THIRI × ElevenLabs Research Suite

6 Interactive Experiments · Closed-Loop Pitch Tracking · SATB Quartets · Solfège Engine
← Return to Reports Hub Voice AI v3.2
⚡ Registered Antigravity & Audio Engineering Skills
dual-format-html-reports GLOBAL & REPO
Builds and maintains interactive dual-format HTML reports (16:9 Landscape & 9:16 Phone), manages camera/talking-head PIP safe zones with zero UI overlap, and auto-updates the index portal.
View SKILL.md → Reports Engine
sound-association-graph GLOBAL & REPO
Maps acoustic corpora into 3D and orchestral semi-circle amphitheater spaces, computes Dan Bruton frequency-to-light octave transposition (380-750 nm), Web Audio synthesis, and multi-track WAV loop exporting.
View SKILL.md → Acoustic Graph
soundtrack-to-picture REPO SKILL
Scores and sound-designs videos to picture. Frame-accurate cue sheets, Csound rendering, dialogue/SFX/music stem separation, sample retuning by attack, and ITU-R BS.1770-4 loudness compliance.
View SKILL.md → Scoring Engine
agentic-radio-station GLOBAL SKILL
Runs 24/7 autonomous live audio-video broadcast streams on YouTube Live and Twitch with real-time audience steering, procedural Python sound generation, and ElevenLabs spoken voice affirmations.
View Global SKILL.md → Autonomous Broadcast
gemini-api-dev & Voice BUILT-IN AGENT
Voice cloning, speech synthesis (TTS), voice replication, closed-loop microtonal intonation tracking, and real-time streaming multimodal model operations.
View Plugin SKILL.md → Voice & Audio AI

"In A Sentimental Mood" SATB Vocal Quartet

Full 4-Part Jazz Polyphony voiced by THIRI Harmonic Engine
Polyphonic Synthesis
⚡ Engineered With: soundtrack-to-picture & dual-format-html-reports
Master Quartet Mix (SATB Combined)
Lily (Soprano) · Sarah (Alto) · Dennison (Tenor) · Charlie (Bass)
SOPRANO (Lily)
Lead Melody · Range: C4 - G5
Gain0 dB
ALTO (Sarah)
Harmonic 3rds & 7ths · G3 - D5
Gain0 dB
TENOR (Dennison IVC)
Counter-Melody · C3 - G4
Gain0 dB
BASS (Charlie)
Root Anchor · E2 - C4
Gain0 dB
How To Run This Experiment
1
Calculate Harmonic SATB Voicings: THIRI Harmonic Engine generates textbook drop-2 or open 4-way jazz voicings for each chord in the progression.
2
Phonetic Syllable Alignment: Lyrics are formatted with explicit hyphenated vowel-lengthening (e.g. "In-a sen-ti-men-taaaal moooood...") to guide the ElevenLabs duration generator.
3
Multi-Model Prompting: Execute parallel requests across distinct ElevenLabs voice IDs: Lily (Soprano), Sarah (Alto), Dennison Clone (Tenor), and Charlie (Bass).
python vibe-studio/out/in-a-sentimental-mood/fetch_11labs_takes.py
4
Phase Alignment & Transient Locking: Run render_11labs_quartet.py to align first onsets within $\pm 2\text{ms}$ and bounce the master 24-bit WAV.
View Python Pipeline Script

Closed-Loop Microtonal Pitch Tracking

Reel 01 · THIRI Deterministic Ear + ElevenLabs Feedback Loop
Closed Loop
⚡ Engineered With: sound-association-graph (FFT / pYIN Cents Gauge) & gemini-api-dev
LIVE INTONATION ERROR (CENTS)
0.0 Cents (In Tune)
Simulate Voice Pitch Deviation0 ¢
The Closed-Loop Feedback Loop
1
Speech-to-Analysis Ingestion: The raw audio output from ElevenLabs is immediately streamed into vibe.ear.hear_window().
2
FFT Peak Frequency Extraction: Signal processing identifies fundamental frequencies and measures deviation in cents:
cents = 100 * (hz_to_midi(f0) - round(hz_to_midi(f0)))
3
Automated Re-Take Verification: If the detected pitch deviates by more than $\pm 18\text{ cents}$ from the target chord scale degree, the engine rejects the take and re-prompts with higher stability (s75).
✓ Eliminates the uncanny valley of robotic AutoTune while guaranteeing textbook jazz intonation.

Harmonica Bend to Vocal Inflection

Reel 02 · Translating Physical Bends into Neural Vocal Nuance
Acoustic Mapping
⚡ Engineered With: soundtrack-to-picture & sound-association-graph (Contour Extraction)
A/B CONTOUR COMPARISON
Listen to how the draw bend on Hole 4 ($D \to Db \to C$) translates into microtonal vocal slide and throat rasp.
How To Run This Experiment
1
Record Acoustic Blue Note: Record an acoustic blues harmonica lick with deep reed bending ($b3$ or $b5$).
2
Track Pitch Contour: Extract the continuous $F_0(t)$ pitch trajectory via librosa pYIN.
f0, voiced_flag, _ = librosa.pyin(y, fmin=65, fmax=1000)
3
Speech-to-Speech (STS) Feeding: Pass the harmonica stem directly into ElevenLabs STS with style exaggeration set to maximum (style=1.0).

Dennison Instant Voice Clone Calibration

Reel 04 · Model Calibration: sp100_s50_sb83
Voice Cloning
⚡ Engineered With: gemini-api-dev (Voice Cloning & Model Calibration)
Dennison Cloned Voice Calibration Take
sp100 (Style 100%) · s50 (Stability 50%) · sb83 (Clarity 83%)
STABILITY (s50)50% (Calibrated Golden Ratio)
Balances emotional expressiveness against pitch jitter.
SIMILARITY BOOST (sb83)83% (Clarity & Throat Resonance)
Preserves authentic vocal tract identity without high-frequency harshness.
STYLE EXAGGERATION (sp100)100% (Maximum Phrasing)
Forces the model to execute jazz breathiness and melodic inflection.
How To Run Voice Clone Calibration
1
Dataset Ingestion: Upload 3 to 5 minutes of dry, 48kHz monophonic speech and vocal singing runs.
2
Grid Sweep Script: Run an automated grid sweep testing permutations of stability $[25, 50, 75]$ vs similarity $[50, 83, 100]$.
3
Intonation Scoring: Score each take using THIRI Ear. The preset sp100_s50_sb83 demonstrated the highest harmonic fidelity.

Interactive Chromatic Solfège Widget

Moveable-Do Ear Training with Speech & Tone Synthesis
Education Tech
⚡ Engineered With: sound-association-graph (Web Audio Triangle Synthesis & Ear Training)
Do1
Re2
Mi3
Fa4
Sol5
La6
Ti7
Do'8
Click any syllable above to hear the pitch generated in real-time alongside spoken solfège voice articulation!
What Problems Does This Solve?
☆
Ear Training Flexibility: Students no longer rely on static audio tracks. They can transpose moveable-do exercises into all 12 keys instantaneously.
☆
Dialect Standardization: Automatically handles European fixed-do (Si) vs American moveable-do (Ti) preventing student confusion.
☆
Phonetic Pronunciation Guardrail: Overcomes TTS voice generator mispronunciations by automatically passing the phonetic substitution Theory to the speech engine while displaying THIRI on screen.

"Eleven Labs Gets a Jazz Vocal Lesson"

Full Video Showcase & Skool Community Funnel
YouTube Master
⚡ Engineered With: soundtrack-to-picture & agentic-radio-station (Broadcast Staging)
Format: 1080p · Subtitles (.srt) & Chapters Included
View Video Copy
YouTube Chapters & Funnel Hook
00:00 - Introduction: Why Voice AI Struggles with Jazz
00:45 - The Ellington SATB Voicing Challenge
02:15 - Closing the Loop: THIRI Ear Pitch Tracking
04:10 - Harmonica Bends & Expressive Inflection
06:30 - The 4-Part Polyphonic Master Render
08:15 - How to Clone Your Own Voice for Music Instruction
Skool Community Pinned Hook:
"We taught ElevenLabs how to sing 4-part jazz voicings using THIRI. Grab the full source code and SATB stems in Classroom Module 01."