No it doesn't
"The panel held that the DCMA requires copyright holders to consider fair use before sending a takedown notification, and that failure to do so raises a triable issue as to whether the copyright holder formed a subjective good faith belief that the use was not authorized by law."
https://law.justia.com/cases/federal/appellate-courts/ca9/13...
Have I missed some precedent that overturns this?
The affirm part is you conceding you committed the crime, and then you argue it was justified.
Lenz didnt confirm it was an affirmative defense, they ruled the way they did despite admitting it was an affirmative defense, one that can only procedurally arise after you've been accused of a crime.
> According to the Supreme Court of Canada, it is more than a simple defence; it is an integral part of the Copyright Act of Canada, providing balance between the rights of owners and users.
"We agree with the district court and hold that the statute unambiguously contemplates fair use as a use authorized by the law. Fair use is not just excused by the law, it is wholly authorized by the law."
In defense of this (among other arguments) they cite portions of the US code that explicitly refer to fair use as a "right":
"See also 17 U.S.C. § 108(f)(4) (“Nothing in this section in any way affects the right of fair use as provided by section 107 . . . .”"
https://cases.justia.com/federal/appellate-courts/ca9/13-161...
So while your understanding of fair use as merely a defense against infringement may be widely held, and arguably is even the historically accurate position, it doesn't reflect the way that the courts currently interpret the statutes.
Now I agree with your interpretation to an extent, however its only the 9th circuit court. Lenz is also odd in that the person claiming fair use is the plaintiff, and the suit itself is that a DMCA claim was frivolous. The video was restored after her counter-notification, and her lawsuit was later.
A DMCA taketdown request is itself not a lawsuit. So my interpretation of the ruling is not that "you cant sue someone" but that you "cant blast out DMCA notices." I read it as an anti-intimidation ruling. It is clear, that it sees fair use as authorized by law, but it also still categorizes fair use as an affirmative defense.
Ergo, fair use gives you a defacto right to not be harassed.
The only time it's closer to trivial to filter out is when the music has been edited in as an audio track, and no additional filters/EQ have been applied. But even then, subtracting a lossless original will leave funky-sounding audio compression artifacts, a "ghost" of the song left behind.
But 1. that's not YouTube's problem, and 2. most content-creators would prefer a segment of weird artefacted audio to their video being demonetized.
For example, let's say most of your voice falls within a range of 120hz to 160hz and there is a guitar playing in the background at various frequencies where a majority of it happens to match your voice's frequency. Now your goal is to cut out the guitar, leaving a pristine copy of your voice.
How would ML or any strategy be able to differentiate a voice or guitar at 140hz?
I deal with a similar problem when recording my voice for video courses in a non-treated room. I have to set up a noise gate to lower the volume of certain frequencies (background computer fans, etc.) until you can't hear them. They tend to give off a low frequency compared to a human voice, but the more loud the fan is relative to my voice requires setting the noise gate to a higher frequency.
If you try to filter too much, then your voice gets cut off for those frequencies and it sounds horrible. It either cuts out entirely or you sound super dirty and metallic since your wave forms are all jaggy instead of smooth lines.
I'm no ML expert but I think this problem would be insanely hard to solve in a way that's high quality. You can bend your voice's frequencies and the frequency of various instruments in a similar way. Then on top of that, there's going to be situations where the singer's voice is going to happen while you're talking. That'll be even harder to separate if you criss-cross frequencies. And on top of that, you might have 2-3 guitars, drums, keyboard, a singer and your voice all hitting the same frequencies. So much overlap!
I'd love to hear the quality of a tool that claims it can remove only the songs out of any audio footage that may include your voice, background voices of other people talking as well as other ambient sound (water falling, people shuffling around, cars honking, etc.). If I were watching someone's vlog I would want to hear the atmosphere of the place. Making it completely mute except for their voice would be extremely weird and unsettling (and still crazy complicated to do I bet).
A known song could be used as a time-series with a similar effect.
Yes, this is what I used to do with REAPER (a digital audio workstation tool). It was even able to do this in soft real time (that's how I recorded my more recent courses) with only 10ms of delay.
But from my understanding it does this by dampening or removing (depending on how aggressive you have it configured) the frequencies of your room or whatever sound you're trying to remove.
A relatively low volume fan hum is pretty easy to remove without distorting the target audio, but if the frequencies start to overlap with varying degrees of volume the whole thing becomes a jumbled mess.
To put this into perspective, the video at the top of https://diveintodocker.com/ used REAPER's noise reduction plugin to remove a pretty loud tower and power supply fan. Even with studio grade headphones at max volume you probably won't notice a hint of hissing or white noise when I'm not speaking, but if you have a trained ear there's a subtle clamping down on the lower end of my voice's range. 99.9999% of people won't notice it, but this is also a best case scenario for removing frequencies (a constant low hum at relatively low volume) that rarely interweaves with my voice's frequency. The whole game changes with multiple instruments and volumes that cut across your range.
Even outside the scope of removing copyright material there's a lot of interesting use cases, like being able to split vocals or instruments from any track so you can practice layering in your own music as a musician (such as playing your own guitar solo over an existing song).
But great point though.
Ambivalent about stock tickers. Sometimes have to look them up, haha.