What first caught my attention was actually the juxtaposition of two different performances of "FIRST KISS". The first was her original role as part of Aa! in 2003. The second was a solo concert version performed in 2007 as part of the °C-ute Autumn Live Tour. Maybe part of it was the lower quality of Youtube audio, but I don't think that changes vocal timbre to this large of an extent. The opening line of the song sounded rather different to my ears in the two versions, and the difference was only further emphasized in the rest of the song. Whereas Airi's voice in the original version slightly childish but overall what I would consider round and resonant, the solo version seemed to have a brighter but also more pinched sound.
Of course, this contrast then made me wonder if her voice was just tired from singing through a long concert, or if it was indicative of some actual trend in Airi's vocal development. And being the speech processing nerd that I am, along with some inspiration from Kawaiirhea, I decided it would be interesting to run some voice analysis software to find out what scientific basis was underlying the changes I perceived.
Because my computer's power supply decided to go kaput recently, I've been using a different computer with inadequate RAM to really run big programs, so I selected four songs I felt were representative of Airi's singing between 2003 and 2008:
* FIRST KISS/Aa!
* Suki ni Naccha Ikenai Hito/H!P All-Stars
* Sakura Chirari/C-ute
* LALALA Shiawase no Uta/C-ute
I then selected segments (or a conglomeration of shorter segments) lasting between 0.5 and 1.0 seconds total. I tried to make my segment choices as uniform as possible by selecting sections where she is singing in about the same range, and holding a note constant rather than going through lots of pitch changes.
Because formants play a large role in determining the perceived timbre of a speaking voice, I figured they would probably also be influential for a singing voice as well. Generally, fundamental frequency and intensity are also important for speaking voice timbers, but fundamental frequency/pitch tracking was not going to be particularly enlightening since the point singing is in fact to modulate and change pitches, and intensity too is variable in a musical manner and not necessarily a property of the voice itself in singing. Thus, I plotted spectrograms for my musical selections, with the formant tracks drawn on them:
FIRST KISS:
Suki ni Naccha Ikenai Hito:
Sakura Chirari:
LALALA Shiawase no Uta:
And to summarize, since it is difficult to glean much information from the spectrograms without closer zoom, I list the average frequency of the first four formants for each song in chronological order (F1 is the lowest, F5 is the highest in the spectrograms):
| F1 | F2 | F3 | F4 | |
| FIRST KISS | 929.77 | 1864.04 | 2810.03 | 3994.20 |
| Suki ni Naccha Ikenai Hito | 1050.59 | 1904.04 | 3140.82 | 3961.91 |
| Sakura Chirari | 979.70 | 2046.94 | 2780.12 | 3918.58 |
| LALALA Shiawase no Uta | 760.45 | 1695.28 | 2915.59 | 4077.39 |
Ignoring all scientific analysis, my ear tells me in listening to these four songs that there seems to have been three different stages in Airi's vocal development in the last five years: the early stage, happening between 2003 and 2004 with the first two song selections, a middle stage around 2006-2007, and the current stage. The sound of the early stage, is childish but has a nice resonance to it, with a hint of breathiness. The middle stage seems to be characterized by a more focused but also more pinched sort of sound. I was relieved though that the sound of the current stage seems to have dropped the pinchedness/brightness while maintaining a good focused core.
These intuitions appear to be supported by my formant analyses as well. The "acuteness" of a timbre, meaning perceived brightness and tenseness, is generally correlated with the frequency of the second formant. A low second formant results in a non-acute sound, whereas a high second formant results in an acute sound. If we track the progression of average F2 frequencies chronologically, they begin at around 1900 Hz with the first two songs, increase and peak at 2046.94 in "Sakura Chirari" (early 2007). They then exhibit an amazing drop down to 1695.28 in "LALALA Shiawase no Uta" (2008), hence the three stages I had perceived.
There could be several factors in this development pattern, not the least of which is puberty. The rising acuteness between 2004 and 2006 I suspect is most likely attributed to voice training. If we observe the distinctiveness of the formant lines in "FIRST KISS", there seems to be a tendency for the second formant to alternately blend with F1 and F3, which I think corresponds to a less focused, more breathy and fuzzy kind of sound. While that can still be a pretty sound, it also means project requires more volume and thus effort, and in general I think a clearer sound is preferred for the style of the H!P Kids groups. The best way to create a focused sound is in fact to make it a brighter tone color, which when taken too far results in some pinchedness in the sound. This probably why the F2 frequency exhibits an increasing trend. As she got more experienced with creating a focused but still open sound, the F2 frequency naturally lowers again. I think the drasticness of the drop is probably due to the her learning how to create an open sound coinciding with puberty causing her voice to fill out and mature.
With the current timbre she is getting, I'm very optimistic about Airi's future vocal development. She has a beautiful voice and she definitely seems to be getting the hang of how to use it. She'll have to make adjustments as it continues to mature, obviously, but she seems to be on a good track thus far. Ganbare Airi!
Disclaimer: I don't claim to be an expert in singing voice analysis, and all of this may just be a coincidence... But the observed patterns seem to make sense as far as I can tell based on my experience with speaking voice analysis.
