Jump to content

Goiter Sanchez

Members
 View Profile  See their activity

  • Posts

    5,692

  • Joined

  • Last visited

  • Days Won

    1

 Content Type 

Posts posted by Goiter Sanchez

    Dostrotime [2024]

    in Squarepusher

    Posted

    Enjoyed the listening event earlier while working in the garden. Hilarious how fast the chat was moving, but I saw a few familiar names pop up 😉

    The album sounds really excellent and I look forward to some more focused listens soon. Akkranen (I think it was called) stood out to me.

    • Like 2

    Underworld - And The Colour Red (2023)

    in New & Upcoming Releases

    Posted

    23 hours ago, LooseLink said:

    Reminded me of "Dark Train", and at times "Kittens", and another song of theirs I can't quite remember atm..... up until when Karl goes "STRAWBERRRRRY JAM GIRL" etc anyway. I think centuries from now future generations will discover Underworld, and maybe they will get what Karl was ever even on about lyrically. But yeah dark and progressive though again, liking it!   

    "STRAWBERRRRRY JAM GIRL" is the most Karl Hyde lyric ever lol

    • Like 1

    New Autechre interview

    in Autechre

    Posted

    1 hour ago, exitonly said:

    i hope this is one where they talk about doing acid with their mates and listening to electro

    They do! Sean also mentions his love for 'Acid Eiffel' which was a nice surprise!

    Phenomenal interview overall. Love the compositonal insights, and how they both describe this ambiguous emotional state that this sort of music can elicit.

    • Like 1

    Stars of The Lid's Brian McBride dies at 53

    in Music Discussion

    Posted

    https://pitchfork.com/news/stars-of-the-lids-brian-mcbride-dies-at-53/

    Never take an artist's work for granted as any album can be their last. SoTL have released some of my favorite music and, while we still have Adam Wiltzie working in the 'A Winged Victory For The Sullen' project, there is a decadence in the SoTL records: a willingness to let a single relatively raw idea simmer and float for a long time in a way that feels relatively unrestrained compared to Winged Victory. I'm forever indebted. RIP

    stars-of-the-life-brian-mc-bride.png

     

     

    • Like 5
    • Sad 8

    Clocolan - Incide (August 17, 2023)

    in New & Upcoming Releases

    Posted

    It's definitely a darker, more discordant record. He posted some interesting info on social media the other day:

    Quote
    'incide' was a look at how #AI invents identity but also blurs and corrupts it through iteration: prompt an #AI engine with an input and you get an output; prompt it again using that output and you get another, slightly different output—a variation. Which one is the real McCoy?
    Likewise, our virtual identities are also inventions, endlessly blurred and corrupted through iteration and variation. Those identities feel strangely artificial—in the same way that #AI generations do.
    What is "real"? (And does it matter whether the inventor is real or not?)
    OK, not to get all Academia on this...
    My initial strategy was to explore [then still nascent] #AI technology and hopefully find novel ways to express that idea: to use it as a sound treatment tool that might somehow obfuscate the "identity" of real voices and instruments.
    That took me down several rabbit holes...
    The first discovery was Magenta DDSP, a DAW "timbre transfer" plugin for which you could train your own models on Google Collab. Training models is tedious, trial-and-error work but I began noticing some behaviors in the "errors".
    The little audio artifacts and playback oddities ended up being useful as a way to control performance (more on this below). Crucially, I got good at sussing out what would make a blah model vs an interesting one: high vs low variation in the training data was key.
    I wrote a lot of dialog about identity and reality for the album. Those recordings produced a model that I used on several tracks. On 'hexis' and 'i want', rather than standard vocal doubling, I used the AI model, tweaking pitch, tone, and timing.
    I discovered that you could control how the plugin responded to incoming audio: routing a Moog into it while altering attack, tone, and vibrato. The plugin reacted to changes by generating predictable elements in the model—vocoder-like. See 'if we can still feel'.
    This was also used as a rhythmic device: by first routing the incoming audio through a filter with a square wave LFO, the plugin would react to changes in tone at tempo. This led me to create a model using the Moog itself and applying it as a rhythmic element (same track).
    Concurrently, I had been experimenting with audio spectrograms, first converting audio to image via @convertingnow's Audio Spectrogram Creator and back again with this convertor: https://nsspot.herokuapp.com/imagetoaudio/. Essentially as a way to recast audio with accumulated errors.
    I used the spectrogram output in several ways. One was to append all its variations to the original recordings and have Kontakt chop it up during playback. 'projectors-introjectors' is a good example of this:
    But one of the most interesting applications was using @harmonai_org's #DanceDiffusion Collab to interpolate the spectrograms with the original audio—essentially isolating shared frequencies. (Reminded me of the spectral shapers in Tom Erbe's Soundhack.)
    Interpolation was a chore and most of it was unusable but the speech elements in 'archetypes += n' and 'hypercathexis' is all interpolation using this method.
    (The opening bars use timbre transfer between FM tones and a tenor voice model.)
    The above example used another element that became central to 'incide'. I had latched onto to the idea of using piano resonance as a way to color instruments and reverb. I also recorded variations on these: singing into the piano sound board while holding cluster chords.
    All of the piano works use one or more of the piano resonance IRs but 'simulacrimosum' is an example of the vocal/sound board resonance effect (triggered here by snippets of dialog).
    #DanceDiffusion also became my noise generator: plugging in noise samples like hiss, vinyl, or just white noise produced very interesting and usable material. Without a noise model the engine seemed to default to the closest matching timbres it could find.
    These would most often be approximations/combinations of rain, hail, wind, water turbulence, ice, fire, electronic glitches, and unidentifiable ones. If #AI is good at anything it's generating noise with no discernible identity.
    Using the output as the input was effective here.
    Musika by @marco_ppasini initially seemed like a plaything rather than a creative tool but as I worked with it I discovered that drum sounds generated by #AI have a thickness and fullness like old analog tape. It got me thinking about rhythm applications.
    Then I stumbled on @tensorpunk's MACE. #AI's facility with noise and the sound quality of generated drum samples made an #AI-powered drum sampler the perfect vehicle—sonically and narratively. A few compound meters later and I knew it was right for 'incide'.
    I struck up a conversation with Jordan Davis, the brains behind MACE, and we talked #AI and music. Jordan's approach was to favor uniqueness over fidelity; do we want *accurate* percussion and effects or *interesting* ones? MACE is all over this album:
    Likewise, I soon discovered that Harmonai's engine was very good at generating interesting drum samples. Thick kick drums and noisy snares with impossible tails ('if we can still feel'). And variations on piano resonance IRs:
    Tuning was another approach for obfuscating identity: Is the note an A or an Ab? Or something in between? I used discrete, scale-based detuning across the board, most noticeable in the piano+Moog interludes and pads.
    In @bleep's write-up for the album they write "on 'incide', the machine sings".
    And the machine is just as real.
    Honorary mention: The #minidisc release of 'incide' wouldn't really have been possible without Stefano Brilli's (@thecybercase) excellent Web MiniDisc Pro interface which made duplication almost effortless. Grazie!

     

    Forum Discussion/Issues

    in General Banter

    Posted · Edited by Goiter Sanchez

    I think watmm was better before post reactions. The ability to 'react' removes some impetus to type out an actual reply to a post. Watmm just feels less alive since it started.

    EDIT: sad reacts only

    • Thanks 1
    • Sad 6

    WAPP480

    in Aphex Twin

    Posted

    2 hours ago, chronical said:

    just noticed that the QR track is playing here too (from 0:30), really great video for anyone who hasn't seen it

    Love this, and lush AFX playing underneath.

×
×
  • Create New...

Important Information

We have placed cookies on your device to help make this website better. You can adjust your cookie settings, otherwise we'll assume you're okay to continue.