'Aim' by fine-tuning the sound: Adjust, first the NOTE and then DETUNE controls so that the original and resynthesized sounds are closely matched (if that is your goal). For example the human voice, plosives, breath and similar untuned percussive sounds tend to acquire a 'vocoded' quality.
A good technique is to fade the BALANCE knob fully left/right after making adjustments, if you have the perfect resynthesis the sound won't change (much) as the BALANCE knob is moved. The resynthesized sound is only a preview, always 'Import charged sample' (described below) to hear the final resynthesized sound. To help resolve these issues the Noise Generator can be used to restore untuned noises to the patch.
The note range C4-C5 also tends to work best for single sounds stretched across the keyboard.
If you intend to have multiple Keyboard Zones use a sample that is in the middle of the Keyboard Zone range that it will serve (the same principle that is applied to multi-sample patches on samplers).
TEST - Starts/Stops a looped play-back of the original and resynthesized 'preview' sound.
NOTE: The resynthesized sound heard here is a preview only (in order to reduce CPU load for testing).This is similar to the 'Root Note' function in a Sampler and is the main control to adjust when matching the resynthesized sound with the original sample.RESOLUTION - Changes the rate at which Breakpoints are laid down (similar to sample rate).If desired you can change the import resolution by selecting the RESOLUTION (Normal, High, Best) options in combination with the 'Import charged sample' step.If you have resynthesized a sample and the sound is still out of tune (and you have set the Resynthesis DETUNE range to the max) the following method may help: 6.1 If you don't need to preserve Detune settings on other Spectra, turn down the 'Control Range' Detune slider (on the main interface) to the minimum setting.The optimization criterion in VAEs being the sum of the reconstruction error and a regularization term, it naturally leads to a lower reconstruction accuracy than DAEs but we show that VAEs are still able to outperform PCA while providing a low-dimensional latent space with nice "usability" properties.We also provide corresponding objective measures of perceptual audio quality (PEMO-Q scores), which generally correlate well with the reconstruction error.A good analogy to keep in mind when Resynthesizing is: 'Load, Aim and Fire'.That is, 'Load' the sample, 'Aim' by adjusting the Resynthesis controls and 'Fire' by sending the sample to the desired Keyboard Zone.The Resynthesis function takes an input sample and recreates the sound as a 'Spectrum' that is a series of Breakpoints.Each Breakpoint represents a harmonic 'snapshot' of the sound at a given moment in time.