Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[methylatedspirit]

I suppose that could be "a thing". I like money, but I'm not sure if I want to take money based on an inconsistent release schedule. I know that's what Dwarf Fortress does, but I guess I'm hesitant about the idea. I'd basically be a freelance researcher. It would be nice to have my fans vet what I'm publishing, though. I try my hardest to be factually correct, but I know I have to be wrong somewhere.

"Purchasing my finest works"... what "finest works"? I maintain a policy of reproducibility first; literally anyone with the right software can make near-identical, if not identical copies of my work. I even put my stuff under CC-BY (or MIT for my code) so that anyone can go around distributing copies of my work, so long as you give attribution and generally follow the licence. I still hold the copyright. You can't retroactively revoke CC licenses, however.

[EuchreJack]

Apparently, there is a market for digital originals: The "First" image.

[methylatedspirit]

Added an addendum to (Ab)using Lossy Audio Compression to Databend Images, Automated image to audio 'bending.

Apparently, there is a market for digital originals: The "First" image.

The "first" image... I guess? It's one of those things where it kinda breaks down when you have perfect replication instructions, no? The idea of an "original" for something that can be so easily replicated (I even gave detailed instructions!) is a bit odd.

Then again, Paulstretch is the one thing that isn't predicable. It produces roughly the same "thing" each time, but it's random where all the colors will end up. Case in point, these two. Same image, two Paulstretch invocations.





The SHA256 sums of the originals are:

Code: (1.raw) [Select]

a9a659adbd91d679bc4055bcb4e517e7229b8ed0282981e02d0ca4199c7540b2

Code: (2.raw) [Select]

2fdbfd2d7fa0e3c7b5ce9e71071a0a4becf664906b0a57591cee97cfc0149181
They're not the same file.

[DogsRNice]

How much experimentation has been done with databending audio formats?
Editing audio as a raw image in what’s basically the reverse of normal data ending
or using the wrong audio codec

[taat]

Most audio and image formats are compressed, which means interpreting them as raw data gives just random noise

[methylatedspirit]

Compressed is one thing-- corrupted FFV1 (without error-detection enabled in the encoder settings) can still be reliably "decoded", despite being compressed. But most formats have error-detection. They will not decode at all if the input isn't correctly checksummed or isn't the right data format.

Though, on the wider topic of doing weird things to data, I've been able to concatenate Vorbis files, used FFmpeg to "correct" them, and got some... interesting noises out of them. Maybe I could try concatenating different formats entirely, and see what happens as the decoder tries to decode invalid data.

[DogsRNice]

Most audio and image formats are compressed, which means interpreting them as raw data gives just random noise

There are raw and uncompressed audio and formats (it’s what the raw data importer is designed for after all)

[methylatedspirit]

How much experimentation has been done with databending audio formats?
Editing audio as a raw image in what’s basically the reverse of normal data ending
or using the wrong audio codec

There are raw and uncompressed audio and formats (it’s what the raw data importer is designed for after all)

Understood the context after the latter post. I've already done some work on that, under Databending with FFmpeg: Misinterpreting Audio as Video for Fun and Profit. That deals with taking raw audio data, converting it into a video, decoding its raw data, and getting the final result. Images tend to be a bit finicky, and they don't really lend themselves well to automation like videos do. Plus, not many lossy image codecs.