I am wondering if the factory preset AMPs are ‘simply’ factory captures of said amplifiers, or were they created via some other process?
The stock amp models are neither plain captures nor are they modeled bottom-up. Some people came up with the term “deep captures” due to the fact that the real amps were apparently captured in multiple (or all) possible knob configurations for authentic knob parameter behavior.
Oh, that’s very cool! And thank you for the reply. I wonder if the QC enhancements might ever include the features needed to allow users to perform similar ‘deep captures’ on their own… that would be a remarkable addition to an ready incredible product. I recently sold a mini van load of all my amps and purchased my QC. I am a home-studio enthusiast, and the decision was one I have never regretted. The QC delivers more than (literally) the van full of equipment did. Thanks Neural!!
As far as I understand, it seems to be a very time consuming process since you need to run the amp through all possible knob configurations. Doug from NDSP mentioned, they’re using robots to automate this task.
Wow, that means, for example a 5 knob amp (gain, vol, low, mid, high) with about 10 positions per control, could require up to 10 ^5, or 100,000 captures! Something that robitics would certainly help accomplish. That approach is really insightful in helping understand the remarkable capabilities of the QC! Once again. Thank you for your reply!
The different knob positions aren’t handled by different captures, they are additional input into the capture process. That is, they are included in the input data set along with the audio tones and are used in the training set used by machine learning to produce the “deep capture”. The robots are used so that the knob position changes can be automated and known for data input.
I think, if we use the term “capture” loosely, @flashtheproducer is not wrong since for each set of knob positions you still need to record some kind of response in order to have training data consisting of input and output. Maybe not a full capture is needed for each set of knob positions but I’d say, it’s still some kind of capture per setting and the amount of these captures goes up non-linearly with the amount of adjustable amp parameters.
Yes, that is correct, and what I intended. On s similar note, I think it would be a powerful feature if captures could consist of one or more “capture data”. In other words, going through the process, multiple times, with a variety of predetermined settings, for example a capture with gain at 12, 3, 6, 9 o’clock, and then having the QC combine those “captures” when it crunches the data. Seems like with a little UI designing and thought, that this “layered” captrure idea might have potential in creating even richer simulation models… just a thought.