Bud Powell behind Sarah Vaughan
Bud Powell recorded with Sarah Vaughan twice. In May 1946 he sat in the rhythm section of the Tadd Dameron Orchestra for four Musicraft sides, alongside Freddie Webster and Kenny Clarke, with nine strings behind them. In July 1950 he turned up again on both sides of a Columbia single under Norman Leyden. Seven titles in all, counting the alternate take of My Kinda Love that the discography lists and that does not appear to survive.
These were the sides missing from the Bud Powell sideman post, and the best transfer of each now plays there alongside the rest of his sideman years. This post is the sweep behind that pick. Same problem as the Morton records: a 78 has no master tape, so what you hear depends on which pressing got transferred and how fast it ran. So for each tune I collected every transfer I could find, the pressings on the shelf and the CD issues alongside collector uploads and 78 channels online. That came to 55 transfers across the six tunes, between eight and eleven per tune.
Vaughan kept singing most of these titles for the next thirty years, so a matching title proves nothing. Every candidate was checked against a known copy of the 1946 or 1950 master by comparing chroma across all twelve transpositions, which is what separates a different pressing from a different recording. That threw out a lot of correctly titled wrong records. A second pass compared the survivors to each other on a log-mel fingerprint, which catches something chroma cannot: the same digital master re-posted under another channel name. True re-posts match at 0.999 while genuinely different masterings of one take spread between 0.78 and 0.98. Six re-posts came out that way.
Every transfer then went through the same pipeline. Source separation pulls the piano out, a tuning pass corrects the transfer speed varispeed-style so pitch and tempo move together, and a transcription model reads the corrected piano stem. Corrections here ran from −34 to +42 cents.
Vaughan then comes off the band stem onto a fader of her own. She is the record, mixed forward the way a 1946 vocal side is, and with one band fader she sits on top of everything behind her. Split out, what is left is Dameron's writing: Freddie Webster's trumpet, Leo Parker's baritone under the strings, and Powell comping inside an arrangement rather than under a singer. She starts at −8 dB so the arrangement is what you hear first. Pull her back up to hear the record as issued.
On the best transfer of each tune the orchestra is peeled apart further: bass, then drums, then guitar, then strings, then winds, each pass fed what the one before it left. The guitar pass only earns a fader on the 1950 sides, where Mundell Lowe comes back at 76 and 81 percent continuity. On the 1946 date it returns silence on three tunes and eight percent of scattered transients on the fourth, which is the answer you want: Dameron's personnel has no guitarist, so a model that found one there would be inventing it. Those six get a fader per section, named for the players where the personnel is known. The other forty-nine keep the three-fader set, since they are there to compare pressings and nobody A/Bs a string section across nine of them.
The drums mostly do not survive that. Kenny Clarke comes back at −85 dB on My Kinda Love and lower still on I Can Make You Love Me, which is silence rather than a quiet stem, so those tunes get no drum fader. He is playing brushes on ballads under nine strings, and brushes are the one thing these models reliably cannot find. Only If You Could See Me Now holds a usable drum stem from 1946. Terry Snyder survives on the 1950 side, where he is playing with sticks.
One transfer is worth hearing for what a worn pressing does to all of this. On I Can Make You Love Me, the copy marked Musicraft 78 (copy B) scores 54.7 percent and yields 202 notes, where the other seven sit between 77 and 87 percent and give 875 notes or more. Chroma confirms it is the same recording, so that is groove wear the separator could not read, not a different take. The pressing filed here as copy A is labelled I Could Make You Love Me, which turns out to be the real Musicraft 396 issue title rather than a mistake, and it matches the Powell side and not the 1945 song of that name.
The two 1950 Columbia sides sit ten to twenty points below the 1946 ones throughout. That is Powell inside a full studio orchestra rather than a separation failure.
Every transfer, ranked
Continuity is the fraction of half-second windows above −50 dB in the piano stem, so higher means fewer holes; notes is how many notes the transcription model heard. The waveform shows where the gaps fall. Neither number settles anything, they just say where to listen first. Click a name to load that transfer in the mixer.