Hello,
The release of a new song in particular for a group as well-known as the Beatles is always an event. "
Now and then" was composed by John Lennon in 1978, and finalized by Paul McCartney and Ringo Starr in 2023.
View attachment 323751
All this was made possible by the AI used to produce the track.
So the question arises as to what artificial intelligence was used, and for what purpose, in the production of this new track.
Generative AI is used to create content, and is increasingly used in all fields, from text to images to music.
But in the case of this piece, it wasn't a generative AI that was used, so there was no creation of new content or new paths in relation to a text, for example.
It's a tool using AI to optimize the separation of a track already mixed on a demo tape, as in this case John Lennon's voice and piano, to produce different tracks with restored quality.
So what's the result when you listen to this song? That's what we're going to look at with the 5 versions tested: maxi45 vinyl record, blue vinyl record, Tidal MAX flac in 24-bit 96 kHz, Amazon UltraHD in 24-bit 96 kHz and Tidal Dolby Atmos.
The waveform of the two vinyl records is not very dynamic. But we can see that the level of the maxi45, which is on a 12″ record, is almost 3 db higher than that of the single vinyl record, which is on a 7″ record.
This is explained by the fact that there is more useful surface area for recording information on the 12-inch record than on the 7-inch record, allowing a wider track cut (and therefore a higher level), as shown in the photo comparing the tracks on the 2 vinyl records.
View attachment 323753
vinyl records : single 7" (blue) vs maxi 12" (black)
The graph below shows the waveforms of the Tidal Max version, the maxi vinyl disc and the Dolby Atmos version downmixed to 2.0 for the song "Now and Then".
All three waveforms are adjusted to -14.7 LUFS for comparison purposes.
The most dynamic version is the Dolby Atmos version downmixed to 2.0. As for the vinyl version, it appears to have been burned directly from the dynamically compressed stereo version.
The difference in DR between the Tidal Max version and the vinyl disc is due to the fact that the vinyl was burned from a source that had undergone a dynamic limiter (see explanation "
Do analog media impose dynamics on music?").
View attachment 323754
Waveform “Now And Then” Digital stereo vs vinyl record vs Dolby Atmos
The graph below compares the spectrum of the Tidal Max version (white curve) with the spectrum of the Dolby Atmos version downmixed to 2.0 (blue curve). The two curves are very different, due to the use of two different mixes. The yellow arrow indicates that the maximum frequency of the Tidal Max version is 24 kHz (for the "Now And Then" sound), despite the use of a 96 kHz sampling frequency.
This raises the question of whether it has been mixed at 48 kHz.
For the Dolby Atmos version, the 20 kHz limitation is due to the Dolby Digital Plus encoding.
View attachment 323756
Spectrum Tidal Max Flac – 2023 (white) vs Tidal Dolby Atmos – 2023 (blue)
Spatialization is perfectly realized while respecting the original spirit. The voice remains in the foreground, with great clarity and nuance, while the music and backing vocals make full use of all channels.
View attachment 323757
Digital technology is capable of both the best and the worst. Let's start with the best: AI was used to separate the voice from the piano with fantastic results, not the AI-generated voice. Unfortunately, the dynamic range compression enabled by digital technology tarnishes the result on stereo editions.
Fortunately, Dolby Atmos is there to save the day, offering a version that respects the dynamics of the mix, while rendering the voice in a cleaner, more natural way.
To convince yourself of this,
simply listen to and compare the various extracts here, including the Atmos 2.0 (stereo) downmix.
Enjoy listening
Jean-François