AIMC Notes · Case Study 09 · Controlled Voice Swap
Sound Chronicles Theme — Same Song, Different Voice
Sound Chronicles Theme. One song. Three audio files. The instrumental never changes — only the voice does. What do machine listening tools hear when an AI voice is replaced by a human singer?
1 · Question
If a human singer re-sings an AI-generated vocal over the identical instrumental, which differences do Cyanite (audio tagging) and Klangio (transcription) report — and which of those differences come from the voice, and which from the tool itself?
2 · Setup
| File | Content | Role |
|---|---|---|
| A · Min (AI) | Sound Chronicles Theme, original release mix. Vocal by the AI persona Min (Suno), incl. Suno backing vocals. | AI condition |
| B · Min – Human Avatar | Same song, sung by a professional session singer (R&B/Soul, booked via Fiverr) over the instrumental of A. | Human condition |
| C · Instrumental | The instrumental of A without lead and backing vocals, mixed from the Suno stems. This is exactly what the singer received. | Control |
- Song: 2:55 · G major · 75 BPM (measured on the control).
- The singer's brief, in two orders: first “sing it like the demo”, then “sing it your own way”. Both deliveries stayed melodically close to Min; the singer named his breathing as the only deliberate difference — “since AI don't have limited air to sing.”
- The human singer is referred to as Min – Human Avatar. His name stays anonymous at his request.
3 · Results — Cyanite (three-way)
3.1 Tempo & key
| C · Instrumental | A · Min | B · Min – Human Avatar | |
|---|---|---|---|
| BPM | 75 | 75 | 150 |
| Key | G major | G major | G major |
Finding: identical instrumental, identical tempo — but the human version is reported at double tempo. The voice alone triggered a tempo octave error.
3.2 Instruments
| Instrument | C · Instrumental | A · Min | B · Min – Human Avatar |
|---|---|---|---|
| Acoustic Guitar | Throughout | Throughout | Throughout |
| Electric Piano | — | Throughout | — |
| Electric Guitar | — | Frequently | — |
| Synthesizer | — | Frequently | — |
| Bass Guitar | Frequently | Frequently | Frequently |
| Percussion | — | Frequently | Frequently |
| Drum Kit | Partially | Frequently | Frequently |
| Piano | Partially | — | Frequently |
| Strings | Partially | — | Partially |
Finding: with the AI voice on top, Cyanite reports four instruments that are not in the instrumental (electric piano, electric guitar, synthesizer, percussion). With the human voice, it reports one (percussion) and upgrades piano and drums. The voice changes what the tool believes the band is playing.
3.3 Mood
| Mood | C · Instrumental | B · Min – Human Avatar | A · Min |
|---|---|---|---|
| Romantic | 0.73 | 0.85 | 0.89 |
| Chill | 0.57 | 0.73 | 0.79 |
| Calm | 0.36 | 0.46 | 0.49 |
| Sad | 0.31 | 0.39 | 0.45 |
| Happy | 0.20 | 0.35 | 0.25 |
| Ethereal | 0.11 | 0.19 | 0.19 |
| Sexy | — | 0.09 | 0.12 |
| Uplifting | 0.07 | 0.11 | 0.07 |
| Epic | 0.05 | — | — |
Finding: both voices raise romance and calm above the instrumental; the AI voice raises them most. The human voice is the only condition that pushes “happy” up.
3.4 Emotional profile & energy
| C · Instrumental | A · Min | B · Min – Human Avatar | |
|---|---|---|---|
| Emotional Profile | Balanced | Balanced | Balanced |
| Emotional Dynamics | Medium | Medium | Low |
| Energy Level | Low | Low | Low |
| Energy Dynamic | Low | Low | Low |
| Valence curve | ≈ −0.4 → peak +0.4 (1:30) → −0.25 | starts ≈ −0.6, rises to ≈ 0 | starts ≈ −0.3, positive from 0:30, peak ≈ +0.35 (2:00) |
| Arousal | negative throughout (≈ −0.25 to −0.48) | negative throughout | negative throughout (≈ −0.33 to −0.50) |
Finding: the AI voice keeps the emotional movement of the music (Medium) and starts darker. The human voice flattens it (Low) and lifts the whole song into positive valence.
3.5 Genre
| A · Min | B · Min – Human Avatar | |
|---|---|---|
| Main genres | Pop 0.42 · Folk/Country 0.39 | Pop 0.32 · R&B 0.28 · Folk/Country 0.26 |
| Free genres | Pop · Acoustic · Chill | Pop · Acoustic · Chill |
Finding: R&B appears only with the human singer — although he imitated the AI melody note for note. His own style is audible to the tool even when he tried to hide it.
3.6 Voice detection
| C · Instrumental | A · Min | B · Min – Human Avatar | |
|---|---|---|---|
| Predominant voice | None | Male | Male |
| Voice presence | None | High (≈ 1.0) | High (≈ 1.0, dip to 0.78 at 2:15) |
| “Female” at 2:15 | ≈ 0.08 | ≈ 0.6 | ≈ 0.24 |
| False voice in instrumental | male up to 0.23 · female up to 0.14 | — | — |
Finding: at 2:15 both singers go high, and the tool reads a high male note partly as a female voice. The control shows no such spike. In the instrumental, the tool still reports a faint “male” voice where there is none.
4 · Results — Klangio transcription (vocal stems)
| Min (AI) | Min – Human Avatar | |
|---|---|---|
| Key · tempo marks | 1 sharp (G) · ♩ 63 / 79 / 63 | 1 sharp (G) · ♩ 63 / 79 / 62 |
| Notes | 202 | 222 |
| Rests (in beats) | 88.3 | 67.9 |
| Held notes (≥ 2 beats) | 1 | 8 |
| Repeated same pitch | 59 | 71 |
| Leaps (≥ 5 semitones) | 29 | 36 |
| Notated range | D3 – A4 | D3 – A4 |
| Most frequent pitches | G · B · A · D · E | G · B · A · D · E |
Finding: the written skeleton is almost identical — same key, same range, same pitch hierarchy. The difference is in the space: the AI leaves more rests and moves on; the human holds notes and fills the gaps.
5 · Results — pitch analysis (between the notes)
| Min (AI) | Min – Human Avatar | |
|---|---|---|
| Sung range | D3 – F♯4 (16 semitones) | E3 – E4 (12 semitones) |
| Median pitch | B3 (247 Hz) | A3 (223 Hz) |
| Pitch changes per second | 10.6 | 7.6 |
| Phrases | 51 | 33 |
| Breaths detected | 24 | 20 |
| Dynamic range (P10–P90) | 10.4 dB | 9.0 dB |
| Notes with vibrato | 42 / 60 | 46 / 75 |
| Vibrato rate · depth | 5.7 Hz · 84 cents | 6.1 Hz · 92 cents |
| Jitter · shimmer · HNR | 1.11 % · 6.0 % · 17.6 dB | 0.78 % · 5.6 % · 16.7 dB |
| Spectral centroid | 1,939 Hz (darker) | 2,291 Hz (brighter) |
Finding: Klangio rounds every bend to the nearest note, so both singers look the same on paper. Between the notes, the AI voice moves more: wider range, more pitch movement, more phrases, more breaths, more dynamics. Where Min bends or rises (0:13, 0:20, 2:00, 2:51), the Min – Human Avatar sings straight; where Min stays plain (2:45), the Min – Human Avatar adds vibrato.
5b · Direct comparison — how closely does the human reproduce the AI identity?
| Measure | Value | Reading |
|---|---|---|
| Time both sing | 72.9 s | shared sung time |
| Only Min (AI) sings | 14.8 s | the human is silent here |
| Only Min – Human Avatar sings | 37.6 s | the human sings into Min's silences |
| Same note (±50 cents) in shared time | 56.6 % | within one semitone: 68.0 % |
| Median pitch offset (Human − AI) | -0.06 semitones | no transposition — same key, same register |
| Melody contour correlation | r = 0.609 | lowest window 2:20–2:30 (r ≈ −0.04), highest at the end (r ≈ 0.92) |
| Phrase onsets matched | 27 | median offset 0 ms, typical deviation 40 ms — he starts exactly where Min starts |
| Median phrase length | 1.46 s · 3.23 s | AI · Human — the human sings phrases twice as long |
| Longest phrase | 5.64 s · 11.98 s | AI · Human |
| Spectral bandwidth | 2891 Hz · 3011 Hz | AI · Human |
| Timbre distance (MFCC 1–12) | 51.3 | clearly different voice colour despite same notes |
Finding: the human copies the AI's timing almost perfectly (phrase starts within ~40 ms) and stays in the same register. He departs in two places: he fills Min's silences (37.6 s sung where Min is silent, against 14.8 s the other way) and he binds Min's short phrases into long lines. The identity he reproduces is the melody and the timing; the identity he does not reproduce is the silence.
6 · What this case shows
- The voice changes the tool's picture of the whole track. Same instrumental, different voice: different instruments, different tempo.
- The AI voice carries an electronic signature that Cyanite turns into four instruments that are not there; the human voice triggers one.
- The human singer's style leaks through (R&B) even when he copies the AI melody.
- The human copies timing, not silence: phrase starts within ~40 ms of the AI, but 37.6 s sung where the AI rests.
- The AI version breathes more in the music — more rests, more phrases, more movement — while the human holds and fills.
- Romance lives in the song, intensity in the voice: 0.73 without a voice, 0.85 with the human, 0.89 with Min.
7 · What this case does not claim
- One song, one human singer, one AI persona. This is a case study, not a general law about Cyanite, Klangio or AI voices.
- The instrumental was mixed from Suno stems and sits slightly lower in level than the original release mix.
- The AI condition contains Suno backing vocals; the human condition and the control do not. Part of the “female” reading at 2:15 in file A may come from backing vocals.
- Klangio transcriptions come from source-separated stems and are estimates. Tempo maps differ slightly (62 vs 67 measures).
- Pitch metrics come from separated vocals, not isolated studio recordings. Loudness, clipping and formant analysis were deliberately not done on these stems.
- Instrument and mood values were read from Cyanite's interface; curve values marked “≈” were read from charts.
- No judgement of quality is made here. The measurements describe difference, not better or worse.
8 · Evidence
| Item | Source |
|---|---|
| Cyanite · A · Min | Contest-MP3 320k_Sound Chronicles Theme.mp3 · Cyanite library · 08.10.2026 |
| Cyanite · B · Min – Human Avatar | Human vocal version, full mix · genre, mood, energy, instruments, voice · 06.11.2026 |
| Cyanite · C · Instrumental | Instrumental (control), as sent to the singer · mood, energy, instruments, voice · 06.11.2026 |
| Klangio | Vocal stem Min (AI) · vocal stem Min – Human Avatar · MusicXML · 06.11.2026 |
| Pitch analysis | Separated vocal stems of A and B · 01.11. and 06.11.2026 |
| All values | CSV: AIMC Fallstudie SC Theme Messwerte.csv (66 rows, source per value) |
| Listening comparison | 6 passages, Min (AI) first, then Min – Human Avatar: 0:13 · 0:20 · 1:51 · 2:00 · 2:45 · 2:51 |