Freesound Forums

BTW, AlienXXX, I can't tell what the disagreement is. Perhaps there is none, really. Or perhaps I am missing your point. When treating image as sonogram and doing inverse dft to go back to sound, using the rectangular approach means you can be changing the magnitude of your lowest frequencies as fast as you change that of the highest frequencies, but it will be nearly meaningless to do so if those low frequency magnitudes are changing more rapidly than the waveform of those frequencies themselves -- you basically end up with a modulation of the pixel-progression rate with those lower frequencies, which is itself a detectable artifact, and likely not a pleasant one. By using the radial approach, the low frequencies will change at a lower rate, the rate per frequency band being linearly proportional across the range. And that aspect might keep the linear approach cleaner, allowing more pixels of change for lower-mid frequencies than would be possible with the "exponential" mapping.

Another way the conversion could be done would be to somehow treat the entire image as just one huge one-dimensional spectrum -- the the inverse fft on the entire thing all at once, to get a waveform in the time domain that is essentially twice as long in samples as there were pixels to start with. Or treat R and G as real and imaginary portions of the complex spectrum and treat B as the left/right balance, and end up with a waveform 4 times as long but in stereo. As usual, you can zero-pad to a power-of-two, you can upsample or downsample afterwards, etc. So if the image were in column-major order, then the similarities between adjacent columns would map to repetitions in a harmonic structure whose base frequency would be affixed to the height of the image. That doesn't sound very promising, does it?

Combining the two ideas, one could use a spiral helix sort of thing from the center outward. You could already see the radial method as being such a thing, if you just say that each cycle around gives magnitude-over-time for an ever-increasing frequency, for doing inverse DFT. But that's linear growth in resolution with frequency increase. We could instead not use full cycles per frequency band, but simply the number of pixels needed for the frequency in question given an exponential approach, allowing the allotment to any given frequency to span more than or less than one (or more) full circles. The offset of the resulting radar-position (angle) at each frequency band could create less spectral smear and offer more isolation to the spectral contents, while still preserving some repetition (due to similarity of adjacent pixels) over greater and greater intervals as the frequencies increase. This would be very difficult to explain to anyone, which of course is kind of what I'm attempting to do here, just hoping you follow the idea.

On second read, not sure I would call it a disagreement. I should have called it perhapd "points to consider"

zimbot wrote:
using the rectangular approach means you can be changing the magnitude of your lowest frequencies as fast as you change that of the highest frequencies, but it will be nearly meaningless to do so if those low frequency magnitudes are changing more rapidly than the waveform of those frequencies themselves

When converting image to sound, the user has control to specify the suration of the sound - which in turn will determine the time duration of each pixel (or, more correctly, each vertical line).
You are right, this could result in modulation/artifacts. You do not even need the duration of the pixel to be less than the duration of a wavecycle - what you need is a rapid changing pixel pattern (e.g. white/black/white/black) and a pixel duration that paces that in the audio range.
Maybe programs like Audiopaint have a smoothing algorhythm to prevent such artifacts - This can be easily determined by experimentation - Samples comming up

zimbot wrote:
By using the radial approach, the low frequencies will change at a lower rate,

In this regard we have no disagreement. But I think I failed ot make my point, which is:
If the process is based on 'reading' a normal X-Y picture (i.e., a normal picture file on a computer) in a radial way, then there will be a very low number of points near the origin.
At the origin there is only 1 point, which would mean the lowest frequency having a constant value for the whole duration of the sound.
When draing very small circles in a computer screen, these are highly pixelated and do not look like circles at all. It is only when the circle is large enough vs the size of the pixel that this effect becomes negligeble.
SAme happens here close to the origin. Although algorythms can be used to interpolate as many points as you want for the purpose of sound generation - the reality is that the data is not there, so all you will do is smooth out any transitions between the small umber of actual data points.
So a more effective sound generation method would map the lowest frequency to a circle at say 10 points away from the origin and ignore any points closer to the origin.
Then you could use linear or exponential frequency mapping.

As for the othe rmethods you sugges, I am not sure I understand them completely.
Would be interesting to discuss furhter but we should do so on IM, as I am not sure this discussion is interesting for most other Dare-26 participants

AlienXXX wrote: As for the othe rmethods you sugges, I am not sure I understand them completely.
Would be interesting to discuss furhter but we should do so on IM, as I am not sure this discussion is interesting for most other Dare-26 participants

I have to admit that 99.9% of what you're both talking about goes completely over my head, but as a participant I have no problem with this discussion continuing in public, and it's definitely on-topic and could be very useful for programmers. I'm all for new methods of sonification and I wonder what Victor Khashchanskiy and others would make of it all. So carry on if you wish.

Used AnalogX Sayit to make beeps, noise and drones.

This http://www.freesound.org/people/toiletrolltube/sounds/235359 becomes this http://www.freesound.org/people/toiletrolltube/sounds/235360

toiletrolltube wrote:
Used AnalogX Sayit to make beeps, noise and drones.

This http://www.freesound.org/people/toiletrolltube/sounds/235359 becomes this http://www.freesound.org/people/toiletrolltube/sounds/235360

Give me a second to get my jaw back from the floor.... okay...
It's AWESOME!!!!

Glad you liked it Copyc4t, I think I just got lucky though. As in most cases I start out with "I wonder what would happen if..." and when I heard the beeps, processing it into a gong like sound seemed the obvious way to go. Probably influenced by that Beautiful Destruction music I heard recently, but there ya go. Still, another unusual and fun way to create sounds with text to speech (I don't know what that 'o' face means, but it's funny )

We have moved into a completely new universe here in terms of sound creation techniques.

I am still stuck in the old universe, as i have a number of experiments still to carry out (and upload)
But you guys go ahead without me, I'll catch up soon.

For some reason, toiletrolltube's latest reminded me that not too long ago I created an AnalogBox circuit that was kind of like a giant, complex, chorusing reverb, and then I tried giving it ridiculous feedback with some limiting to ensure things didn't get out of control, and then tied it to a cheap mic input on my computer. I could vary all sorts of parameters and even throw in some occasional vocalizations to help feed the circuit while it swirled away. I had never heard anything quite like it, and it seemed kind of freakish and scary, so I saved some to a file and posted it on freesound. If you're interested, it's here:

http://www.freesound.org/people/zimbot/sounds/204356

toiletrolltube wrote:
Used AnalogX Sayit to make beeps, noise and drones.

This http://www.freesound.org/people/toiletrolltube/sounds/235359 becomes this http://www.freesound.org/people/toiletrolltube/sounds/235360

Very, very fine work!

I recreated your effect using chain GRM Comb -> GRM Evolution -> SparkVerb. Interesting method.

Trebblofang wrote:

toiletrolltube wrote:
Used AnalogX Sayit to make beeps, noise and drones.

This http://www.freesound.org/people/toiletrolltube/sounds/235359 becomes this http://www.freesound.org/people/toiletrolltube/sounds/235360

Very, very fine work!

I recreated your effect using chain GRM Comb -> GRM Evolution -> SparkVerb. Interesting method.

I would like to hear that and I'll have a look into Evolution.

I'm actually processing, or have been for the last two days, your http://www.freesound.org/people/Trebblofang/sounds/234165 sample by stretching and time compressing repeatedly until all the boundaries blur and merge. Sounds good with lots of drone potential, but getting the final sound right is my only obstacle at present, but I'll get there.

Zimbot, probably the mention of "analog" that reminded you. I've downloaded and will have a go at exploring that sound, definitely on my 'next' list. Also interesting to learn about more unusual analog methods of sound creation.

toiletrolltube wrote:

I would like to hear that and I'll have a look into Evolution.

I uploaded my sound here: http://www.freesound.org/people/Trebblofang/sounds/235528/
It waits for moderation, but keep checking & stay tuned for next expriments

Hi all, I'm only just catching up, last night I got blown away by the amazing samples created. Some of the conversation is way over my head too but makes for interesting reading and is definitely one you can keep coming back to, maybe when I've progressed a bit! So many bits of software and techniques to try out. The in depth descriptions I like because it's much more fun when you know why something works the way it does. So thank you for sharing.

AlienXxx I loaded the drumkit samples you created into flstudio, really nice and easy. Ive only got the demo which means you can save but not open again. Doh! Drums have never been a strong point but i will upload a few as an example of the drumkit in action. Copycat I'm working out a way to work with flac, maybe have to convert. Trebblofang and toiletrolltube your samples are awesome!!! I'm feeling a little bit inspired you guys, thanks again. I feel lucky to be involved in this discussion

@luckylittleraven:

Sorry for the inconvenience with FLAC, I tend to prefer it to WAV, but Audacity will easily convert my sounds to WAV for you

If you want to have drum fun that you can save and reopen, without switching to a different DAW, you might try this standalone drum machine:
http://www.hydrogen-music.org/hcms/node/21

Dear friends and daring Freesounders

I have been a little absent from these discussions due to workload and personal life requirements.
By the time I am done with the two above it is past 11pm and I have no energy left.

I am a bit behind on listening to the sounds you have been uploading, but I have kept up with the thread postings (thanks to my iphone).
Weeked is coming and there wil be time to pick up some of the stuff I have pending at the moment: more sounds, listenning and comenting to what has been going here, working on a pice where I include some souds derived from data-bending and image-to-sound...
... and...
...a new dare!

@copycat on no, it's cool! Isn't is a windows/mac thing? or is that ogg? Anyway thankyou for the tip about audacity I didn't think of that, another nice piece of free software that I've recently discovered. Yes I do like my trusty old paid for honestly DAW, so at the moment I wouldn't change and invest in something else, but there is so much awesome freeware out there as I have recently discovered. I do really like the idea of 'creating' a drumkit with whatever sounds you want . . . Hydrogen sounds good, I will check that out sometime.

@luckylittleraven:
Well, some formats are native to some platforms, even with narcissistic references in the format names like Windows Media Audio (wma) and Apple Lossless Audio Codec (alac) but most of them are "more or less" usable anywhere.
That "more or less" is because of patents, we don't have to pay to use the mp3 format for instance, but if their inventors decided to put a fee for the registered patents, they could make us pay. Maybe it will never happen, but that's why personally I prefer completely free formats like flac and ogg
(if you are curious about the various formats, this page is interesting: https://en.wikipedia.org/wiki/Audio_file_format )

Audacity is indeed a great tool! Even though I make music with a DAW, Audacity is always part of my workflow, to prepare single tracks (e.g. noise removal on recordings) and sometimes convert them (some collaboration tracks I worked with were wma, aac, aiff), and to convert the final products from flac to ogg or mp3.

Hydrogen is what I used for the drums in Speed Limit; drum VSTi's have few slots, so they're good when you have already chosen which sounds to load into them; but Hydrogen can load many more sounds, helping you find the ones that sound best together, and it also supports effect plugins, so it makes experimenting quite easy.

@AlienXXX:
I'm really curious to hear your final product from this Dare! But take your time, this one has no deadlines

After a tension-building wait, here comes the last pack from Trebblofang's goldmine, the TrebbloNoises pack:
http://www.freesound.org/people/copyc4t/packs/14483/

Two of the sounds planned for this pack have been assigned to the TrebbloGUI pack:
http://www.freesound.org/people/copyc4t/sounds/235652/
http://www.freesound.org/people/copyc4t/sounds/235735/

You know what, i'm thinking of converting image to sound in a different way.

All around me I see objects/architecture with the potential to be converted into a sequence of values. If I build up these values over time, I can plot them on a 360 degree oscillation axis and play the values represented as an SPL. I would have to come up with some algorithm to sort the values in a sequence that would render the waveform consistent in amplitude; as to not get a distorted waveform through sudden and drastic changes in the value.

I would love to try this.

Is this completely crazy?

Headphaze wrote:
You know what, i'm thinking of converting image to sound in a different way.

All around me I see objects/architecture with the potential to be converted into a sequence of values. If I build up these values over time, I can plot them on a 360 degree oscillation axis and play the values represented as an SPL. I would have to come up with some algorithm to sort the values in a sequence that would render the waveform consistent in amplitude; as to not get a distorted waveform through sudden and drastic changes in the value.

I would love to try this.

Is this completely crazy?

Do you mean like converting a skiline into a waveform and then playing it (looping)?
Please note that static waveforms loop-played always produce a static tone - at best you can expect an organ-like sound.
You need to get that waveform evolving over time for an intersting sound (use filters with envelopes, or LFOs. Use chorus / flanger / phaser ...)

Programs like Mathematica and Mathcad have the functionality to do that - convert any list of values to a wav.

luckylittleraven wrote:
I do really like the idea of 'creating' a drumkit with whatever sounds you want . . .

Remind us, what DAW you use?
Your DAW probably has some built in drum machine that can load samples. Or a sample that can load different sounds simultaneously and map them to different keys.

There are many simple VST drum machines that can load samples:
You can find some here:
http://www.vst4free.com/

For example this one
http://www.vst4free.com/free_vst.php?plugin=Midinaut_Drum_Station&id;=1676
This one I think also loads samples
http://www.vst4free.com/free_vst.php?plugin=DPC_3&id;=385