Home / Lab / Vocal space test

Result in this run

Only A4 clearly delivered the intended distant, reverberant vocal. B1, B2, and B4 clearly delivered close, mostly dry vocals; B3 stayed dry but sounded somewhat farther back. This is a repeated field comparison, not a seed-matched controlled experiment.

Listen to the closest musical match

Start close, then move farther back

B1 and A4 were the closest pair in singer character, melody, tempo, and arrangement. Start with B1, then switch to A4. The two links begin at corresponding listening moments rather than the same clock time.

B1 · close and mostly dry

Creator observation: little obvious reverb; lead vocal sounds close. Start at 1:13.

Cover artwork for The Last Word B1 Close Dry
Use the official player, or open B1 at 1:13.
A4 · roomy and distant

Creator observation: more obvious reverb; lead vocal sounds farther away. Start at 1:17.

Cover artwork for The Last Word A4 Roomy
Use the official player, or open A4 at 1:17.

What stayed stable

The generation plan used the shared title The Last Word; the public records were then labeled A1–A4 and B1–B4 for review. All eight used the same lyrics, the same v5.5 model, and the same musical core: an intimate indie-pop ballad at 78 BPM in a minor key, restrained verses, a gently lifted chorus, warm female alto lead, sparse electric piano, soft bass, brushed drums, and a faint analog pad. Both conditions also asked for relatively dry, restrained instruments so the vocal-space request remained the intended variable.

A conditionLead farther back in a large studio room; audible early reflections, spacious ambience, and a long smooth reverb tail.B conditionLead close to the microphone and forward; minimal reflections, almost no ambience, and a very short subtle reverb tail.RepeatsFour generations per condition; every record was retained rather than selecting only successful outputs.Listening focusPerceived near-to-far distance, audible reverb amount, and phrase-ending tail.

All eight generation records

Retain the misses as part of the result

RecordRequested conditionCreator listening observationClassification
A1Roomy / distantNo obvious reverb; vocal sounds close.Direction not clearly followed
A2Roomy / distantNo obvious reverb; vocal sounds close.Direction not clearly followed
A3Roomy / distantNo obvious reverb; vocal sounds close.Direction not clearly followed
A4 · 1:17Roomy / distantReverb is more obvious; vocal sounds farther away.Clear condition match
B1 · 1:13Close / mostly dryNo obvious reverb; vocal sounds close.Clear condition match
B2Close / mostly dryNo obvious reverb; vocal sounds close.Clear condition match
B3Close / mostly dryNo obvious reverb; vocal sounds somewhat farther away.Partial condition match
B4Close / mostly dryNo obvious reverb; vocal sounds close.Clear condition match

What the result supports

The main pair supports a concrete listening observation: B1 sounds closer and drier, while A4 sounds farther back with more obvious ambience. The full set supports a second observation about consistency in this run: the close/dry request transferred more often than the roomy/distant request.

Evidence boundary: these generations were not produced from a shared fixed seed, and musical details can vary between renders even when the inputs are held stable. The records show what happened in this eight-generation run. They do not prove that either phrase caused every difference or that another session will produce the same match rate.

Back to the evidence collection: Browse all Lab studies