Woah, thank you for using my little web app in this video! I've fixed a few bugs since this video came out, but if anyone has any issues feel free to hit me up!
I believe the way it works is each pixel's brightness represents the amplitude of a sample of audio, and it reads left to right like a book, so when he's copying and pasting vertical layers, theoretically it would be like repeating a line of text, and that's why it chops the sample rather than blurs it. I could be wrong though, but it makes the most sense to me. Edit: got further into the video, that also explains the distortion effect, because with more contrast, bright pixels get brighter and dark pixels get darker, which should also stretch the waveform in a similar way. It's kinda hard to explain, but I can sorta visualize how it's working.
@@stxnw It's basically how hard you play a note, it's in practically every DAW and it's kinda like volume but not exactly. I'm surprised someone can watch this guy without knowing what it is, but maybe you just started watching him.
how this website actually works is related to the way the golden record from voyager 1 and 2 works. how it’s related is because both ways the record’s encoded image audio and the way the image is converted into sound uses the same encoding technique. i used to have a hyper-fixation over this, which is how i know how it worked. let’s say you input an image with a height of 432 pixels. the waveform that it outputs is actually divided into 432 parts, with each part corresponding to one pixel row of the image. how the converter encodes each part of the waveform (which corresponds to one pixel row of the image as i said earlier) is by using the crests (aka peaks or high parts) and troughs (aka valleys or low parts) of the waveform as different brightnesses. crests corresponding to lighter colors of each pixel row, and troughs corresponding to darker colors of each pixel row. so the converter scans through each row of the image left-to-right and outputs them as one part of the waveform. the converter scans through the image and outputs them as a waveform until it completely finishes generating. this is how the images are converted into waveforms.
PNG and WAV files are probably the best formats and they are my favorite for image and sound. WAV format is just uncompressed bytes of sound. PNG uses 4 bytes of data per pixel (RGBA), and usually WAV uses 4 bytes characters as far as I know, so it's perfect conversion.
It would also be cool to try to paint a full understandable picture with recognizable objects in the picture that make a song when converted to a wav file.
I think that the best explanation I have seen is that it reads left to right top to bottom, red channel is probably the left channel and green is the right channel. That would make the most sense to me. That would explain the popping sounds, the color of the image, and also means it would be hard to use image editing software to actually edit it due to the way it is formatted Edit: This is almost definitely how it works. If it was formatted differently that would make editing it much easier
how this website actually works is related to the way the golden record from voyager 1 and 2 works. how it’s related is because both ways the record’s encoded image audio and the way the image is converted into sound uses the same encoding technique. i used to have a hyper-fixation over this, which is how i know how it worked. let’s say you input an image with a height of 432 pixels. the waveform that it outputs is actually divided into 432 parts, with each part corresponding to one pixel row of the image. how the converter encodes each part of the waveform (which corresponds to one pixel row of the image as i said earlier) is by using the crests (aka peaks or high parts) and troughs (aka valleys or low parts) of the waveform as different brightnesses. crests corresponding to lighter colors of each pixel row, and troughs corresponding to darker colors of each pixel row. so the converter scans through each row of the image left-to-right and outputs them as one part of the waveform. the converter scans through the image and outputs them as a waveform until it completely finishes generating. this is how the images are converted into waveforms.
i think the weird stuttering you're experiencing, which you speculate at 3:55 is the sample rate, is due to the actual png resolution. i suspect that each one of those delay/echo effects is occurring with every pixel in the image, so if there were some way to increase the vertical resolution of the images that the converter program uses, you could have less choppy results.
@@arcioko2142 if the .png resolution was the sample rate, either the images would be much much taller or we wouldn't be able to see all the little oscillations in the resulting waveform, like at 6:59 we can clearly see the waveform's oscillations occur more quickly than the stutter effects, and based on how many of those oscillations fit across the screen at once, we can easily see how if there was even one pixel for each oscillation, the .png would be so much taller than it is
It's because the audio is encoded into one single horizontal line of pixels that is then wrapped vertically. It should be read right to left, and then when you reach the end of the line on the right it continues on the next row on the far left. Think of it like reading this comment. If you did a vertical motion blur all you are doing is duplicating letters vertically across words in different lines. Example: This is a sentence about ducks. Quack! Gets incoded as: This is a sentence about ducks. Quack! Turns into: Tahbiosu t idsu cak sse.n tQeunaccek ! Tahbiosu t idsu cak sse.n tQeunaccek ! Which would turn back into audio as: Tahbiosu t idsu cak sse.n tQeunaccek ! Tahbiosu t idsu cak sse.n tQeunaccek ! Rather than: TTThhhiiiss iiisss aaa ssseeennnttteeennnccceee aaabbbooouuuttt ddduuuccckkksss... QQQuuuaaaccckkk!!! This is why everything got stuttery. Because he was blurring the sounds vertically across multiple rows of time. Even when blurring horizontally, the blur doesn't wrap with the pixels so the audio at the left and right edges gets messed up. If you could unwrap this image into a single horizontal row of pixels the blur would probably sound a bit more like reverb.
@@ORyanMcEntire (with the exception of the motion blur on your ducks example) that actually makes a lot of sense. after experimenting with it a bit myself, I see that your explanation makes much more sense than what I said. interesting that this means there's no connection between sound frequencies and the x coordinate in the image (in the sense that the lower frequencies are not to the left of the higher frequencies or vice versa)
Now do the reverse. In audacity, if you click "import raw audio" it will accept ANY file type as audio. Ive found using weird file types with weird data (.AVI, .blend, .apk, etc.) Gives the best results.
I'm still watching, but the first thing I would do is convert the sound into a png and then BACK to audio without changes to make sure it even does that properly in the first place. Shakiness of audio might be just a png compression artefact or something like that
have you found out about blob opera yet? I spent a couple days going through a bunch of covers people made with the 'experiment'. its an a.i powered choir.
I think the way you are assuming the audio got encoded as the image might be the reason the experiments didn't sound great. I'm pretty sure the audio is encoded into one single horizontal line of pixels that is then wrapped vertically. It should be read right to left, and then when you reach the end of the line on the right it continues on the next row on the far left. Think of it like reading this comment. If you did a vertical motion blur all you are doing is duplicating letters vertically across words in different lines. Example: This is a sentence about ducks. Quack! Gets incoded as: This is a sentence about ducks. Quack! Turns into: Tahbiosu t idsu cak sse.n tQeunaccek ! Tahbiosu t idsu cak sse.n tQeunaccek ! Which would turn back into audio as: Tahbiosu t idsu cak sse.n tQeunaccek ! Tahbiosu t idsu cak sse.n tQeunaccek ! Rather than: TTThhhiiiss iiisss aaa ssseeennnttteeennnccceee aaabbbooouuuttt ddduuuccckkksss... QQQuuuaaaccckkk!!! This is why everything got stuttery. Because you are blurring the sounds vertically across multiple rows selection of time stacked vertically. Even when blurring horizontally, the blur doesn't wrap with the pixels so the audio at the left and right edges gets messed up. If you could unwrap this image into a single horizontal row of pixels I bet the blur might sound a bit more like reverb.
Now you gotta make a song where you hide some memes or something in it. The sounds come out very strange though when you use a picture and I think it may have to be black and white. I played around with doing it a while ago but couldn’t find a good way to hide the sounds without it seeming obviously out of place
If you want multiple layers to show up the same amount, you can’t set the opacity for 25 for all of them. You need to make the bottom 100, the second 50, the third 25, and so on. Otherwise the top layer shows well but the lower ones are all fighting to be seen
Awesome video! You should have compared the original loop to one that was converted to an image and then directly converted back into sound so you can see how much information is lost in the conversion process every time
i'v done this with Adobe audition and paint shop pro, but i used bmp, with raw import in psp. i assume this app work in pure bitmap, yes? .. as far as i remember png often deals with compression. however in my end its more rainbowish but the reverb effect was similar. i also think i used 16 bit audio which will result in a odd 1 + an half PCM per color, but if i used 24 bit audio , each PCM would be each color pixel. ** ok going down the rabbit hole, and figured out i had more success with 8-bit audio to 8-bit grayscale image, other only distorted audio since i guess i didn't get a proper 1:1 convertion
Woah, thank you for using my little web app in this video! I've fixed a few bugs since this video came out, but if anyone has any issues feel free to hit me up!
Yes, where can I find this app?
@@jordoneaton7083 Description man directmusic.me/wav2png/
@@jaczob666 Thank you. My screen narrator has been glitchy lately and appears to have missed that.
Do you have this up on GitHub or anywhere?
You really have to make this a VST
Ah, music to my eyes 😅
Cursed
Now we can say "I can see music notes"
🤣
@@Euphoric_Brunette perfect pitch In the eyes
OW IT SOUNDS SO BRIGHT
I believe the way it works is each pixel's brightness represents the amplitude of a sample of audio, and it reads left to right like a book, so when he's copying and pasting vertical layers, theoretically it would be like repeating a line of text, and that's why it chops the sample rather than blurs it. I could be wrong though, but it makes the most sense to me.
Edit: got further into the video, that also explains the distortion effect, because with more contrast, bright pixels get brighter and dark pixels get darker, which should also stretch the waveform in a similar way. It's kinda hard to explain, but I can sorta visualize how it's working.
I can explain it more simply: The X axis is frequency, the Y axis is time and the brightness of each pixel is velocity.
@@farmerchuck7294 wtf is velocity
@@stxnw It's basically how hard you play a note, it's in practically every DAW and it's kinda like volume but not exactly. I'm surprised someone can watch this guy without knowing what it is, but maybe you just started watching him.
@@farmerchuck7294 so its amplitude?
@@stxnw Pretty much
this reminds me of throwing pngs into serum wavetable
Or harmor
poo
Hello Dylan
how this website actually works is related to the way the golden record from voyager 1 and 2 works.
how it’s related is because both ways the record’s encoded image audio and the way the image is converted into sound uses the same encoding technique. i used to have a hyper-fixation over this, which is how i know how it worked. let’s say you input an image with a height of 432 pixels. the waveform that it outputs is actually divided into 432 parts, with each part corresponding to one pixel row of the image. how the converter encodes each part of the waveform (which corresponds to one pixel row of the image as i said earlier) is by using the crests (aka peaks or high parts) and troughs (aka valleys or low parts) of the waveform as different brightnesses. crests corresponding to lighter colors of each pixel row, and troughs corresponding to darker colors of each pixel row. so the converter scans through each row of the image left-to-right and outputs them as one part of the waveform. the converter scans through the image and outputs them as a waveform until it completely finishes generating. this is how the images are converted into waveforms.
I put random images into Audacity
Can we just all appreciate the quality of your videos
PNG and WAV files are probably the best formats and they are my favorite for image and sound. WAV format is just uncompressed bytes of sound. PNG uses 4 bytes of data per pixel (RGBA), and usually WAV uses 4 bytes characters as far as I know, so it's perfect conversion.
can we get a round of applause for the editing in this video!
Yo you keep disappearing and appealing out of nowhere with great content 🔥🔥🔥
He is very appealing indeed!
Task failed successfully
he really makes grate con tent
Appearing
yo i keep doin your mom
Of course here a challenge... Make a song out of pngs if that's even possible.
i already done it
@@sergejsdarznieks321 pics or it didn't happen
It would also be cool to try to paint a full understandable picture with recognizable objects in the picture that make a song when converted to a wav file.
I think that the best explanation I have seen is that it reads left to right top to bottom, red channel is probably the left channel and green is the right channel. That would make the most sense to me. That would explain the popping sounds, the color of the image, and also means it would be hard to use image editing software to actually edit it due to the way it is formatted
Edit: This is almost definitely how it works. If it was formatted differently that would make editing it much easier
There is a thing called databending where you open an image in a audio software and add effects it looks super trippy!
1:50 why does that sound so damn good wow
Your content is so different from other music producers. I love that!
You should've tried changing the hue of the image
You just gave me the perfect tool to make weird sounds for video games, and dubstep. I can not thank you enough for this video.
14:08 - That reminds me of scanning through radio stations sound.
how this website actually works is related to the way the golden record from voyager 1 and 2 works.
how it’s related is because both ways the record’s encoded image audio and the way the image is converted into sound uses the same encoding technique. i used to have a hyper-fixation over this, which is how i know how it worked. let’s say you input an image with a height of 432 pixels. the waveform that it outputs is actually divided into 432 parts, with each part corresponding to one pixel row of the image. how the converter encodes each part of the waveform (which corresponds to one pixel row of the image as i said earlier) is by using the crests (aka peaks or high parts) and troughs (aka valleys or low parts) of the waveform as different brightnesses. crests corresponding to lighter colors of each pixel row, and troughs corresponding to darker colors of each pixel row. so the converter scans through each row of the image left-to-right and outputs them as one part of the waveform. the converter scans through the image and outputs them as a waveform until it completely finishes generating. this is how the images are converted into waveforms.
These faster uploads are amazing.
I would actually love to see a part 2 to this video! That was awesome! ^_^
You should try converting audio to PNG to compressed jpg and back. Might be interesting.
the noise added to the song when you first tried it, it sounded cool like it had a lofi-ish vibe
i think the weird stuttering you're experiencing, which you speculate at 3:55 is the sample rate, is due to the actual png resolution. i suspect that each one of those delay/echo effects is occurring with every pixel in the image, so if there were some way to increase the vertical resolution of the images that the converter program uses, you could have less choppy results.
what if the png resolution is the same as the sample rate
@@arcioko2142 if the .png resolution was the sample rate, either the images would be much much taller or we wouldn't be able to see all the little oscillations in the resulting waveform, like at 6:59 we can clearly see the waveform's oscillations occur more quickly than the stutter effects, and based on how many of those oscillations fit across the screen at once, we can easily see how if there was even one pixel for each oscillation, the .png would be so much taller than it is
@@wyntrr_end oh ok
It's because the audio is encoded into one single horizontal line of pixels that is then wrapped vertically. It should be read right to left, and then when you reach the end of the line on the right it continues on the next row on the far left. Think of it like reading this comment. If you did a vertical motion blur all you are doing is duplicating letters vertically across words in different lines.
Example:
This is a sentence about ducks. Quack!
Gets incoded as:
This is a sentence
about ducks. Quack!
Turns into:
Tahbiosu t idsu cak sse.n tQeunaccek !
Tahbiosu t idsu cak sse.n tQeunaccek !
Which would turn back into audio as:
Tahbiosu t idsu cak sse.n tQeunaccek ! Tahbiosu t idsu cak sse.n tQeunaccek !
Rather than:
TTThhhiiiss iiisss aaa ssseeennnttteeennnccceee aaabbbooouuuttt ddduuuccckkksss... QQQuuuaaaccckkk!!!
This is why everything got stuttery. Because he was blurring the sounds vertically across multiple rows of time. Even when blurring horizontally, the blur doesn't wrap with the pixels so the audio at the left and right edges gets messed up.
If you could unwrap this image into a single horizontal row of pixels the blur would probably sound a bit more like reverb.
@@ORyanMcEntire (with the exception of the motion blur on your ducks example) that actually makes a lot of sense. after experimenting with it a bit myself, I see that your explanation makes much more sense than what I said.
interesting that this means there's no connection between sound frequencies and the x coordinate in the image (in the sense that the lower frequencies are not to the left of the higher frequencies or vice versa)
There's a function in Serum where you can use PNG images as wavetables
In harnor in fl Studio you can do that too
Its in the paid version of vital too I think
@@noface718 you can try it in the free version i think
@@raoufbensalem3417 nope Tested it You cant
@@noface718 i think its text to speech not this
0:01 omg the fucking nostalgia from the gta san andreas destination marker sound
Omg I’ve ALWAYS wondered how this would work
1:36 sounds so lofi
3:39 *TRIPLE DRUMS*
7:01 poppy guitar and drums
10:59 DISTORTED BOI
Now do the reverse.
In audacity, if you click "import raw audio" it will accept ANY file type as audio. Ive found using weird file types with weird data (.AVI, .blend, .apk, etc.) Gives the best results.
Good to know!
I can’t believe I’m watching this in the middle of class
Levi with another banger as usual
WAIT THIS IS EVERYBTING IVE BEEN LOOKING FOR
10:06 "That's a bulgy boi"
Beat proceeds to shart on everything.
Hmmmmmm
The end result had a faint Daft Punk kind of vibe to it. Sounds interresting, could be a great thing for sound design.
Puft dank
4:20 (not intended) that would make a good sound for like a machine gun.
I am loving the upload consistency recently! Keep up the amazing work!
guy literally be making beats in Microsoft paint. i thought excel was something, but now the music community has peaked. this is the ultimate DAW
Definitely a lot of fun and quite useful as well. Thanks a lot! Great video.
I can easily imagine Andrew Huang making music with that.
do this but try changing the orange color completely to blue or green or something. that could be really interesting
It does nothing probably
I'm still watching, but the first thing I would do is convert the sound into a png and then BACK to audio without changes to make sure it even does that properly in the first place. Shakiness of audio might be just a png compression artefact or something like that
Pretty true
I did it
It's not lol
@@Kai_On_Paws_4298 You mean it does not convert back to audio properly? :D Thought so
I love how the song at end actually sounds pretty good lol, btw can you try making music in a video editor like premiere?
You are a madman. I love it.
Another upload so quick??? This is great! Love the videos!
Aw yea i used to do this all the time as a kid. I loved using it for data moshing
Im surprised you didnt try using random images or doodling on the image
or pure sounds, like sine/saw/square waves
@@dacolib I used a sine wave-
You gotta drop the full release of the first finished product at the end that was actually sounding good 🙌
12:26 laser gun yes
Jeez man you just keep coming back with bangers
dang, levi is on a roll
My boy direct!
now i have some sounds to use in my bandcamp experimental album!
This video was made with *red heart emoji*
it really was made with [ *red heart emoji* ] and it shows 😍
have you found out about blob opera yet? I spent a couple days going through a bunch of covers people made with the 'experiment'. its an a.i powered choir.
14:20 flipped all those layers and ended up sounding like a Blanck Mass song
Thanks. I'm glad someone figured this out
This is a really cool concept. I wonder what adding like visual distortion does to a guitar like does it actually distort it?
I use png to wav to add a visual watermark to my sound effects in the same way Mick Gordon made the pictures in the Doom soundtrack.
Nobody:
3:39 Travis Scott: thats fire 🔥🔥
- Hey man, what's your DAW ?
- Have you heard of MS Paint
why is this actually kinda cool
2:09 the best banger dubstep ever
now you can save music to paper!
This is insanely cool
Thanks for this video
Helped me a lot
Very sick vid honestly 👌
Yo that's so cool
2 Levi Niha's is a lot of Levi Niha's
this could be used to hide a message in an ARG :D .
That what i was thinking XD
Gave me an idea
You are crazy. I'm here for it.
Oh hell yea, another Affinity enjoyer.
I did made dubstep glitch sounds way back in my college day's .
It was fun .
But you must use EXE files or folders files .
Something of that jazz .
sending stems as pngs is actually a great idea
I wonder if it reverses if you flip the image
Edit: no it doesn't, it just makes its sound glitchy
This could be crazy for making glitchy/weirdcore/hyperpop/experimental tracks
An assault both on the eyes and the ears.
Magnificent 😂😂
You're a legend, you should make a skillshare course because you're amazing at this! I'd love to learn from you!
I think the way you are assuming the audio got encoded as the image might be the reason the experiments didn't sound great.
I'm pretty sure the audio is encoded into one single horizontal line of pixels that is then wrapped vertically. It should be read right to left, and then when you reach the end of the line on the right it continues on the next row on the far left.
Think of it like reading this comment. If you did a vertical motion blur all you are doing is duplicating letters vertically across words in different lines.
Example:
This is a sentence about ducks. Quack!
Gets incoded as:
This is a sentence
about ducks. Quack!
Turns into:
Tahbiosu t idsu cak sse.n tQeunaccek !
Tahbiosu t idsu cak sse.n tQeunaccek !
Which would turn back into audio as:
Tahbiosu t idsu cak sse.n tQeunaccek ! Tahbiosu t idsu cak sse.n tQeunaccek !
Rather than:
TTThhhiiiss iiisss aaa ssseeennnttteeennnccceee aaabbbooouuuttt ddduuuccckkksss... QQQuuuaaaccckkk!!!
This is why everything got stuttery. Because you are blurring the sounds vertically across multiple rows selection of time stacked vertically. Even when blurring horizontally, the blur doesn't wrap with the pixels so the audio at the left and right edges gets messed up.
If you could unwrap this image into a single horizontal row of pixels I bet the blur might sound a bit more like reverb.
yay new levi niha video :D
my friend : what songs you listen to
me : it’s complicated...
that random sample is used in a song with 122M hahaha
Rels B, Dellafuente - BUENOS GENES
me he quedado igual al escucharlo JAJAJAJ
i wonder what a picture of you would sound like XD
Now you gotta make a song where you hide some memes or something in it. The sounds come out very strange though when you use a picture and I think it may have to be black and white. I played around with doing it a while ago but couldn’t find a good way to hide the sounds without it seeming obviously out of place
If you want multiple layers to show up the same amount, you can’t set the opacity for 25 for all of them. You need to make the bottom 100, the second 50, the third 25, and so on. Otherwise the top layer shows well but the lower ones are all fighting to be seen
Damn, this is really cool
Awesome video! You should have compared the original loop to one that was converted to an image and then directly converted back into sound so you can see how much information is lost in the conversion process every time
Now I can definitely prove I can hear images
I have that same shirt.... i get compliments on it every time i wear it. So... nice shirt! Lol
a DAW you never thought you needed.
This will be future music production
Photoshop is my favorite music program
I did this and I did content aware fill on a part of the png I erased and you could here other parts of the song blended in. It was super interesting.
this is such a brilliant idea! that’s so sick omg
This would be a beast for Lofi hiphop
Do this again, but with your knowledge from this video! This could give a unique style if mastered!
there are actually other ways to databend. if you convoert an image to bmp(bitmap) and throw them into audacity you can do some pretty funky stuff.
Very cool idea
Looks cool . i tried chopping and miss placing pieces over top each other gives cool effect if just doing vocals
Heeyy levi i am here to BEG YOU to do an extended version of the old outro that sounds like a distorted violin
I just love it too much
Yeah but mix the two sounds!
Also have you tried saturation, contrast and other light effects? For example tint? Or patterns?
i'v done this with Adobe audition and paint shop pro, but i used bmp, with raw import in psp.
i assume this app work in pure bitmap, yes? .. as far as i remember png often deals with compression.
however in my end its more rainbowish but the reverb effect was similar.
i also think i used 16 bit audio which will result in a odd 1 + an half PCM per color, but if i used 24 bit audio , each PCM would be each color pixel.
** ok going down the rabbit hole, and figured out i had more success with 8-bit audio to 8-bit grayscale image, other only distorted audio since i guess i didn't get a proper 1:1 convertion
So this is what they mean by hearing images. Thank you sir
great experiments. you could use izotope rx8 to get rid of the pops