I analyzed Arpasing 0.2.0 reclist written by Kanru Hua (aka. the developer of Moresampler).
And the results of analyzing were following:
ch_ao, dh_ao, jh_ao, ng_ao, sh_ao, v_ao, z_ao, k_ao*, p_ao*, s_t_ao
dh_aa, ng_aa, sh_aa, th_aa, v_aa, s_k_aa, s_p_aa
ng_iy, th_iy, t_iy*, s_p_iy
ch_uw, dh_uw, g_uw, jh_uw, ng_uw, th_uw, v_uw, z_uw, s_p_uw
jh_eh, ng_eh, th_eh, z_eh
ch_uh, d_uh, dh_uh, f_uh, hh_uh, jh_uh, m_uh, n_uh, ng_uh, q_uh, r_uh, s_uh, th_uh, v_uh, z_uh, s_k_uh, s_p_uh, s_t_uh
s_k_ah, s_p_ah
ch_ae, ng_ae, sh_ae, v_ae, w_ae, z_ae, s_p_ae
jh_er, ng_er, r_er, th_er, s_k_er, s_p_er, s_t_er
ch_ey, jh_ey, ng_ey, th_ey, k_ey*, s_p_ey
ch_ay, dh_ay, ng_ay, th_ay, y_ay, k_ay*, s_p_ay
ch_ow, jh_ow, th_ow, v_ow, z_ow, s_k_ow, s_p_ow, s_t_ow
ch_aw, g_aw, jh_aw, k_aw, ng_aw, p_aw, s_aw, sh_aw, v_aw, w_aw, y_aw, z_aw, s_k_aw, s_p_aw, s_t_aw
ch_oy, d_oy, dh_oy, f_oy, hh_oy, k_oy, l_oy, m_oy, n_oy, ng_oy, p_oy, q_oy, r_oy, s_oy, t_oy, th_oy, v_oy, w_oy, y_oy, z_oy, s_k_oy, s_p_oy, s_t_oy
*Does have this sound combination but not as aspirated (I checked this by looking into index file if there was "s" sound beginning of k/t/p or not).
ao_b, ao_ch, ao_d, ao_dh, ao_g, ao_hh, ao_jh, ao_m, ao_p, ao_r, ao_sh, ao_th, ao_v, ao_w, ao_y, ao_z
aa_dh, aa_g, aa_hh, aa_jh, aa_ng, aa_r, aa_sh, aa_th, aa_v, aa_w, aa_y
iy_ng
uw_ch, uw_jh, uw_ng
eh_b, eh_ch, eh_dh, eh_hh, eh_jh, eh_ng, eh_th, eh_w, eh_y, eh_z
ih_b, ih_hh, ih_w, ih_y
uh_b, uh_ch, uh_dh, uh_f, uh_g, uh_hh, uh_jh, uh_m, uh_n, uh_p, uh_s, uh_th, uh_v, uh_w, uh_y, uh_z
ah_y
ae_b, ae_ch, ae_dh, ae_g, ae_hh, ae_jh, ae_jh, ae_r, ae_th, ae_w, ae_y
er_ch, er_jh, er_ng, er_r, er_sh
ey_ch, ey_ng, ey_r
ay_ch, ay_dh, ay_ng, ay_th, ay_y
ow_ch, ow_g, ow_jh, ow_ng, ow_r, ow_th
aw_b, aw_ch, aw_f, aw_g, aw_hh, aw_jh, aw_k, aw_m, aw_ng, aw_p, aw_r, aw_sh, aw_th, aw_v, aw_w, aw_y
oy_b, oy_ch, oy_d, oy_dh, oy_f, oy_g, oy_hh, oy_jh, oy_k, oy_l, oy_m, oy_ng, oy_p, oy_r, oy_s, oy_sh, oy_t, oy_th, oy_v, oy_w, oy_y
Note: There's none vowel_q combination in any vowel.
I interpret that if wanted to add "ax" sound (schwa) into reclist (which it lacks currently), the easiest/fastest way to do so is making a copy of cells that has "ah" sound, swap copied line's "ah" sound to "ax" and add the end of index file and ta-da! Fast made "reclist" for ax sounds!
I think that missing "hh" sound's VC aren't required to record because VC of h sound sounds often quite ugly in CVVC style voicebanks (yes, Arpasing is technically CVVC too but all sounds being purely diphones) and because English doesn't have ending h sound in any word.
I didn't check CC combination deeply but as I have been using Arpasing for months, I haven't found any missing CC so far. I checked that there were properly two types of "kr/tr/pr" recordings (aspirated or not), each one at least one as can seen below:
s_k_r = 1
k_r = 4
s_t_r = 2
t_r = 8
s_p_r = 1
p_r = 3
...so I think that adding any CC is not required.
Also, did you realize a HUGE about of CV for some y/r sounds? Well... The reason: many this recordings included CC recording like "cute"="k y uw t", "kribe" = "k r ay b" etc. (There's, of course, normal CV recordings for these consonants, not just CCV - just saying.)
I did this huge searching to help people (it took days to check this all, this wasn't one-click-magic-instant analyzing), so you all can quickly check what sounds Arpasing 0.2.0 reclist has and what not. If a person feels that they need xxxxx, then can add missing sounds themselves onto index and reclist file.
In my opinion, Arpasing 0.2.0 reclist is great and even better if people modify/edits reclist to fit own accent etc. I hope that you guys find this helpful and if I missed anything, feel free to comment.
And the results of analyzing were following:
ch_ao, dh_ao, jh_ao, ng_ao, sh_ao, v_ao, z_ao, k_ao*, p_ao*, s_t_ao
dh_aa, ng_aa, sh_aa, th_aa, v_aa, s_k_aa, s_p_aa
ng_iy, th_iy, t_iy*, s_p_iy
ch_uw, dh_uw, g_uw, jh_uw, ng_uw, th_uw, v_uw, z_uw, s_p_uw
jh_eh, ng_eh, th_eh, z_eh
ch_uh, d_uh, dh_uh, f_uh, hh_uh, jh_uh, m_uh, n_uh, ng_uh, q_uh, r_uh, s_uh, th_uh, v_uh, z_uh, s_k_uh, s_p_uh, s_t_uh
s_k_ah, s_p_ah
ch_ae, ng_ae, sh_ae, v_ae, w_ae, z_ae, s_p_ae
jh_er, ng_er, r_er, th_er, s_k_er, s_p_er, s_t_er
ch_ey, jh_ey, ng_ey, th_ey, k_ey*, s_p_ey
ch_ay, dh_ay, ng_ay, th_ay, y_ay, k_ay*, s_p_ay
ch_ow, jh_ow, th_ow, v_ow, z_ow, s_k_ow, s_p_ow, s_t_ow
ch_aw, g_aw, jh_aw, k_aw, ng_aw, p_aw, s_aw, sh_aw, v_aw, w_aw, y_aw, z_aw, s_k_aw, s_p_aw, s_t_aw
ch_oy, d_oy, dh_oy, f_oy, hh_oy, k_oy, l_oy, m_oy, n_oy, ng_oy, p_oy, q_oy, r_oy, s_oy, t_oy, th_oy, v_oy, w_oy, y_oy, z_oy, s_k_oy, s_p_oy, s_t_oy
*Does have this sound combination but not as aspirated (I checked this by looking into index file if there was "s" sound beginning of k/t/p or not).
ao_b, ao_ch, ao_d, ao_dh, ao_g, ao_hh, ao_jh, ao_m, ao_p, ao_r, ao_sh, ao_th, ao_v, ao_w, ao_y, ao_z
aa_dh, aa_g, aa_hh, aa_jh, aa_ng, aa_r, aa_sh, aa_th, aa_v, aa_w, aa_y
iy_ng
uw_ch, uw_jh, uw_ng
eh_b, eh_ch, eh_dh, eh_hh, eh_jh, eh_ng, eh_th, eh_w, eh_y, eh_z
ih_b, ih_hh, ih_w, ih_y
uh_b, uh_ch, uh_dh, uh_f, uh_g, uh_hh, uh_jh, uh_m, uh_n, uh_p, uh_s, uh_th, uh_v, uh_w, uh_y, uh_z
ah_y
ae_b, ae_ch, ae_dh, ae_g, ae_hh, ae_jh, ae_jh, ae_r, ae_th, ae_w, ae_y
er_ch, er_jh, er_ng, er_r, er_sh
ey_ch, ey_ng, ey_r
ay_ch, ay_dh, ay_ng, ay_th, ay_y
ow_ch, ow_g, ow_jh, ow_ng, ow_r, ow_th
aw_b, aw_ch, aw_f, aw_g, aw_hh, aw_jh, aw_k, aw_m, aw_ng, aw_p, aw_r, aw_sh, aw_th, aw_v, aw_w, aw_y
oy_b, oy_ch, oy_d, oy_dh, oy_f, oy_g, oy_hh, oy_jh, oy_k, oy_l, oy_m, oy_ng, oy_p, oy_r, oy_s, oy_sh, oy_t, oy_th, oy_v, oy_w, oy_y
Note: There's none vowel_q combination in any vowel.
I don't want to make a list of missing VV sounds because, in fact, Arpasing lacks most of VV combinations: it has the most "common/frequent" ones from American English only.
People with a British accent might want to add iy_ax, eh_ax, uw_ax, ay_ax and oy_ax VV to support British accent/pronouncing better.
000:
ay_er
er_iy
iy_ah
ah_iy
001:
ih_iy
iy_ay
ay_ih
ih_uw
002:
ow_ih
ih_uw
uw_ah
ah_ow
003:
ah_iy
iy_ih
ih_ow
ow_ay
004:
uw_ih
ih_uw
uw_ay
ay_ey
005:
ey_ih
ih_er
er_aw
aw_er
006:
iy_ey
ey_ay
ay_ow
ow_ah
007:
ao_aw
aw_ay
ay_ay
ay_ae
008:
eh_er
er_ah
ah_uw
uw_aa
009:
er_ay
ay_ah
ah_eh
eh_aw
and there's 320-334 string for "isolated vowels" meaning that each vowel will be recorded as long expect diphthongs are recorded twice like "ああいああい" or "ええいええい".
If recording full 5-mora for each VV recordings (000-009 in reclist) feels frustrating or difficult, then you can "slice" into smaller 2-mora chunks (or 3-mora whatever you prefer), editing index file and reclist file.
To be honest, it's not hard to edit index.csv file with Exel if knowing what doing but I might make an own result of that in future if requested a lot. It's totally ok to modify to reclist for personal use and to fit own needing better, don't afraid to edit reclit to fit to you!
A Tip: to replace missing VV sound, use "CV" sound of w* or use q (glottal stop)! E.g. "you and" = "y uw" "w eh" "eh n " OR "y uw" "q eh" "eh n". Change lyric and tune in a way that sounds best in the situation.
*at least with my voicebank, v sounds better than w for this purpose, but that's probably because my UTAU has heavy foreign accent *hahah*.
People with a British accent might want to add iy_ax, eh_ax, uw_ax, ay_ax and oy_ax VV to support British accent/pronouncing better.
000:
ay_er
er_iy
iy_ah
ah_iy
001:
ih_iy
iy_ay
ay_ih
ih_uw
002:
ow_ih
ih_uw
uw_ah
ah_ow
003:
ah_iy
iy_ih
ih_ow
ow_ay
004:
uw_ih
ih_uw
uw_ay
ay_ey
005:
ey_ih
ih_er
er_aw
aw_er
006:
iy_ey
ey_ay
ay_ow
ow_ah
007:
ao_aw
aw_ay
ay_ay
ay_ae
008:
eh_er
er_ah
ah_uw
uw_aa
009:
er_ay
ay_ah
ah_eh
eh_aw
and there's 320-334 string for "isolated vowels" meaning that each vowel will be recorded as long expect diphthongs are recorded twice like "ああいああい" or "ええいええい".
If recording full 5-mora for each VV recordings (000-009 in reclist) feels frustrating or difficult, then you can "slice" into smaller 2-mora chunks (or 3-mora whatever you prefer), editing index file and reclist file.
To be honest, it's not hard to edit index.csv file with Exel if knowing what doing but I might make an own result of that in future if requested a lot. It's totally ok to modify to reclist for personal use and to fit own needing better, don't afraid to edit reclit to fit to you!
A Tip: to replace missing VV sound, use "CV" sound of w* or use q (glottal stop)! E.g. "you and" = "y uw" "w eh" "eh n " OR "y uw" "q eh" "eh n". Change lyric and tune in a way that sounds best in the situation.
*at least with my voicebank, v sounds better than w for this purpose, but that's probably because my UTAU has heavy foreign accent *hahah*.
I interpret that if wanted to add "ax" sound (schwa) into reclist (which it lacks currently), the easiest/fastest way to do so is making a copy of cells that has "ah" sound, swap copied line's "ah" sound to "ax" and add the end of index file and ta-da! Fast made "reclist" for ax sounds!
I think that missing "hh" sound's VC aren't required to record because VC of h sound sounds often quite ugly in CVVC style voicebanks (yes, Arpasing is technically CVVC too but all sounds being purely diphones) and because English doesn't have ending h sound in any word.
I didn't check CC combination deeply but as I have been using Arpasing for months, I haven't found any missing CC so far. I checked that there were properly two types of "kr/tr/pr" recordings (aspirated or not), each one at least one as can seen below:
s_k_r = 1
k_r = 4
s_t_r = 2
t_r = 8
s_p_r = 1
p_r = 3
...so I think that adding any CC is not required.
Also, did you realize a HUGE about of CV for some y/r sounds? Well... The reason: many this recordings included CC recording like "cute"="k y uw t", "kribe" = "k r ay b" etc. (There's, of course, normal CV recordings for these consonants, not just CCV - just saying.)
I did this huge searching to help people (it took days to check this all, this wasn't one-click-magic-instant analyzing), so you all can quickly check what sounds Arpasing 0.2.0 reclist has and what not. If a person feels that they need xxxxx, then can add missing sounds themselves onto index and reclist file.
In my opinion, Arpasing 0.2.0 reclist is great and even better if people modify/edits reclist to fit own accent etc. I hope that you guys find this helpful and if I missed anything, feel free to comment.
