Perfecting Live Zeppelin Recordings Using AI

[Unseenguest]

Perfecting Live Zeppelin Recordings Using AI

I was inspired to write this post after reading about how director Peter Jackson, Post Road Post Production [NZ] and their team utilized Artificial Intelligence [ AI ] to demix existing audio of the Beatles from the extended footage from Twickenham Studios in 1969 used for the documentary ‘The Beatles : Get Back’ and improve the sound and quality of many of those low fidelity recordings.  I understand that in many ways this new AI software development and coding used to restore sound and improve audio is still very much a developing technology but as an enormous fan of Led Zeppelin, especially of their live performances, for those of you not already aware of using AI for audio I thought I would be a great idea to share this new technology with all the fine people currently listening to and/or restoring existing live Led Zeppelin recordings. Perhaps in the future one could apply this new technology combined with traditional audio post production methods and work towards creating the best remasters of available Led Zeppelin audio recordings so that we may enjoy them and preserve them for future generations.  Many thanks to the fine individuals who continue to appreciate the band’s music and contribute to the various Led Zeppelin online forums. 

The potential is exciting and as Robert Plant says during the Baton Rouge, LA 1975 concert: “ Hang on to your heads ! “

 

“To me the sound restoration is the most exciting thing. We made some huge breakthroughs in audio.” [ Audioholics.com - Jacob Green. March 13, 2022 ]

 

Jackson’s team utilized a technology called “demixing,” which, according to the director, is “an idea that’s been around for a while,” but has never been used with this level of success. The clever people at Park Road Post built custom software using an AI-based machine learning system that allowed them to teach the computer “what a guitar sounds like, what a bass sounds like, what a voice sounds like. In fact we taught the computer what John sounds like and what Paul sounds like.” Given an adequate education, the software was then able to analyze the mono tracks from Twickenham and split them into multiple tracks, thus allowing the music to be remixed cleanly, as if the voices and instruments had all been recorded to separate tracks.  [ Audioholics.com - Jacob Green. March 13, 2022 ]

 

For those interested in more details and examples on the AI process of audio source separation, here are some links:

James Clarke, Senior Software Engineer from Abbey Road Studios, London, UK on audio source separation / demixing for production : 24 minutes

https://www.youtube.com/watch?v=j4BvmpMLeFk 

Giles Martin, sound engineer at Abbey Road Studios, London, UK highlights improvements on The Beatles Track ‘ Taxman ‘ using AI and audio improvements: 2 minutes

https://www.youtube.com/watch?v=Z6yUGrbNrvA 

Marc Gosselin - Remixing old 1960's recordings with AI technology (A/B comparison) - The Beatles - Ed Sullivan Show / 1 minute.

https://www.youtube.com/watch?v=odg40BPSEmQ

 

 

Jimmy Page quote on recording :

“The whole idea, the way I see recording, is to try and capture the sound of the room live and the emotion of the whole moment and try to convey that across.  You’ve got to capture as much of the room sound as possible. That’s the very essence of it”

 

 

Fan Remastering of Led Zeppelin live.

There have been some excellent remastering of live Led Zeppelin concert audio over the years, especially the past 10 years or so as improvements to consumer and ‘prosumer’ [ an amateur who purchases equipment with quality or features suitable for professional use ] audio production software has become more advanced.  Many thanks to all the dedicated individuals who have spent their time and money taking existing Led Zeppelin recordings and improving the final audio for everyone to enjoy.  There are many websites, networks and forums which feature some very impressive live Led Zeppelin original recordings, videos and remasters. To the amazement and wonder of most of us new live recordings of the band continue to appear in an unpredictable and staggered manner. 

 

There was an excellent post by Guest WD52 on how fan remastering works on this Led Zeppelin Live Forum.

To highlight a few of specifics of how this works: 

[ from SteveZ98 ]. “Although definitions vary, to me a remaster is just changing the sound characteristics of an existing record (more bass, less treble, adding reverb, etc.), while remixing is changing the layout of the instruments in the stereo field (e.g. moving the guitar to the right channel.) Within the past couple of years, software has been released that can analyze a recording and create an individual track for each instrument. Those tracks are called stems, and once you have them you can remix the instruments into any positions you want…..you're impacting all of the instruments simultaneously, trying to put a final polish on the song. “

[ from TMTOMH ]  “The most common technique is probably EQ, changing the frequency energy with an equalizer plugin or a similar tool. Graphic EQ can boost or cut specific frequency ranges to alter the overall balance of the source, adding high-end sparkle, cutting excessively boomy low end, or boosting or cutting the midrange to change the loudness of much of Plant's vocals and Page's guitar relative to Bonham and Jones. Parametric EQ allows more fine control over the frequency range that's boosted or cut, and is often used to, for example, add some snap to the snare drum without altering the overall sound of the midrange, or to tame an overpowering bass frequency without cutting the overall level of bass on the recording…. folks will sometimes narrow the apparent stereo channel difference by doing a version of what Steve notes above, either narrowing the L-R spread in an audio editor, or else making a duplicate of the stereo source, flipping the channels, lowering the volume of the flipped version, and mixing it back in with the original. …

An exciter adds distortion to the signal, but it adds only even-order (or maybe just 2nd order, I can't recall) harmonic distortion, so what it's designed to do is increase the high end sparkle of a source by synthesizing sounds at double the original frequency of the sound. The restored harmonics extend the treble, brighten and liven up the sound, increase the body and volume (since new sounds are being added all over in the form of those new harmonics), and potentially increase the feeling of air and ambience…”

Material gathered from:  [ https://forums.ledzeppelin.com/topic/27272-how-does-fan-remastering-of-bootlegs-work/ ]. 

 

One possible method for remastering live Led Zeppelin is the combination of soundboard and audience recordings into what is referred to as a Matrix recording. The idea of the Matrix mix method is to combine the best of both worlds - the soundboard feed for clarity and definition, the audience part for vibes and ambience of the entire venue. The soundboard is primarily dominant in the final mix and a tasteful touch of the audience’s sonic perspective adds some nice character to the final product.  Matt Houghton of UK magazine Sound on Sound wrote that the ideal live recording would be “capturing a great performance, complete with all the atmosphere of the night.“ 

Techniques used in remastering include a multitude of methods and software programs that include: band noise reduction, dynamic processing, specific panning, applying reverb, signal processing, equalization, multi band compression, exciters, stereo widening, applying audio filters, and many other procedures. The use of stems is now considered an extremely useful tool in mixing and remastering audio. Music stems are basically an audio item or group mixed down to its own file.  It is a manageable and customizable format used within different music production tasks including audio mixing, audio assembly and final mastering. It can increase the engineers power to construct, deconstruct, remix and customize the audio within their digital workstation. Stems can provide increased flexibility allowing for greater control and creativity.

A note of appreciation and recognition must be given to all the original Led Zeppelin sound crew members who did an excellent job miking the band, mixing the live sound and on occasion, recording the live performances out of the soundboard.  Ultimately the ideal mixed sound is achieved by correcting any mix imbalances and enhancing particular sonic characteristics. AI audio post-production and professional mastering adds an entirely new level of impressive enhancements and offers amazing possibilities to improve recorded sound. 

“Quality is to bootlegs as location is to real estate.“ - Great quote by Comixx : Sat Dec 07, 2013 9:02 pm. Royal Orleans Post

I invite those who have remastered live Led Zeppelin recordings to share how they were able to improve the audio from soundboard or audience captures.  Please feel free to share what DAW [ digital audio workstation ] audio interfaces, signal processors, plug-ins, and software were incorporated. If anyone is currently using any AI Audio Enhancers available in the consumer audio world  [ examples: AudioAI, Audio Audition, iZotope, etc ] I welcome you, if you like, to share your experience with your fellow Led Zeppelin enthusiasts. Details about how the recording was improved from the source material via processing [ compression, eq, reverb, enhancers, etc ] would be most welcome and appreciated. 

 

 

Explanation of Artificial Intelligence enhancement on audio.

AI audio enhancer refers to a technology that utilizes artificial intelligence algorithms to analyze and enhance the quality of sound recordings.By leveraging machine learning techniques, this software can identify and isolate various audio elements, such as vocals, instruments, and background noise, to enhance each component individually.The primary goal of AI audio enhancer is to provide users with the tools to achieve cleaner, more balanced, and polished sound recordings.

An AI audio enhancer employs advanced algorithms to analyze the waveform and spectral characteristics of an audio file. It breaks down the sound into its constituent parts, applying sophisticated signal processing techniques to refine and enhance each element.

This can involve noise reduction, equalization, dynamic range adjustment, stereo widening, and other audio enhancement processes.

Enhanced Sound Clarity: AI audio enhancer can significantly improve the clarity of your recordings by reducing background noise, hums, clicks, and other unwanted artifacts. This ensures that your music sounds polished and professional.

Balanced Audio Elements: With AI audio enhancer, you can achieve a more balanced mix by adjusting the volume levels and frequency response of different audio components. It allows you to highlight vocals, bring out the details in instruments, and create an overall cohesive sound.

Noise Reduction: AI audio enhancer excels in reducing noise in sound recordings, even in challenging environments. Whether it’s hiss, room ambience, or microphone noise, the software can intelligently identify and suppress these unwanted elements, resulting in cleaner and more focused recordings.

[ from Hiphoppush.com article on AI Audio Enhancement ]

 

For instance, AI algorithms can now analyze the spectral content of an audio signal and automatically apply equalization to achieve a desired tonal balance. This can be particularly useful for novice engineers who may not have the experience or knowledge to make these adjustments manually. Similarly, AI-powered tools can identify and eliminate unwanted noise and artifacts from recordings, such as clicks, pops, and hums, which can be a tedious task when done manually.

Another area where AI is making a significant impact is in the realm of sound synthesis and design. Traditionally, creating new sounds and textures required a deep understanding of synthesis techniques and access to expensive hardware synthesizers. However, AI-driven software synthesizers are now capable of generating a vast array of sounds by analyzing and learning from existing audio samples. This opens up a world of possibilities for sound designers and composers, who can now create unique and complex sounds with relative ease.

Moreover, AI is also playing a crucial role in enhancing the quality of audio restoration and remastering. By analyzing the characteristics of the original recording and applying advanced algorithms, AI-powered tools can effectively restore and enhance the audio quality of old and damaged recordings. This has significant implications for the preservation and restoration of historical audio archives, as well as for the remastering of classic albums for modern listening formats.

In conclusion, the integration of AI into the audio engineering industry is undoubtedly revolutionizing the way audio professionals work and the quality of the end product. The automation of various tasks, the enhancement of sound synthesis and design, and the improvement of audio restoration and live sound are just a few examples of the significant impact AI is having on the future of audio engineering. As AI technology continues to advance and become more accessible, it is expected that its influence on the audio engineering industry will only grow, leading to even more exciting developments and innovations in the years to come.

[ The Impact of AI on the Future of Audio Engineering by Marcin Frąckiewicz in Artificial intelligence, TS2 Space / 1 May 2023 ]

 

Spectrograms could visualize signals, but the technique described in the paper—called non-negative matrix factorization—was a way of processing the information. If this new technique worked for video signals, it could work for audio signals too, Clarke thought. “I started looking at how instruments made up a spectrogram,” he says. “I could start to recognize, ‘That’s what a drum looks like, that looks like a vocal, that looks like a bass guitar.’” About a year later, he produced a piece of software that could do a convincing job of breaking apart audio by its frequencies. His first big breakthrough can be heard on the 2016 remaster of the Beatles’ Live at the Hollywood Bowl, the band’s sole official live album. The original LP, released in 1977, is hard to listen to because of the high-pitched shrieks of the crowd.

https://www.wired.com/story/upmixing-audio-recordings-artificial-intelligence/

AI processed sound separation, demixing, remixing and remastering offers the finest possible options when the program is operating with as much information related to the source data as possible. Machine Audio Learning [ MAL ] and Deep Reinforcement Learning [ DRL ] use algorithms that are able to examine a huge amount of source material with supervised learning and solve complex tasks to optimize the objective given from the programmer. A method proven useful to AI audio producers is to convert the audio data into images called spectrograms [ images that portray signal strength and frequencies ] and then process those images to incorporate any new isolated audio into their final mix.  AI will use a variety of algorithms to isolate desired signals and frequencies.  Through detailed data examination, clustering, predictions, separation, specific classifications and a host of other processes these programs can provide previously unobtainable methods of detailed sound isolation and audio production. 

“Spatial audio is an interdisciplinary field of research that brings together experts from audio engineering, acoustics, computer science, applied psychoacoustics, and other domains. The aim of spatial audio is to recreate an acoustic environment or synthesize a new one by using a proper combination of sound recording, processing, and reproduction techniques. Within such an objective, it is not only important to preserve the fidelity of the audio content but also the spatial attributes of the sound scene resulting from the actual locations of the sound sources and the properties of the acoustic environment 

The processing stage may also manipulate the sound scene, for example, by separating direct sound components from diffuse reverberation and recombining them such that this results in a change of the characteristics of the reverberation or in a change of the apparent location of a source. The ultimate goal could be to decompose a sound scene into all its independent conceptual components, i.e., the individual source signals and all components of the reverberation that each source produces. This would allow for unrestricted manipulation. This goal still lies in a considerably distant future so that the available methods rather target different subsets of the sound scene components.”

[ https://asmp-eurasipjournals.springeropen.com/articles/10.1186/s13636-022-00242-x ]

 

A great basic overview of Demix Technology as presented by Abbey Roads Studios, London UK

 

Consultation: Detailed analysis of your brief against the existing audio, to access feasibility and the best De-mixing method.

Forensic Research: into the best possible source material, thanks to our in-house archiving expertise and extensive network.

Audio Separation: Designing ‘audio source models’ to enable De-mixing and the audio separation process.

Remix / Remaster: Once separation is complete, our engineers can begin to remix or remaster the original base recording, perhaps to stereo or 5.1.

Review: Client review and approval which allow us to revisit the process and finesse specific details.

 

It would seem that Jimmy Page has a residence in London which is about 4 miles from Abbey Road Studios. 15 minutes in a car, 1 hours 11 minutes walking, and 23 minutes by bicycle… [ thanks Google Maps ! ]   🤔

 

 

Led Zeppelin Live Recordings.

“I had a dream, crazy dream….”  It would seem at this point in time, many years since the ‘official end’ of Led Zeppelin, that the architect of the band Jimmy Page or anyone at Led Zeppelin HQ will not release any more live official Led Zeppelin concert recordings. [ I hope that is not the case. ] Jimmy Page has mentioned several times in different interviews that he has a collection of live recordings from the band from 1968 to 1980 and suggested he would compile yearly or tour specific releases to highlight certain performances. Unfortunately this has not come to fruition.  Everyone was most grateful and appreciative of the early BBC Sessions, official Led Zeppelin DVD Box Set, How The West Was Won recordings, the remasters of the soundtrack from The Song Remains The Same and the Celebration Day capture of the 2007 London O2 reunion show.  

Alas it might come down to "Do what thou wilt—then do nothing else." [ A.C.]

Theoretically Jimmy Page would lead such a remastering project and would oversee a professional team who could utilize the absolute best studios and equipment to improve existing raw recordings and showcase the band at their best live.  There might even be material in his ‘vaults’ or in the hands of other individuals that has yet to come to light. One can speculate that the Led Zeppelin estate has the finances to do a whole host of projects if they so desired. If Robert Plant and John Paul Jones want to be involved at any level of any potential new releases, I am sure all would welcome them with love and appreciation.  Jason Bonham is known to have an encyclopedic knowledge of all things Led Zeppelin and is a dedicated collector of live Led Zeppelin recordings so I would speculate he would love to be involved.  [ nota bene: Jason was absolutely amazing drumming for the 2007 O2 London, UK show. Brilliant! ] Given the success that Peter Jackson and his team of audio specialists at Park Road Post Production in New Zealand and the staff at Abbey Roads Studios in London, UK have had in improving and remastering The Beatles material using AI, one could hope that Jimmy Page and the Led Zeppelin Estate could use comparable methods to improve existing live concert recordings and share this wonderful music with the world. [ there is already a little connection of Led Zeppelin’s love of The Hobbit / The Lord of The Rings and Peter Jackson who did an amazing job bringing the JRR Tolkien novels to the big screen “ … and in the darkest depths of Modor. “]  I am confident that there are many talented audio engineers and Led Zeppelin enthusiasts who be very enthusiastic and honored to be involved in any new live Led Zeppelin recordings.

 

“I seem to have tireless energy when I get involved in things, on an almost OCD basis, which is a good way to do things because if you're gonna do something, you'd better make sure you do it well.” - Jimmy Page

“I knew it was a long haul, that it would involve hundreds of hours of tape,” he said in an interview in New York on Wednesday. “I had to listen to everything, every bootleg that was out there, too. But it has to be done if you’re going to do something really authoritative. I wanted to be sure this holds up, and I hate to think, if I wasn’t around, what was going to happen.”   - Jimmy Page [ NY Times 05.17.14. Article/Larry Rohter ]

 

There are also several individuals who have worked on the audio remastering of the BBC / DVD / How The West Was Won / The Song Remains The Same live audio and would be potential candidates to be involved in any new projects.  Here is an incomplete list of individuals known to have worked on [ primarily live ] audio for Led Zeppelin and/or Jimmy Page :

Benjamin 'Benji' Lefevre - Led Zeppelin engineer and soundman.

Kevin Shirley - Led Zeppelin DVD engineer.

Tim Summerhayes - Led Zeppelin 2007 O2 engineer.

John Catlin - Celebration Day release engineer.

Richard Lancaster - Celebration Day release Pro Tools engineer.

Roy Williams - 2007 02 concert / Robert Plant sound.

MIck Hughes - 2007 O2 concert.

Barry Diament - 2007 TSRTS soundtrack release mastering.

Alan Moulder - Celebration Day release engineer / Pro Tools engineer.

Eddie Kramer - Five time engineer for Led Zeppelin.

Phil Duderidge - Led Zeppelin's first dedicated live sound engineer.

Bob Ludwig - Led Zeppelin II album mastering engineer.

Stuart Epps - Led Zeppelin Box Set engineer.

Pete Ritzema - BBC Sessions release.

Tony Wilson - BBC Sessions release.

John Astley - BBC Sessions mastering.

 

Abbey Road Studios in London, UK and Park Road Post Production in Wellington, NZ are possible locations and candidates for any future remastering Led Zeppelin projects. The Real World Studios, in Box, Wiltshire, UK founded by Peter Gabriel is a greatly respected recording facility with amazing studios and has a variety of dedicated post-production and mixing rooms. The aforementioned studios and post-production audio labs are some of the finest in the world and have an excellent reputation in providing impressive professional sound restoration.

 

 

Possible recordings to remaster using AI from the Led Zeppelin Archives:

1969: Supershow. UK Television.

1969: Denmark TV Byen, Danmarks Radio

1970: Royal Albert Hall - Vic Maile - Pye Mobile Unit - 8-track one-inch tape

1970: Bath Festival - [ audio recording method unknown ]

1971.04.01 -- Paris Theatre, London, England, UK -- Concert recorded for John Peel's In Concert series; officially released in 1997 on BBC Sessions.

1971: Japan - ??? - Warner-Pioneer Recording Unit? - 2 8-tracks two-inch tape?

1972: LA Forum / Long Beach Arena - Eddie Kramer - Wally Heider Mobile Studio - 16-track two-inch tape

1973: Southampton - Ron Nevison / Eddie Kramer - Rolling Stones Mobile Studio - 16-track two-inch tape ???

1973: Madison Square Garden - Eddie Kramer - Wally Heider Mobile Studio - 16-track two-inch tape

1975: Earl's Court - George Chkiantz - Rolling Stones Mobile Studio - 16-track two-inch tape

1979: Knebworth - George Chkiantz - Rolling Stones Mobile Studio - 24-track two-inch tape

Thanks  to [ RoyalOrleans Post ] thenobs70 » Fri Jul 13, 2018 / known Led Zeppelin live multi-track recordings.

Note 1: Roger Daltrey was at the Royal Albert Hall gig that Vic Maile recorded. Apparently The Who wanted some of Zeppelin's magic so they utilized his skills for their Leeds gig in February. We all know how that turned out. [Actually they used their soundman Bob Pridden. A mistake on my part. But they did use the Pye Mobile Unit.]

 

 

Potential Future Led Zeppelin shows for AI remastering believed to be multi-tracked.

·1971.09.28 -- Festival Hall, Osaka, Japan -- This concert is widely believed to have been multi-tracked.
·1971.09.29 -- Festival Hall, Osaka, Japan -- This concert is widely believed (and very likely) to have been multi-tracked.

Informational link about the Japan shows:  http://www.royal-orleans.com/phpBB/viewtopic.php?f=21&t=26340&start=50

10/17/69 Carnegie Hall shows in NYC are confirmed as being multi-tracked through Eddie Kramer.

6/29/29 RAH shows are rumored to be multi-tracked.
9/23 and 9/24 71 Tokyo confirmed mulit-racked on 2 8 track machines using 2 inch tape.
9/28 and 9/29 Osaka are confirmed also, same equipment as Tokyo.
Japan 71 recordings are confirmed here from an interview in Crossbeat magazine.

http://www.royal-orleans.com/phpBB/view ... 0&start=50

Jimmy Page himself or Peter Grant confirmed one of the Surrey University shows in 68 was multi-tracked. I don't recall if the date was 10/15/68 or 10/25/68 since they played shows at Surrey University on both dates. [ Thanks  Sathington Willoughby. August 2, 2015 ]

According to the latest information available there are at least 98 Soundboard recordings of Led Zeppelin circulating and hundreds of audience captures.

 

Thank You

Once again, many thanks to all those on the Led Zeppelin Forum and everyone who loves the band and continues maintaining their musical legacy.  The main goal of improving any existing concert audio would be to provide the ultimate representation of Led Zeppelin’s impressive live performances.  Excellent high fidelity and upgraded audio will allow the listener to fully appreciate the devastating power, subtle nuances and rich complexity of the band’s musicianship. Thanks to John Bonham, John Paul Jones, Jimmy Page, Robert Plant, Jason Bonham and Mr. Peter Grant. 

 

“ Best. Band. Ever. “ - Jack Black [ 2012 Kennedy Center induction of Led Zeppelin. ]

 

PS: I love this video. A Japanese trio called NoMorePanda delivering some fine Led Zeppelin music.  Check them out playing ‘Achille’s Last Stand.’  Brilliant!

https://www.youtube.com/watch?v=LH0cwQaa8xs 

 

 

Here are some related articles for those who wish to further explore the history of live Led Zeppelin recordings :

Audio technology article on 2007 Zeppelin

https://www.audiotechnology.com/PDF/FEATURES/AT59_LED_ZEPPELIN_LIVE.pdf

Kevin Shirley - mixing live DVD

https://www.soundonsound.com/people/kevin-caveman-shirley

MixOnline article on audio for DVD

https://web.archive.org/web/20080706200630/http://mixonline.com/recording/interviews/audio_led_zeppelin/index.html

Article on recording Zeppelin 2007 - Big Mick Luges

[https://www.audiotechnology.com/features/led-zeppelin-live]

Website for Park Road Post Production studios in New Zealand.

https://www.parkroad.co.nz/

“About those 1971 - 1980 soundboards”  by blackmikito » Sun Jul 12, 2020 2:22 am Royal Orleans - [ I love this post! The attention to detail!  Blessings to ‘blackmikito’ for all of this historic information related to early recording methods, microphone placement and his amazing photographs. Brilliant! ]

https://www.royal-orleans.com/phpBB/viewtopic.php?f=3&t=46925

Mixing zeppelin live in 1977 article

https://worldradiohistory.com/hd2/IDX-Audio/Archive-Modern-Recording-IDX/IDX/70s/Modern-Recording-1978-02-OCR-Page-0054.pdf

Phil Dudderidge - 1st. Sound eng. working with zeppelin

https://www.listennotes.com/podcasts/a-history-of-live/pt-13-phil-dudderidge-led-atSh1xJK0Pu/

Phil Dudderidge - on how 10 channels were used to record early Led Zeppelin [ to put this in a historic perspective, over 130 tracks were used ultimately in the final mastering of the Celebration Day / 2007 O2 concert audio ]

[https://soundgirls.org/phil-dudderidge-on-the-road-with-led-zeppelin-the-first-time-around/]

Informative article on AI and how spectrograms are used in modern audio production. “Audio Deep Learning Made Simple.”

https://towardsdatascience.com/audio-deep-learning-made-simple-part-1-state-of-the-art-techniques-da1d3dff2504 

Excellent Youtube.com video on how AI audio demixing and remixing works: Music source separation YT Video by Stipe Kabic

https://www.youtube.com/watch?v=DnUcj0B4BWA

Detailed information on the history of live ‘alternative’ Led Zeppelin recordings.

http://www.argenteumastrum.com/bootlegs.htm

 

“Light and shade... Whisper to the thunder. It invites you in... It's intoxicating." - Jimmy Page

[SteveZ98]

That was quite an informative post. It's already possible to do much of what you described. Turning one of Zep's mono BBC recordings into a stereo recording is easy using AI, or rather the AI can easily split the songs into individual tracks (bass, drums, and vocals become individual instruments, and whatever remains is all put together in a single track) and those can then be manually mixed like you would a four track recording, and then mastered as desired. There is AI that will automate mixing. It does a great job on studio songs, but isn't as useful on live recordings where the goal is to present all of the instruments on an equal footing. And there is also AI that will try to apply the best EQ and other effects to to make music sound as good as possible. The downside, from what I've seen/heard, is that all AI works best on music that was well recorded to begin with, which most bootlegs are not.

The two main types of bootlegs, audience recordings and soundboards, each present different challenges and opportunities. I've been able to convert a number of soundboards to stereo and remaster them to varying degrees of listenability, but they are fairly easy to work with because most of them are at least decent sounding and consistent (i.e. no unexpected volume changes, etc.) before they're remastered. I think what a lot of people are hoping is that AI can take something like the awful Bath 1970 audience recording and turn it into an end product that sounds as good as How The West Was Won. I'm not sure if that will ever be possible. I can see a time when it is, but there would need to be considerable effort put into training the AI and I don't know who will do that. If bands like Zep think they can use AI to do that and are willing to pursue it because there's money to be made, it may happen. If it's just left up to fans, I'm not sure how many people have the technical knowledge and time train AI.

Additionally, there are a few things AI currently struggles with when trying to parse some Zep songs that may remain challenging. In particular, the vocal and guitar duets that Robert and Jimmy engaged in during 1969 are difficult for AI to separate adequately. It generally just puts both those sounds in the vocal track. Also, most AI has been trained on acoustic pianos, so they struggle at times when JPJ uses an electric piano or other type of keyboard. It should be technically easy to resolve that by training the AI on the type of keyboard he's using during a given show, but that training takes time and a lot of samples of the individual instruments and the complete songs to be done correctly. The other issue is when JPJ and Jimmy are both playing guitars/mandolins/three neck acoustical things. AI cannot currently separate those instruments. It's possible to create a pseudo-stereo spread of them based on AI putting them all in one stem, but it's not the same as having each instrument in its own track.

After all that, hearing is worth a thousand words. Here's an example from something I'm currently working on. Having AI parse the songs into individual tracks took a couple of hours, and in this case it did a particularly good job separating the piano from the rest of the music.. Mixing the tracks by hand took another couple of hours. Making the resulting stereo songs sound as good as possible will take a couple of months, if not more, although I work very slowly so someone else might be able to finish a remaster like this much more quickly.

 

[NorthShoreBlues]

Damn. That sounds good

[john morrison]

being a tech luddite much of the discussion passed over my head ... but I can appreciate the product . I doff my cap to you in recognition of your efforts 

[John M]

15 hours ago, SteveZ98 said:

Here's an example from something I'm currently working on. Having AI parse the songs into individual tracks took a couple of hours, and in this case it did a particularly good job separating the piano from the rest of the music.

That is outstanding.  

[donb8]

Great post! Great follow up as well. On audience recordings, the goal should just be improving what we have. Especially the historical shows. I digress though. It will be difficult since these recordings are of varying quality. Baltimore 1972 and 1973 are examples of this.

[SteveZ98]

3 hours ago, John M said:

That is outstanding.  

Thanks. The piano wanders from left to right channel and back early on in that sample. It's easy to fix, but I left it in as an example of the kinds of things that need to be handled manually. There's other stuff like that that needs to be cleaned up by hand in a lot of the songs because the AI is really good, but it's not (yet) perfect. The ones where it does the best are straight forward rock songs like Sick Again and The Song Remains The Same. Interestingly, the AI doesn't know what a guitar sounds like. Instead, it puts the guitar and anything else it doesn't recognize in an "Other" track. The reason those songs work so well is that there's nothing else aside from the guitar in that track.

[rm2551]

One day, we will have releases (either official or otherwise) that utilise these kinds of technologies. The latter Zep live years are very under represented. LA 1977 soundboard is a kind of grail for enhancing potential (especially if the SB is well balanced, which is a BIG 'if' for '77 boards). Best case scenario is Jimmy has the boards and they are decent, and someone convinces him that they can be turned into a near immaculate release, and he decides to go ahead with a project. This scenario is sadly unlikely. What is much more likely is at some point good to very good sounding boards and audience recordings are turned into incredibly good recordings. Unfortunately for us, this is likely years, possibly decades away.

It might all just be wishful thinking too.

[Inthelateafternoon]

2/28/75- Something to look forward to.

[Dirty Work]

AI really should be able to fix whatever is wrong with the 5/23/75 multitracks. I think the things that would benefit the most from demixing are 4/1/71 BBC, and Danmarks Radio, as well as 9/23/71, 9/24/71, 9/28/71, and 9/29/71.

PS. For what it is worth, I do not think Carnegie ‘69 and Albert Hall ’69 were recorded. Those are simply rumors. Bath ‘70, however, is more than a rumor.

-things I cross my fingers for: Winterland late 1969 shows, various 1969 gigs, especially the Fillmore’s/Bill Graham venues, anything from Europe in 1970 or 1971, Montreux 1971 and 1972. 1972 US tour soundboards. 1972 Down Under soundboards. 3/12/75 soundboard. Anything 1977 soundboard on a night where the band is playing well. 9/19/70 soundboard (hell, anything from the year 1970).

Edited November 7, 2023 by Dirty Work
Added 9/19/70 to my wishlist

[BShea]

I was just thinking about how it would be great to have AI “fill in” those little glitches in some recordings. Spots of 1 or 2 seconds of silence, can AI simply create filler to span the missing second?

Also, having AI learn “bootleg ears” the way we all learned to hear the music in super muddy recordings.

I listen to some of my old favorites after listening to soundboards and I wonder how I was pleased with it. Our brains seemed to filter out the static, pops, and mud to allow us to enjoy the performance. Then when you try to share it with a friend that doesn’t have bootleg ears, you see them cringe.

I got hopes. Crap, just the availability of so many shows on YouTube is blowing my mind. We used to have to hunt for tapes and CDs in strange places. Then we had to find downloads and hope people knew how to rip a CD properly.  Now we have YouTube.