Voice belt: two replacements for the dropped third voice

Both of these were in the 45 creators I measured yesterday, and I rejected both for a weak consonant gap: 14.2 dB and 12.4 dB against a 15 dB bar. That measurement was taken on the same 16 kHz files that caused the robotic clones, and a 16 kHz file has already thrown away half the consonant band the measurement reads. Re-screened at 24 kHz they come back at 23.3 dB and 22.1 dB. The bad sample rate was rejecting good voices as well as producing bad clones.

Both are deeper than anything in the belt except Jack Henry's creator, both read clean on all three takes, and both sit inside the voiced-share band every voice you approved occupies. Same two sentences of the Rainbow Passage as the others.

m30-theophilus — Theophilus Aladejebi, fragrance and style reviewer ← my pick

clone pitch 98.2–99.4 Hz · sample 12,000 Hz, clarity 22.1 dB · Whisper clean on all three takes · voiced share 0.26–0.33

The deeper of the two, and he lands his line endings hardest: his phrase fall reads 3.79 to 6.25 semitones, the widest of any voice in the belt. His source audio also had the widest pitch range of all 45 creators I measured at 8.03 semitones.

m30-huntervought — Hunter Vought, cologne and style reviewer

clone pitch 99.7–106.7 Hz · sample 12,000 Hz, clarity 23.3 dB · Whisper clean on all three takes · voiced share 0.13–0.17

The steadier read of the two, and the closest in texture to Wolf of Williamsburg, who you approved. Calmer and more even than Theophilus, which suits a reviewer walking through a product rather than selling hard.

Against the rest of the slot

These two bring the man 30-45 slot to six voices. The other four for comparison:

m30-aterencehughes — 87.7–94.7 Hz, re-cloned this morning, awaiting your verdict

m30-balancegrooming — 91.4–93.8 Hz, re-cloned this morning, awaiting your verdict

m55-drgeorge — 110–128 Hz, your first pick for authority reads

m30-mattytingles — dropped this morning on your verdict, here so you can hear what the new two are replacing

Every number

voice	pitch	voiced share	phrase fall	seconds	sample	Whisper
m30-theophilus	98.2–99.4 Hz	0.26–0.33	3.79–6.25 st	7.5–7.9	12,000 Hz / 22.1 dB	clean ×3
m30-huntervought	99.7–106.7 Hz	0.13–0.17	1.78–4.33 st	8.0–8.5	12,000 Hz / 23.3 dB	clean ×3
m30-aterencehughes	87.7–94.7 Hz	0.22–0.23	1.21–4.58 st	8.2–8.6	12,000 Hz / 24.4 dB	1 word out in 3
m30-balancegrooming	91.4–93.8 Hz	0.19–0.24	4.00–4.63 st	7.8–8.6	12,000 Hz / 20.4 dB	2 words out in 3
m30-wolfwilliamsburg (approved)	120–126 Hz	0.11–0.13	3.44–4.16 st	8.1	10,851 Hz / 18.9 dB	clean
m55-drgeorge (approved)	110–128 Hz	0.15–0.20	−3.00–7.25 st	8.2–8.3	12,000 Hz / 28.6 dB	2 of 3 misread a word
m30-mattytingles (dropped)	92–107 Hz	0.58–0.60	3.72–4.87 st	9.0–9.7	12,000 Hz / 25.7 dB	clean

What the screen now checks, and what it cannot

The hunt screens a source window on four numbers before anything is cloned: consonant gap at or above 15 dB, bandwidth at or above 10,000 Hz, pitch at or below 130 Hz for this slot, and a voiced share inside 0.05 to 0.45. Nine deep candidates went through it and four cleared every bar.

The voiced share is the number that condemned Matty Tingles, and the screen cannot catch it in advance. I checked: his source window measures 0.313, which is ordinary, and Dr Scott Walter measures 0.39 at source and is approved. The clone took 0.313 to 0.58. The defect is created by the cloning, so that number can only be read after the takes exist, which is where I now read it. Both voices above were checked there and both passed.

Spend on this hunt was six takes at $0.033. The downloads and both clone registrations were free. The Out There ad batch remains stopped where you left it.