← The Corpus·The Source Voice
Tragedy Academy Studios
A synthetic voice cannot exceed its source.
There is a voice model of Jay Hicks in the Respeecher marketplace. These seven films are the instrument it was made from — seven genres, six characters, one performer, and the full breadth of what is actually in there.
Why this page exists
These are not add-ons. They are what the voice is made of.
None of this material is in the corpus. No character work was ever recorded for training, and none of it needed to be. A voice carries the instrument that produced it — the elasticity, the reach, the places it can go under load — and every hour of conversational audio in the archive came out of the same instrument that does all six of these.
That breadth doesn't get added to a voice later. It is either in the source or it isn't, and it colours everything the source ever says — including an ordinary sentence with no character in it at all.
This is that instrument at full extension. Six characters, six genres, one performer. Start at the top — the first one is the furthest thing from a podcast voice on this page.
Each film is a single performance, not a compilation. No line-by-line assembly, no comping between takes.
01 — Supper Time
Elias Rook · 0:53 · Survival horror · Photoreal
A cattle farmer holds a family supper that has gone very wrong. Constricted throat placement with heavy vocal fry, authentic southwest regional vowel shifts, harsh glottal attacks, ragged unfiltered inhalations driving the tempo.
He is at peak volume for forty-seven straight seconds. Then the last word drops twenty-five decibels to a whisper. The furthest point in the reel from the commercial voice.
02 — Fish & Chip
Chip · 1:06 · Deadpan comedy · 2D animation
A french fry tows a goldfish around a county fair planning a heist. High nasal mix with tight laryngeal constriction, a twelve-decibel dynamic window, breath completely suppressed, diction overly formal on purpose.
Never raises his voice once. The hardest register to sustain, because nothing is allowed to help you.
03 — Unpunished!
“Broken Leg” · 1:17 · Theatrical tragedy · Dark fantasy / narrative
A ruined performer in a collapsed opera house, named for “break a leg.” Roughly thirty-five decibels between murmur and scream. Deep wounded chest pushing into high pharyngeal strain, theatrical gasps as punctuation, articulation holding crisp under distortion.
At 0:54 the theatrical grandeur drops away and the hatred becomes personal. He tears the mask off at the end and bows bare-faced to an empty house.
04 — Three Minutes
Dr. Anselm Vosk · 0:39 · Manic · Steampunk / fantasy
An inventor in a lantern-lit workshop, three minutes from something. Extended upper pitch ceiling with head-voice glides, octave jumps on emotional peaks, explosive laughter, articulation staying hyper-precise through extreme tempo shifts.
At 0:22 the explanation breaks into uncontrolled laughter mid-thought. That transition is the whole performance.
05 — Divided & Dying
0:25 · Epic narration · Fantasy / trailer
An unseen narrator over a fallen empire in the rain. Low chest and sub-glottal resonance, compressed dynamics, elongated vowels, deep controlled inhalations carrying the weight.
No character on screen at all. The voice has to build the world by itself.
06 — Onboarding
Professor Gantt Flotsam · 0:28 · Warm institutional · Animated family / education
A professor welcomes a corridor of schoolchildren using the language of a corporate onboarding deck. Mid-high placement with a brightened smile vector, narrow dynamic range, hidden breath, pitch lifts at phrase ends.
The comedy is entirely in the mismatch — growth mindset and agility, delivered to eight-year-olds with total sincerity.
The zero point
And here is the same voice being ordinary.
Range only means something against a reference. This is the professional instrument doing paid, unremarkable, competent work — five brands, five stances, no character. Deep chest for premium automotive, warm mid-chest for financial advisory, light nasal conversational for consumer, gritty chest-throat for apparel, softened attack for healthcare.
Commercial reel · 1:18 · Everything above is measured as distance travelled from this.
What holds it together
One instrument, six configurations.
Independent acoustic review of the full set identified the through-line under every character: over-articulated consonant termination — crisp stops, hard T and K — and rhythmically structured breath intakes placed before emotional pivots rather than after them.
That signature survives vocal fry, head voice, nasal constriction and full theatrical strain. It is what makes the range legible as range rather than as six different actors.
The characters live in the voice.
White light reads as one plain thing until it meets a prism. The glass adds nothing — the whole spectrum was in the beam the entire time.
Break the ordinary conversational audio down and the same material is already sitting in it. The same over-articulated consonant terminations, the hard T and K. The same breath arriving before an emotional pivot rather than after it. The same registers moving against each other inside a single unbroken take. These films do not add range to the speaker. They split it.
The same voice has been licensed as a synthetic model built with Respeecher's Emmy-awarded team, trained on a curated corpus of unscripted long-form conversation.
That corpus contains no character work at all — no animation scripts, no healthcare copy, no game dialogue. It is one person talking. The breadth on this page was never trained in; it was in the speaker the whole time, which is why it is present in ordinary conversational audio and audible in what that audio produces.
The masks
Same man. Same microphone. No performance at all.
Everything above this line was written, directed and performed. These two were not. Same room, same broadcast chain, same afternoon — two behavioural registers that were simply happening while the recorder was running.
Nothing was tightened afterward. No filler removal, no pause trimming, no Studio Sound. The picture is cut; the audio is untouched. The dynamic range is the data.
Comic
Detention, live · four people in the room · 1:40
A story nobody planned, in a room where the reactions are half the comedy. The audio runs unbroken from first frame to last — every hesitation, restart and overlap left exactly where it landed.
Teacher → Vulnerable
Long-form interview · one take, no notes · 1:01
A guest offers two words — stained glass. They get taken and built into a full image, live, across a minute of unbroken speech. The metaphor was not prepared. It was assembled while it was being said.
A model trained on one of these cannot produce the other. Not because the voice is different — it is the same voice, the same chain, the same afternoon. Because the behaviour is different, and behaviour is the part that never gets recorded.
Blackmagic cinema · multi-angle · broadcast microphones · picture edited, audio untouched
The range was always in the voice.
Every character on this page was written, performed, directed and finished in-house — six worlds built to find the edges of one instrument. Nothing here was recorded for a model. It is simply what the voice contains.
The synthetic voice is licensed through Respeecher. The corpus it was trained on is licensed directly.
Authenticity Amplified — the corpus →
Tragedy Academy Studios · The Source Voice
← The Corpus·The Source Voice