I was asked by a member of our online user group (the UILIST) about how I was able to avoid artifacts when resynthesizing the guitar harmonics to create the FH01 and FH02 examples. The text is from a posting to the list. If time permits, I will polish the presentation and provide more of a tutorial. Apologies if the writing here is neither elegant or concise -- my time is limited so this is a bit more stream-of-consciousness than usual (concision takes time). You may know all of this already. If not, let me know. I never know what is obvious and what isn't.
SIDE NOTE
This part may only be of interest to geeks like me. If not interested in the 'background' of this scroll down for the tips and techniques.
The long example sounds are not where I thought I would go when I started working on the examples. As often happens for me when working in MetaSynth, I make accidental discoveries that I become intrigued by which take me away from my original intention. My original goal was to go through my electronica library, find some great filter sweeps, and "steal" them by using various techniques in the Image Synth, Image Filter and Spectrum Synth and then apply them to samples destined for multisamplers or to notes in a groove. While exploring one particular filter sweep, I got curious about what would happen if I applied it over a much longer time domain than the original sample. The original filter sweep for the second example was maybe 1/2 to 1 second long. It worked nicely and seemed 'rich' to me. So I wanted to hear it stretched out to see how rich it was -- and I ended up really liking the result.
Something that interested me was that the result (in this case) was that applying the filter worked best if I applied it to the guitar harmonics before doing the time dilation (in the Spectrum Sequencer). But that could also be worked around by tweaking the filter (probably reducing some of the low level noise in the filter) and applying it after the time dilation.
ABOUT THE TECHNIQUES FOR FH01 and FH02
These notes apply only to FH01 and FH02
The filtered harmonics sounds started with the short sampled of a plucked harmonic (two strings ringing) you hear in the Sources section. That short sample (1.2) is the 'core' sound. The 'Filtered Harmonics' examples don't involve any voice samples in the process.
Quick synopsis of the technique: I applied an Image Filter (derived from the analysis of a synthesizer note -- from a sampled recording -- whose filter envelope I liked) to the short sample then analyzed and resynthesized in the Spectrum Sequencer where I changed the duration of the sound.
I describe the main process below and then later some tips that are useful at each stage. If there is enough interest (people will need to post to indicate that they are interested) I can post some pictorial examples that demonstrate some of the techniques.
Both of the examples used the same basic process:
==Stage 1: Create a filter by analyzing a sound
Find a recording where there are nice fat synth notes that have a filter envelope that one likes. For these examples, I used two different sources. Both are no longer and 1 second or so (one might be 1/2 second). The synth note needs to be reasonably well isolated from other instruments (preferably the synth is the only instrument playing
-- My filter sources: the opening synth note from Kalpol by Autechre, a synth pad from Kalpol (a few seconds into the song), a phrase a couple of note long from The Crystal Method's Busy Child. The Busy Child notes come at around beat 3 of the second measure of the songs main groove.
==Stage 2: Adjust and Apply the filter
In the Image Filter Room, filter the source.
-- Stage note: at this stage you may have to tweak the filter a little to fit the source material if there is not enough (or if there is too much) of an intersection between the source and the filter, the filter (or the source sound) will need to be tweaked. In the examples I didn't tweak the source other than to crop out the pick sound at the beginning of the guitar sample.
==Stage 3: Resynthesize in the Spectrum Sequencer
) In the Spectrum Sequencer, analyze the sound
) Shorten the last event in the sequence so that the rendered sound ends with silence (read the Spectrum Synth docs if you are not sure why this has to be done)
) Use the tools to adjust the balance of the partials.
--Stage note: for these examples I did very very little processing in the Spectrum Room because I did not need to do much. The filtered sound was clean enough that I only needed to make very minor adjustments using the Increase Contrast tool
TIPS AND IDEAS
The order of stages 2 and 3 can oftentimes be reversed. Sometimes it is interesting to resynthesize the source before filter. Or, it can be interesting to re-filter (with the same or different filter) after resynthesis.
My simplified tips for avoiding artifacts when using the Spectrum Room:
- ) Generally the best results for this sort of time dilation comes from sounds that don't have much noise and don't have rapid transients (which in essence are a kind of noise). (That being said you can get some very interesting sounds from noisy sources but they take a lot of work if you want something 'warm'). In this case, I cropped out the very beginning of the guitar harmonic because there was a quiet click (the pick or my fingernail touching the string) that made the results unsatisfactory. Applying the filter before going to the Spectrum room also reduces stray harmonics. For creating long pads in the spectrum room, you want a harmonically stable base. Sounds caused by vibrating strings often work nicely (but avoid consonants like 'b', 't', 'd' and s' when using these techniques because they are noisy.
- Cleaning up the sound in the Image Filter before spectrum analyzing can be very helpful. I tend to see if there are noisy spots or spots with too much high frequency action. When there is a lot of high frequency action, I find that spectrum sequencing gets more difficult to manage. Analyzing to the grid layer in the image filter often makes the troublesome areas really easy to see -- so I will notch those areas out. In these particular examples, I did not have to that because guitar harmonics are an almost ideal source.
- Investigate the low-amplitude harmonics by using the the Decrease Contrast tool. You will often discover that there is a lot of low level noise. When going for artificial sounds and wild percussive effects, it can be nice to emphasize this noise (and then filter in the filter room or treat with inertia or resonance later). But, when going for warm pads, you need to use the increase contrast tool to suppress them.
- I generally do not use the Tune tools when trying to preserve warmth. This tool and its variants are great for creating electronic sound events but they often remove the warmth from the sound.
Image Filter Tips
Here are some tips for creating useful filters from analyzed sounds.
- Do the initial analysis in the Image Filter since you don't have to do the whole duration adjustment process. You can just select the portion of the audio to analyze then click on the Analyze Spectrum button.
- Create a wide picture (1024 or 2048 pixels wide) and use semitones or quartertones. (Adjust the height of the picture as necessary).
- For the next stage of refinement, it can be useful to work in the Image Synth. There are a few reasons. One is that layers can be useful while developing the filter. You can copy and paste between layers as you work on variants which is handy. (But you will have to merge layers or split the layers into separate presets before using them). Another reason to work in the Image Synth is that by synthesizing the filter in the Image Synth, you can hear all of the frequencies that the filter contains. It takes only a little bit of experience using the same image as synth and filter to get a sense of of they correspond.
- While refining the filter, I will always work from a copy of my original analysis because some of the experiments fail and I need to go back to scratch.
- Use the Contrast and Luminance tool, to explore the low-level content that may not be visible on your screen (most people's gamma adjustment is wrong and you won't see the low luminance pixels). In some cases, you actually want to emphasize the low-level content and in others you want to suppress it. You need to just experiment and see what gives the results that please you the most. I often end up with several different presets from my original analysis, some where I have brought the low level pixels up and others where I have suppressed them.
- If there are big gaps between the pixels, it is often useful to use option-nudge up/down or left/right to thicken the filter. (You can get fancy by select all, transfer mode: maximum, copy, paste, arrow-key up or down, THEN use the color palette to reduce the amplitude or change the panning of the pasted copy)
- Motion Blur with a smallish Effects Grid setting is often useful
- Back in the filter room, open up a sound to filter and preview the filter and explore the following:
- Adjusting the filter reference pitch up and down to hear what it sounds like
- Explore what happens when EITHER the bottom line(s) of the canvas is black and when it is white/yellow (keeping in mind that all frequencies below this line will be filtered with this line's setting). I often find that I want all those frequencies to pass through the filter and in other cases I want to filter them out.
- Explore what happens when EITHER the top line(s) of the canvas is black and when it is white/yellow (keeping in mind that all frequencies above this line will be filtered with this line's setting)
- I also like to explore making slightly different modifications in the red and green channels which creates a lot of depth in some cases.
- In conjunction with the previous tip, make Micro 8 or Micro 16 version of the filter (which will require you to increase the height of the canvas in order to cover the same pitch range). The quickest way to do this is: copy, adjust the height of the canvas, paste, set the new tuning space.