would still have a difference when comparing within Cortex.
That for sure is another way to work around it.
Yep; you got it! That’s why I pay not intention to the ‘instrument level’ in the capture window & why I only toggle between the ‘cortex’ & ‘reference’ tones to check that the capture worked & there’s no obvious issue. I perform all tonal comparisons & required tweaks etc. by A/Bing to the original device when plugged straight into it (not by going through the QC first).
FYI 0.0db input on the QC most certainly doesn’t equate to the blocks in the grid reacting the same way as the real-world devices would (be it captures, amps or effects) - I’ve found that a significant boost at the input is required for all to work as it should.
Suffice-it-to-say, I don’t believe that the issue is that captures have too little gain, I believe that the issue is the discrepancy in how things work when the QC is in capture mode & that people aren’t all that clear on the importance of the global input level to all of the sounds in the QC. How captures sound in the capture window is irrelevant as this is not how/where one uses them anyway - it only matters how they sound once inserted in the grid & in use. This is why the global input level is key &; if you get it right, you’ll find that all of the blocks benefit from it (not just captures).
a little off-topic/tangential here, but you mentioned you are doing ‘thousands’ of captures. Are you saving that many? Cloud only? Just curious how you’re going about that aspect of it. I’ve never found any info on if there’s a limit to how many can be stored in the cloud
Don’t know if you’ve seen this already, but this guy has quite a complex setup in order to gain-stage the capturing process: Gain staging when creating Captures with the Quad Cortex - YouTube
He’s also talking about the QC missing 12dB on it’s outputs within the first 3 Minutes. Might be interesting to watch for the ones following this thread.
I think i’ve uploaded about 500 to the cloud, no issues so far. Thousands i’ve made counting all captures i’ve done, including tests, variants, etc. to get to good results. I’m not aware of any limitations regarding this.
1 Like
Yes - if you’re amp gain is on 10, you’re out of luck without an external boost. But it still shouldn’t be so far off that upping the capture gain a little won’t resolve it. Anyone you share it with is going to end up adjusting it anyway
1 Like
BTW, sharing captures as a preset with the capture embedded would preserve other than default settings for the capture.
1 Like