Mastodawn

🧵 … at ~1200€, #Alphalabs #Defeedback is quite expensive for a solution that you have to explain to your client in #EventTech. Perhaps certain clients will like it because it is #AI, and AI is always the modern way to go. (So, they might like it for the bullshit, not for the effect.)

If the main contractor in my last #CorporateEvent would have provided me with cardioid headset #microphone|s instead of omni ones, this would probably have reduced feedback on a similar level already.

De-Feedback does not do the wonder of eliminating feedback from every bunch of shitty gear used in the wrong way by volunteers. Yet this is what advertisement videos of the manufacturer suggest it would do.

As a preliminary result of my test, it De-Feedback appears to be a very tasty icing on the speech sound cake.

I certainly would love to use it. Yet if the icing is more expensive than the rest of the cake, clients won't pay me for this.

Show thread

Nielso 1d ago

🧵 … so #Alphalabs #Defeedback does reduce feedback, it's certainly not a hoax.

Alphalabs offer a dedicated Windows 11 IOT PC that is optimized for De-Feedback for $699. The license for the plugin is $499. Without connectivity, which can be a simple Focusrite 2i2 interface (analog I/O), your console's USB port, if it has one, or something like the #RME #Digiface #Dante (~1300€). It also runs on #Waves #Livebox ($7000).

A #MacMiniM4 with Waves Superrack Performer would be around 700€, too.

So, approx 1200€ for one single local #AI solution that does its thing but no wonders, if you own the I/O already.

(2/x)

Nielso 1d ago

Testing #Alphalabs #DeFeedback for #EventTech. This is an #AI #microphone #feedback remover.

And as usual with AI, people say it does wonders. First impression: Also as usual with AI, it can do its thing, but no wonders.

This is not to be confused with generative AI running in data centers. This is more like traditional #MachineLearning. It runs entirely local, and my #Waves system based on a #MacMiniM1 can run 1 instance of it at 64 samples buffer size with a #Rme #Digiface #Dante. Which is 2 channels of de-feedback. The #plugin itself runs at zero latency.

This is all about gain before feedback: How far can you push the faders before feedback sets in?

While this is easy to measure with traditional tools like feedback destroyers, EQs, automixers, it is hard to say here. Because this tool reduces certain frequencies in a way that apparently depends on many things, not entirely on gain.

🧵 1/x

Nielso 4d ago

Wieso reden wir eigentlich nicht so langsam mal von #KiKrise?

Immer mehr sehe ich Artikel in Online-Medien, dass AbsolventInnen keine Jobs mehr kriegen wegen #KI.

Also aus meiner Warte: Weil Firmen glauben, Berufseinsteiger durch KI ersetzen zu können (und sich damit den Nachwuchs abwürgen, Resultat kommt noch).

Der Markt für Gebrauchs-Kreativität wie Illustration, Animation, Musik für Imagefilme, etc. ist schon länger eher tot.

Der Umweltverbrauch, insb. in den USA, durch KI-Rechenzentren ist enorm. Speicherchips sind sehr teuer geworden für normale Kunden, wegen KI.

Wenig überraschend scheint KI vor allem ein Tool dafür, Geld aus der Breite in die Höhe zu schaufeln.

Wenn ich das #LLM frage, ist der Anteil inhaltlich falscher Antworten enorm hoch. Auch das kostet, Zeit, Nerven… Mal ehrlich: Würde ChatGPT kosten, ich würde nicht dafür zahlen, es ist ein unbefriedigendes Produkt.

Es bleiben nützliche Anwendungen wie #DeFeedback… aber das läuft lokal auf dem Rechner und kostet $499.

Nielso 4d ago

So while I was experiencing some issues with omni headset #microphone|s today, as usual in #eventtech in corporate events without soundcheck, a colleague used #alphalabs #defeedback on another stage of the event, apparently succesfully.

De-Feedback is an #AI-based #audioPlugin for Windows and macOS, and depending on whom you ask, it's a magic wand, or not that helpful.

It's $499, and you need a decent host to run on (they sell a $699 box with Windows IOT 11), or #Waves #Livebox ($7000), I myself have a MacMini M1 with a touch screen and an #Rme #DigifaceDante (~2000€) that could handle ~2 channels of it…

… But the difficult part is to convince the client that he should pay for that feature that nobody needed before. Because if you screw your sound with feedback-preventing EQing, it's still good enough for corporate events.

Same holds for Waves #xfdbk, a pretty decent clasic #Feedback destroyer.

I must say I am a little frustrated. I know how I could improve things, but it's not working out economically.