The No.1 Website for Pro Audio
 Search This Thread  Search This Forum  Search Reviews  Search Gear Database  Search Gear for sale  Search Gearslutz Go Advanced
My LANDR Test Multi-Ef­fects Plugins
Old 1st September 2018
  #61
Lives for gear
 
robert82's Avatar
Honestly, I can't imagine how anyone who is going for a marketable, serious, finished sound would trust their project to an automated mastering service. I think Fabien is right on - that the automated services are targeting this whole new market of amateurs and enthusiasts, made possible by the proliferation of great digital recording technologies. I'd be interested to know how many professional mastering engineers are truly worried about AI "mastering".
Old 1st September 2018
  #62
Gear Nut
 

Broadcast processing can only homogenize the input. It can't for example make the groove work better or enhance the best attributes of a mix to support the content, more so while minding translation. We certainly use robots in the form of audio processors for a lot of tasks but they can't shine by themselves.

There is demand of course, and Landr is a great product catering to the market that doesn't need/want mastering. Some just don't know, others don't want anyone else's creative input or have bad experiences with poor mastering, others just see it as an extra option or safety net that costs next to nothing. The fact that Landr is unable to master is its main advantage in this case (besides price and speed) and I don't think they were aiming for that at all, but rather ended up with something people might use by restricting its processing span to a tight range.

Landr could do better even with less intelligent algos if the user could steer the result via a targeted questionnaire. It would also do better if it had its framework set up purely for what its real strength is: a sense of safety and a pat on the back. I would imagine that if Landr is taking someone's work, it's likely they weren't very careful, or they weren't in the business of mastering to begin with.

AI projects with massively more capital and research behind them are useless when it comes to the real party, so I think it's impossible for puny Landr et al to deliver mastering any time soon.
Old 1st September 2018
  #63
DAH
Lives for gear
 
DAH's Avatar
Quote:
Originally Posted by z-plane View Post
Broadcast processing can only homogenize the input. It can't for example make the groove work better or enhance the best attributes of a mix to support the content, more so while minding translation. We certainly use robots in the form of audio processors for a lot of tasks but they can't shine by themselves.

There is demand of course, and Landr is a great product catering to the market that doesn't need/want mastering. Some just don't know, others don't want anyone else's creative input or have bad experiences with poor mastering, others just see it as an extra option or safety net that costs next to nothing. The fact that Landr is unable to master is its main advantage in this case (besides price and speed) and I don't think they were aiming for that at all, but rather ended up with something people might use by restricting its processing span to a tight range.

Landr could do better even with less intelligent algos if the user could steer the result via a targeted questionnaire. It would also do better if it had its framework set up purely for what its real strength is: a sense of safety and a pat on the back. I would imagine that if Landr is taking someone's work, it's likely they weren't very careful, or they weren't in the business of mastering to begin with.

AI projects with massively more capital and research behind them are useless when it comes to the real party, so I think it's impossible for puny Landr et al to deliver mastering any time soon.
You dont understand how quickly it will evolve. Now it is just the beginning. Soon the analyse will be done in deeper layers, like recognizing the genre, monitoring the genre's top hits and preferences and listening chain etc. and creating a dynamically adjusted to all these factors output. Hire me, LANDR lol.
Old 1st September 2018
  #64
I wouldn't believe these A.I. tales.

A simple reality check. Google, Amazon and Facebook lead in A.I development, both in man power and budget. And here's the shocking aspect: Everybody can see how impressively cheesy this A.I. works when targeting ads (these companies' core business!).

Same is true for robot-advisors in the financial industry. They obviously do not work (this is well documented).

Basically, A.I. fails at anything that involves the human psyche. BTW, a fact educated informaticians have long been aware of.



BUT, where it shines is in the business of automating simple, banal mechanisms, and finding equally banal patterns at great speed. That's what state of the art A.I. does today.
Old 2nd September 2018
  #65
Lives for gear
 
Ben F's Avatar
 

Verified Member
Can I just correct a term.

This is not AI, there is no “intelligence”, it’s machine learning marketed as AI.
Old 2nd September 2018
  #66
Lives for gear
 
robert82's Avatar
Quote:
Originally Posted by Ben F View Post
Can I just correct a term.

This is not AI, there is no “intelligence”, it’s machine learning marketed as AI.
Good call.
It's not like Skynet is taking over mastering duties.


Last edited by robert82; 2nd September 2018 at 04:59 PM..
Old 2nd September 2018
  #67
Gear Nut
 
ChrisNunchuck's Avatar
I find Landr service to be interesting from an academic perspective but also recognize they have a demographic that benefits from their service. Why are so many up in arms when we already have similar approaches with things like Toontracks EZMix or Ozone's Master Assistance? I have demo'd Ozone and found the master assistance regularly on a different page than myself but on some occasions it did recognize a suggested frequency bump or dip that I might not have considered yet. Even the Sonible EQ does some automated things which I take as a nice suggestion to speed things up so I can focus on music.

Landr's 4 million in seed money came from investors that feel this business model has potential. I could see their platform sold to another company that would combine it with another music (hosting? or streaming?) service they are already offering. Nice insight FabianTDR.
Old 3rd September 2018
  #68
Quote:
Originally Posted by eternalsound View Post
Exactly. There is nothing "intelligent" about software. It's flawless with what it's been programmed, but is endlessly dumb beyond that.

Try telling/asking Amazon Alexa: "I just had chicken yesterday and I'm in the mood for something else. What do you think my stomach is calling for?"

You'll probably find the computer logic ...quiet dumb after that one.


That's the nature of what auto-master deals with too in other senses. Dumb ..dumb ..dumb.

Not much dumber than "I had a rough night, didn't sleep well so I will put too much bass because reasons"
Old 13th September 2018
  #69
Quote:
Originally Posted by gregorydutchman View Post
hello slutz

ive made my own comparison with landr after hearing so much about it from beginning artist that want a quick pass in better sound on a low budget .. so i thought leys tango ... no better lets do it matrix style shootout ..

YouTube


so took a mix from a producer that was already compressed a bit and very full and some errors and gave it to landr . landr masterd the wav and then we went to the studio and masterd it also and put them up against eachother for comparison. the priducer wanted it to sound fresch like the sam smit song no money on my mind.. that is very uogront fresch and loud sounding . so with that in mind we got to work.. we did not altert levels or gain in the comparisson video .. just what you get vs what we deliver ..

my personal opinion is that landr is a hoax, but check the link and let me know what you think . or if you not made up your mind .. now you might .

take care
I prefer the LANDR version, less compression and not as bright and distorted as yours
Old 13th September 2018
  #70
Here for the gear
 

Verified Member
Quote:
Originally Posted by cyjanopan View Post
I prefer the LANDR version, less compression and not as bright and distorted as yours
thanks for the feedback .

yes every one has its own flavour , but its not distortet. making every thing sound dark with blankets on top was just not the goal of the client its open and you hear every detail now. and thats a good or sometimes a bad thing .. compression yes . the ref track was very hot to match from sam smith money on my mind with rms levels around -4 db . most important the client was very happy with the result. it will defitnitly win the war on loudness lol

take care

Last edited by gregorydutchman; 13th September 2018 at 07:06 PM.. Reason: spelling mistake
Old 15th September 2018
  #71
Gear Maniac
 

The main issue with today’s A.I. implementations is that up to date computer architectures were not designed to deal with problems A.I. needs to solve in order to make it effective, they were designed to perform extremelly fast basic arithmetic. Today it just allows to collect and process (not real time however) big amount of data in a statistical way, but (un)fortunately this is not how intelligence works by any means. Machine learning is just a very small part of the equation and it is nothing more than just recording and reusing user response, that still needs amount of supervision to work properly. This is one of the reasons why even simple A.I attempts fail at this point of time and space .

On the other hand, far before A.I term became that appealing, there were powerful technologies and tools devloped to the extents to allow easily mislead minds with just few simple buzz words.

Art
Old 21st September 2018
  #72
Here for the gear
When technology can get to the point where Excel doesn't say "not responding" every 5 minutes, then I might look into A.I. as something viable.

I think something like LANDR might work well for something like "library tracks", where someone just can't afford to have 100 tracks mastered in one go. Otherwise, may as well master yourself with Ozone 8.

That being said, I will always get my more "personal" musical releases mastered by a professional human mastering engineer, without question.
Old 21st September 2018
  #73
Lives for gear
 
robert82's Avatar
Quote:
Originally Posted by nn2079 View Post
When technology can get to the point where Excel doesn't say "not responding" every 5 minutes, then I might look into A.I. as something viable.
Aaand, who created Excel?

Billysoft wouldn't be my first choice for pioneer in AI.
Old 22nd September 2018
  #74
Lives for gear
 
ionian's Avatar
Quote:
Originally Posted by robert82 View Post
Aaand, who created Excel?

Billysoft wouldn't be my first choice for pioneer in AI.
I still think one of the most glorious events ever was when Microsoft unleashed their A.I. twitter bot "Tay Tweets" upon the world and in under a day she became a racist, sex obsessed slut causing them to pull her off twitter in a frenzied panic. I have a folder on my hard drive of her tweets and each one is funnier than the next. They must have been going into seizures at Microsoft once they let her loose and saw what direction she went once she started "learning" and forming her own opinions hahaha
Old 22nd September 2018
  #75
Gear Guru
 
UnderTow's Avatar
 

Verified Member
Quote:
Originally Posted by ionian View Post
I still think one of the most glorious events ever was when Microsoft unleashed their A.I. twitter bot "Tay Tweets" upon the world and in under a day she became a racist, sex obsessed slut causing them to pull her off twitter in a frenzied panic. I have a folder on my hard drive of her tweets and each one is funnier than the next. They must have been going into seizures at Microsoft once they let her loose and saw what direction she went once she started "learning" and forming her own opinions hahaha
Which only really says something about the quality of discourse on Twitter as that is what the bot was learning from. Basically the project got trolled.

Alistair
Old 22nd September 2018
  #76
Gear Nut
 
eternalsound's Avatar
Try LANDR.

Then compare to an ME.
Old 22nd September 2018
  #77
Lives for gear
 
Thomas W. Bethe's Avatar
 

Verified Member
Quote:
Originally Posted by eternalsound View Post
Try LAND#R.
Isn't that what this whole topic has been about??? Trying it???
Old 23rd September 2018
  #78
Lives for gear
 
ionian's Avatar
Quote:
Originally Posted by UnderTow View Post
Which only really says something about the quality of discourse on Twitter as that is what the bot was learning from. Basically the project got trolled.

Alistair
Oh, I know. Twitter stands as the example of the worst that social media can be. They're biased, they shadowban people, they let people threaten with violence and let pedophiles roam free, while instantly banning anyone it thinks is taking part in wrongthink. I saw screenshots a few days ago of someone who saw a pedophile on twitter describing it openly in detail so they reported the pedo to twitter and twitter banned the person who reported the pedo instead! The pedo suffered no repercussions. Twitter is nothing more than a vortex of sh!t who's time has long come and gone.

And yes, I know it got trolled. I remember when microsoft released it, I knew it was gonna crash and burn amazingly, I just didn't realize how quickly! I just thought it was a fun thing to bring up since you mentioned microsoft and A.I. and it's amazing how Microsoft never had the forethought to realize how bad of an idea the whole thing was.
Old 23rd September 2018
  #79
Gear Maniac
Quote:
Originally Posted by eternalsound View Post
The thread was about an individual's LANDR test.

My post was just strait to the point in that EVERYONE should try LANDR.

So, everyone ...TRY LANDR!!

No. But thanks for yelling it at me.
Old 23rd September 2018
  #80
Gear Maniac
Quote:
Originally Posted by ionian View Post
Oh, I know. Twitter stands as the example of the worst that social media can be. They're biased, they shadowban people, they let people threaten with violence and let pedophiles roam free, while instantly banning anyone it thinks is taking part in wrongthink. I saw screenshots a few days ago of someone who saw a pedophile on twitter describing it openly in detail so they reported the pedo to twitter and twitter banned the person who reported the pedo instead! The pedo suffered no repercussions. Twitter is nothing more than a vortex of sh!t who's time has long come and gone.

And yes, I know it got trolled. I remember when microsoft released it, I knew it was gonna crash and burn amazingly, I just didn't realize how quickly! I just thought it was a fun thing to bring up since you mentioned microsoft and A.I. and it's amazing how Microsoft never had the forethought to realize how bad of an idea the whole thing was.
Twitter makes money when it's users stay ferouciously engaged, it needs them to stay addicted so they can stay profitable and in business. People are more engaged when outraged or getting likes/retweets/comments, even with as many puppies and kittens there are to share GIFs of. It doesn't matter whether people messed up Twitter or whether Twitter messed up people, either way it's a platform that requires this cycle of addictive behavior in order to survive. It's a symbiote and users are the host.
Old 23rd September 2018
  #81
Lives for gear
 
Thomas W. Bethe's Avatar
 

Verified Member
Quote:
Originally Posted by eternalsound View Post
The thread was about an individual's LANDR test.

My post was just strait to the point in that EVERYONE should try LANDR.

So, everyone ...TRY LANDR!!

And that begs the question "WHY"?
Old 1 week ago
  #83
Gear Maniac
 
artnoiser's Avatar
 

Quote:
Originally Posted by Thomas W. Bethe View Post
And that begs the question "WHY"?
For education, for context, for empirical purposes. Paying for one month doesn't cost much, and if a lesson is learned in what automated mastering can and can't do, it's well worth the handful of dollars it costs.

There aren't just two options - DIY or a professional mastering engineer. There are a ton of musicians sending their stuff to friends who have no idea what they're doing too. The amount of possibilities to fudge up something has no end, even when it comes to a semi-pro who has a solid idea of what he or she is doing, but doesn't end up doing it the way the client wanted to.

I firmly believe that the human element in mastering (mixing, recording, composing, etc) is integral in translating emotion through musical expression. But if one wants quick "mastering" (aka "maximizing and loudening") of demos, soundbites, samples, or what-not, Landr isn't completely out of the picture. Just because Landr doesn't doesn't fulfill the potential of the goodness that could be done to a piece of audio, doesn't mean that the person needing mastering willa actually look up or find the best possible solution for his needs.

Without wanting to sound ungracious to people with potential...some people aren't really at a place where they could actually top what an automated process could give to a piece of audio, even if the automated process may be riddled with imperfections (and I'll mean that generously).

Having said all that, I'm sure some tracks will be mastered with Landr based on the illusory promise it presents when it would have been better to hand those tracks into the capable hands of a skilled person. It happens. People go and record, get their stuff produced, mixed by persons who aren't as good as the ones who are clearly way better. Whole records by bands who started out could have been serviced better, without any doubt whatsoever. But it happens, and it happens every day. It's unfortunate, and it's real.
Old 3 hours ago
  #84
Lives for gear
Quote:
Originally Posted by FabienTDR View Post



BUT, where it shines is in the business of automating simple, banal mechanisms, and finding equally banal patterns at great speed. That's what state of the art A.I. does today.
Wouldn’t a simple algorithm do just as well for automating simple banal stuff?

But anyway machine learning has proven itself by beating the best humans in complex games with myriads of potential strategies and outcomes like Go (board game) and Chess. You couldn’t teach a monkey to understand the rules of chess.

The future of AI will be for setting interest rates at central banks because humans are too easily biased and swayed by short term interests.
Old 2 hours ago
  #85
Quote:
Originally Posted by miscend View Post
Wouldn’t a simple algorithm do just as well for automating simple banal stuff?
Depends what you mean by "simple algorithm". Most human beings are already overwhelmed by standard arithmetic.

Yes, for a traditional engineer well familiar with the state of the art, it will be easy to stick the right solution to the right problem, and end up with sensible results. Most normal people however will interpret this machine as magical, if not even intelligent.

Path finding algorithms, advanced equation solvers, fuzzy logic, fast search algorithms, advanced data compression, fractal generators, or advanced control systems all feel intelligent (while still being only made of "traditional" techniques).

Advanced control systems like driving assist or fly by wire are magical. Even basic controls systems like the PID feel smart:



Machine learning is primarily associated with trained neural networks. They have impressive advantages in certain areas (finding and comparing patterns), but also come with clear limitations.


Quote:
Originally Posted by miscend View Post
But anyway machine learning has proven itself by beating the best humans in complex games with myriads of potential strategies and outcomes like Go (board game) and Chess. You couldn’t teach a monkey to understand the rules of chess.
Thanks for mentioning this. The limitation is the assumption of a fixed rule set. This is simply not existent for anything related to the human psyche, i.e. a good artist knows that the game constantly moves, the only rule is that there are no rules.

Any of the artist's "move" directly changes the underlying rules. The same is the case for robot-advisors/automated trading: Works for 3 years at best, until it doesn't. Simply because humans are impressively creative at responding to completely new environments and situations.

Maybe we'll understand to synthesize this "divine touch" at some point, but today, we really aren't there. Not even close!
Post Reply

Welcome to the Gearslutz Pro Audio Community!

Registration benefits include:
  • The ability to reply to and create new discussions
  • Access to members-only giveaways & competitions
  • Interact with VIP industry experts in our guest Q&As
  • Access to members-only sub forum discussions
  • Access to members-only Chat Room
  • Get INSTANT ACCESS to the world's best private pro audio Classifieds for only USD $20/year
  • Promote your eBay auctions and Reverb.com listings for free
  • Remove this message!
You need an account to post a reply. Create a username and password below and an account will be created and your post entered.


 
 
Slide to join now Processing…
Thread Tools
Search this Thread
Search this Thread:

Advanced Search
Forum Jump
Forum Jump