Is there any guides here for making my own VCV reclists?

[WoofMatcha]

Tbh I feel some reclists are not fitting my voice, some are even having too many lines to record... while missing some kana lines or having a "weird" base oto that makes SetParam confused and saying missing wav files while I've loaded the right one.

I actually want to extract the reclists from 狼音アロ (Rouon Aro)-Nova but well, I don't want to cause trouble with the owner so I'd not do that. Tbh his voicebank is sounds good and also having less samples too, it just 153 samples and also efficient. I bet the VP, 牙音 is making a custom reclist for Aro.

So... I was thinking about making a tailored reclist for myself to make it efficient and also timesaving while also having less samples, or maybe having shorter mora amounts to keep my tone consistent? My breath is kinda short so I think the only best thing is making the reclist as short as possible, especially if I'm recording a long line I might get coughing so bad because my body is sucks. Gets sick so easy LMAO

And it causing my voice is kinda sounded weak or maybe too soft for a male voice LOL

Post automatically merged: May 19, 2025

I want it to be efficient that I can have time to record multipitch too.

 

[WoofMatcha]

I was thinking to make 12 mora reclist for my voicebank but it might not good for my vocal cord.
So... does 6 mora good enough for this? I want to make it kinda balanced.

Post automatically merged: May 20, 2025

While I trying to avoid the metallic voice on some transitions to match my old test voicebank.

 

[SaKe]

6 mora is probably a little short for it to be efficient.
I wouldn’t see the problem with extracting the reclist from another voicebank. With Japanese I’d highly doubt it’s an original reclist- those in general are pretty hard to find, since what’s the point in writing a reclist you aren’t going to release?

 

[WoofMatcha]

I found 8 mora reclist by Gyromancy, it's kana-encoded and have no Romaji.

I tried to use GitHub Copilot on VS Code to turn it to Romaji for making OREMO comment, and I see all of the lines are kinda easy for my voice... I mean rolls well with my tongue.
But probably inaccurate, and yeah I asked it to make the same sounding letters to be kept as is on the sounding. Cuz the "ro" on ろ often read as "lo" in my opinion, cuz I tried an UST of "Hello" but the "hello" part turned to be sound as "herro". LMAO

Maybe you guys could help me to correct the mistake by the rough conversion by the AI:

Spoiler: Romaji reclists that converted by AI

_a-a-i-a-u-e-a
_i-i-u-a-e-i-e
_u-u-i-o-a-o-i
_e-e-u-o-e-o-o
_o-u-n-a-n-n-u
_n-i-n-e-n-o-n
_ka-n-ka-ka-ki-ka-ku-ka
_ki-n-ki-ki-ku-ki-ke-ki
_ku-n-ku-ku-ke-ku-ko-ku
_ke-n-ke-ke-ko-ke-ka-ke
_ko-n-ko-ko-ka-ko-ki-ko
_sa-n-sa-sa-si-sa-su-sa
_si-n-si-si-su-si-se-si
_su-n-su-su-se-su-so-su
_se-n-se-se-so-se-sa-se
_so-n-so-so-sa-so-si-so
_ta-n-ta-ta-ti-ta-tu-ta
_ti-n-ti-ti-tu-ti-te-ti
_tu-n-tu-tu-te-tu-to-tu
_te-n-te-te-to-te-ta-te
_to-n-to-to-ta-to-ti-to
_na-n-na-na-ni-na-nu-na
_ni-n-ni-ni-nu-ni-ne-ni
_nu-n-nu-nu-ne-nu-no-nu
_ne-n-ne-ne-no-ne-na-ne
_no-n-no-no-na-no-ni-no
_ha-n-ha-ha-hi-ha-fu-ha
_hi-n-hi-hi-fu-hi-he-hi
_fu-n-fu-fu-he-fu-ho-fu
_he-n-he-he-ho-he-ha-he
_ho-n-ho-ho-ha-ho-hi-ho
_ma-n-ma-ma-mi-ma-mu-ma
_mi-n-mi-mi-mu-mi-me-mi
_mu-n-mu-mu-me-mu-mo-mu
_me-n-me-me-mo-me-ma-me
_mo-n-mo-mo-ma-mo-mi-mo
_ya-n-ya-ya-i-ya-yu-ya
_yu-n-yu-yu-ie-yu-yo-yu
_ie-n-ie-ie-yo-ie-ya-ie
_yo-n-yo-yo-ya-yo-i-yo
_ra-n-ra-ra-ri-ra-ru-ra
_ri-n-ri-ri-ru-ri-re-ri
_ru-n-ru-ru-re-ru-ro-ru
_re-n-re-re-ro-re-ra-re
_ro-n-ro-ro-ra-ro-ri-ro
_wa-n-wa-wa-wi-wa-u-wa
_wi-n-wi-wi-u-wi-we-wi
_we-n-we-we-wo-we-wa-we
_wo-n-wo-wo-wa-wo-wi-wo
_u-we-u-wo-i-yu-i-ie
_ga-n-ga-ga-gi-ga-gu-ga
_gi-n-gi-gi-gu-gi-ge-gi
_gu-n-gu-gu-ge-gu-go-gu
_ge-n-ge-ge-go-ge-ga-ge
_go-n-go-go-ga-go-gi-go
_za-n-za-za-zi-za-zu-za
_zi-n-zi-zi-zu-zi-ze-zi
_zu-n-zu-zu-ze-zu-zo-zu
_ze-n-ze-ze-zo-ze-za-ze
_zo-n-zo-zo-za-zo-zi-zo
_da-n-da-da-di-da-du-da
_di-n-di-di-du-di-de-di
_du-n-du-du-de-du-do-du
_de-n-de-de-do-de-da-de
_do-n-do-do-da-do-di-do
_ba-n-ba-ba-bi-ba-bu-ba
_bi-n-bi-bi-bu-bi-be-bi
_bu-n-bu-bu-be-bu-bo-bu
_be-n-be-be-bo-be-ba-be
_bo-n-bo-bo-ba-bo-bi-bo
_pa-n-pa-pa-pi-pa-pu-pa
_pi-n-pi-pi-pu-pi-pe-pi
_pu-n-pu-pu-pe-pu-po-pu
_pe-n-pe-pe-po-pe-pa-pe
_po-n-po-po-pa-po-pi-po
_kya-n-kya-kya-ki-kya-kyu-kya
_ki-n-ki-ki-kyu-ki-kye-ki
_kyu-n-kyu-kyu-kye-kyu-kyo-kyu
_kye-n-kye-kye-kyo-kye-kya-kye
_kyo-n-kyo-kyo-kya-kyo-ki-kyo
_sha-n-sha-sha-shi-sha-shu-sha
_shi-n-shi-shi-shu-shi-she-shi
_shu-n-shu-shu-she-shu-sho-shu
_she-n-she-she-sho-she-sha-she
_sho-n-sho-sho-sha-sho-shi-sho
_cha-n-cha-cha-chi-cha-chu-cha
_chi-n-chi-chi-chu-chi-che-chi
_chu-n-chu-chu-che-chu-cho-chu
_che-n-che-che-cho-che-cha-che
_cho-n-cho-cho-cha-cho-chi-cho
_nya-n-nya-nya-ni-nya-nyu-nya
_ni-n-ni-ni-nyu-ni-nye-ni
_nyu-n-nyu-nyu-nye-nyu-nyo-nyu
_nye-n-nye-nye-nyo-nye-nya-nye
_nyo-n-nyo-nyo-nya-nyo-ni-nyo
_hya-n-hya-hya-hi-hya-hyu-hya
_hi-n-hi-hi-hyu-hi-hye-hi
_hyu-n-hyu-hyu-hye-hyu-hyo-hyu
_hye-n-hye-hye-hyo-hye-hya-hye
_hyo-n-hyo-hyo-hya-hyo-hi-hyo
_mya-n-mya-mya-mi-mya-myu-mya
_mi-n-mi-mi-myu-mi-mye-mi
_myu-n-myu-myu-mye-myu-myo-myu
_mye-n-mye-mye-myo-mye-mya-mye
_myo-n-myo-myo-mya-myo-mi-myo
_rya-n-rya-rya-ri-rya-ryu-rya
_ri-n-ri-ri-ryu-ri-rye-ri
_ryu-n-ryu-ryu-rye-ryu-ryo-ryu
_rye-n-rye-rye-ryo-rye-rya-rye
_ryo-n-ryo-ryo-rya-ryo-ri-ryo
_gya-n-gya-gya-gi-gya-gyu-gya
_gi-n-gi-gi-gyu-gi-gye-gi
_gyu-n-gyu-gyu-gye-gyu-gyo-gyu
_gye-n-gye-gye-gyo-gye-gya-gye
_gyo-n-gyo-gyo-gya-gyo-gi-gyo
_ja-n-ja-ja-ji-ja-ju-ja
_ji-n-ji-ji-ju-ji-je-ji
_ju-n-ju-ju-je-ju-jo-ju
_je-n-je-je-jo-je-ja-je
_jo-n-jo-jo-ja-jo-ji-jo
_bya-n-bya-bya-bi-bya-byu-bya
_bi-n-bi-bi-byu-bi-bye-bi
_byu-n-byu-byu-bye-byu-byo-byu
_bye-n-bye-bye-byo-bye-bya-bye
_byo-n-byo-byo-bya-byo-bi-byo
_pya-n-pya-pya-pi-pya-pyu-pya
_pi-n-pi-pi-pyu-pi-pye-pi
_pyu-n-pyu-pyu-pye-pyu-pyo-pyu
_pye-n-pye-pye-pyo-pye-pya-pye
_pyo-n-pyo-pyo-pya-pyo-pi-pyo
_tsa-n-tsa-tsa-tsi-tsa-tsu-tsa
_tsi-n-tsi-tsi-tsu-tsi-tse-tsi
_tsu-n-tsu-tsu-tse-tsu-tso-tsu
_tse-n-tse-tse-tso-tse-tsa-tse
_tso-n-tso-tso-tsa-tso-tsi-tso
_fa-n-fa-fa-fi-fa-fu-fa
_fi-n-fi-fi-fu-fi-fe-fi
_fu-n-fu-fu-fe-fu-fo-fu
_fe-n-fe-fe-fo-fe-fa-fe
_fo-n-fo-fo-fa-fo-fi-fo
_va-n-va-va-vi-va-vu-va
_vi-n-vi-vi-vu-vi-ve-vi
_vu-n-vu-vu-ve-vu-vo-vu
_ve-n-ve-ve-vo-ve-va-ve
_vo-n-vo-vo-va-vo-vi-vo
_kwa-n-kwa-kwa-kwi-kwa-ku-kwa
_kwi-n-kwi-kwi-ku-kwi-kwe-kwi
_ku-n-ku-ku-kwe-ku-kwo-ku
_kwe-n-kwe-kwe-kwo-kwe-kwa-kwe
_kwo-n-kwo-kwo-kwa-kwo-kwi-kwo
_gwa-n-gwa-gwa-gwi-gwa-gu-gwa
_gwi-n-gwi-gwi-gu-gwi-gwe-gwi
_gu-n-gu-gu-gwe-gu-gwo-gu
_gwe-n-gwe-gwe-gwo-gwe-gwa-gwe
_gwo-n-gwo-gwo-gwa-gwo-gwi-gwo
_tya-n-tya-tya-ti-tya-tyu-tya
_ti-n-ti-ti-tyu-ti-tye-ti
_tyu-n-tyu-tyu-tye-tyu-tyo-tyo
_tye-n-tye-tye-tyo-tye-tya-tye
_tyo-n-tyo-tyo-tya-tyo-ti-tyo
_dya-n-dya-dya-di-dya-dyu-dya
_di-n-di-di-dyu-di-dye-di
_dyu-n-dyu-dyu-dye-dyu-dyo-dyu
_dye-n-dye-dye-dyo-dye-dya-dye
_dyo-n-dyo-dyo-dya-dyo-di-dyo
_ra-n-ra-ra-ri-ra-ru-ra
_ri-n-ri-ri-ru-ri-re-ri
_ru-n-ru-ru-re-ru-ro-ru
_re-n-re-re-ro-re-ra-re
_ro-n-ro-ro-ra-ro-ri-ro

I take the reclist (JP VCV 8-mora) from here: https://gyromancy.wixsite.com/home/reclists

 

[FelineWasteland]

The basic string length for any VCV is going to be V+1 where V is the number of vowels (or other nuclei) included. For Japanese, this is 7 because of the five vowels and syllabic nasal. This is because you need one line beginning with each V for the [- CV] sample, a second CV with that vowel, and then an alternating pattern covering every other VV pair, cutting off the pattern before any duplicates would arise.

Longer string lengths won't really work unless you're omitting the [- CV]s or recording them separately from the rest of the reclist.

So, the basic string pattern is:

ka-ka-ki-ka-ku-ka-ke
ki-ki-ku-ki-ke-ki-ko
ku-ku-ke-ku-ko-ku-n
ke-ke-ko-ke-n-ke-ka
ko-ko-n-ko-ka-ko-ki
n-n-ka-n-ki-n-ku

If identical string lengths throughout the reclist isn't important, there are some ways you can make the pattern more efficient by rearranging redunant parts of the string. For example, you'll already be pulling [N V]s from the vowel strings, so you don't need all those [N]s throughout every CV string, so you can make a new vowel pattern than excludes [N] and attach the N-CV segment to the end of the string, resulting in 8-mora.

This will result in a shorter reclist length overall because now there's one less string for each consonant:

ka-ka-ki-ka-ku-ka-n-ka
ki-ki-ku-ki-ke-ki-n-ki
ku-ku-ke-ku-ko-ku-n-ku
ke-ke-ko-ke-ka-ke-n-ke
ko-ko-ka-ko-ki-ko-n-ko

You can use this strategy for any CVs you want to omitt; just repattern the string without them, and then attach them at the end. For example, you can do the same the [Cy] strings to exclude [Cyi] by creating a pattern that excludes [i ] in addition to [N] and attaching both to the end.

You can split up the string into a new line at any point as long as you remember to double the mora at the splitting point so that you don't leave out that [V CV]. For example, if you're splitting ka-ka-ki-ka-ku-ka-ke after the 3rd [ka], remember to add another [ka] to the new string, like ka-ka-ki-ka and ka-ku-ka-ke.

If this results in uneven string lengths, you can try to combine the "leftover" CVs into their own strings, but it doesn't always math out well. This is why 5-mora is usually the next most common length after 7-mora, because the strings are a bit shorter but still result in even lengths. I've yet to find a good pattern for 6-mora but I won't say it's impossible.

Splitting strings can drastically increase the amount of strings you have to record overall, though, because it adds +however many additional strings in the pattern to every consonant. You can't really have it both ways of making a list with both shorter strings and also fewer strings.

These are the patterns I used for writing my own VCV lists. If you tried one of them and it gave you an error, do you mind telling me which list it was? It's possible there's just a typo in either the list or the base oto that I can correct for future users.

The other option too is to do a lite list where you don't include every possible [V CV] pair, instead just including regular [CV]s pulled from the middle of the string in addition to some [V CV]s, though I'm not sure how well that plays with SetParam's oto generation. Personally I don't like this type of reclist so I haven't written any, but they do exist. The other other option is to do CVVC; those lists are much shorter than VCV even with shorter string lengths, and can sound just as smooth.

 

[WoofMatcha]

The basic string length for any VCV is going to be V+1 where V is the number of vowels (or other nuclei) included. For Japanese, this is 7 because of the five vowels and syllabic nasal. This is because you need one line beginning with each V for the [- CV] sample, a second CV with that vowel, and then an alternating pattern covering every other VV pair, cutting off the pattern before any duplicates would arise.

Longer string lengths won't really work unless you're omitting the [- CV]s or recording them separately from the rest of the reclist.

So, the basic string pattern is:

ka-ka-ki-ka-ku-ka-ke
ki-ki-ku-ki-ke-ki-ko
ku-ku-ke-ku-ko-ku-n
ke-ke-ko-ke-n-ke-ka
ko-ko-n-ko-ka-ko-ki
n-n-ka-n-ki-n-ku

If identical string lengths throughout the reclist isn't important, there are some ways you can make the pattern more efficient by rearranging redunant parts of the string. For example, you'll already be pulling [N V]s from the vowel strings, so you don't need all those [N]s throughout every CV string, so you can make a new vowel pattern than excludes [N] and attach the N-CV segment to the end of the string, resulting in 8-mora.

This will result in a shorter reclist length overall because now there's one less string for each consonant:

ka-ka-ki-ka-ku-ka-n-ka
ki-ki-ku-ki-ke-ki-n-ki
ku-ku-ke-ku-ko-ku-n-ku
ke-ke-ko-ke-ka-ke-n-ke
ko-ko-ka-ko-ki-ko-n-ko

You can use this strategy for any CVs you want to omitt; just repattern the string without them, and then attach them at the end. For example, you can do the same the [Cy] strings to exclude [Cyi] by creating a pattern that excludes [i ] in addition to [N] and attaching both to the end.

You can split up the string into a new line at any point as long as you remember to double the mora at the splitting point so that you don't leave out that [V CV]. For example, if you're splitting ka-ka-ki-ka-ku-ka-ke after the 3rd [ka], remember to add another [ka] to the new string, like ka-ka-ki-ka and ka-ku-ka-ke.

If this results in uneven string lengths, you can try to combine the "leftover" CVs into their own strings, but it doesn't always math out well. This is why 5-mora is usually the next most common length after 7-mora, because the strings are a bit shorter but still result in even lengths. I've yet to find a good pattern for 6-mora but I won't say it's impossible.

Splitting strings can drastically increase the amount of strings you have to record overall, though, because it adds +however many additional strings in the pattern to every consonant. You can't really have it both ways of making a list with both shorter strings and also fewer strings.

These are the patterns I used for writing my own VCV lists. If you tried one of them and it gave you an error, do you mind telling me which list it was? It's possible there's just a typo in either the list or the base oto that I can correct for future users.

The other option too is to do a lite list where you don't include every possible [V CV] pair, instead just including regular [CV]s pulled from the middle of the string in addition to some [V CV]s, though I'm not sure how well that plays with SetParam's oto generation. Personally I don't like this type of reclist so I haven't written any, but they do exist. The other other option is to do CVVC; those lists are much shorter than VCV even with shorter string lengths, and can sound just as smooth.

I forgotten which one, because I'm focused a lot on your Efficiency VCV and 7 mora ones. I didn't touched the 5 mora ones, so it's neither the Efficiency or 7 mora ones that had missing sample error on SetParam.

It's the Kana ones.