nnsvs: https://github.com/r9y9/nnsvs/pull/81

ustは https://www.nicovideo.jp/watch/sm39319737 を使わせていただきました。ust -> musicxml に変換して、nnsvsで合成しました。

モデル情報: 
Statistical parametric SVS (sampling rate: 48000)
Time-lag model: {
    "_target_": "nnsvs.model.MDN",
    "in_dim": 344,
    "out_dim": 1,
    "hidden_dim": 32,
    "num_layers": 3,
    "dropout": 0.5,
    "num_gaussians": 1
}
Duration model: {
    "_target_": "nnsvs.model.MDN",
    "in_dim": 344,
    "out_dim": 1,
    "hidden_dim": 256,
    "num_layers": 3,
    "dropout": 0.5,
    "num_gaussians": 1
}
Acoustic model: {
    "_target_": "nnsvs.model.ResF0Conv1dResnet",
    "in_dim": 348,
    "out_dim": 206,
    "hidden_dim": 128,
    "num_layers": 6,
    "in_lf0_idx": 300,
    "out_lf0_idx": 180,
    "in_lf0_min": 5.162578582763672,
    "in_lf0_max": 6.491110801696777,
    "out_lf0_mean": 5.892984643557185,
    "out_lf0_scale": 0.2296959673213171
}

2022/03/21 Kiritan test SVS Roshin Yuukai

This song is made from Oniku Kurumi's singing voice database(onikuru.info/db-download/).
Time-lag model: {
"_target_": "nnsvs.model.MDNv2",
"in_dim": 337,
"out_dim": 1,
"hidden_dim": 32,
"dropout": 0.5,
"num_layers": 3,
"num_gaussians": 1,
"init_type": "kaiming_normal"
}
Duration model: {
"_target_": "nnsvs.model.VariancePredictor",
"in_dim": 337,
"out_dim": 1,
"hidden_dim": 256,
"num_layers": 5,
"kernel_size": 5,
"dropout": 0.5,
"use_mdn": true,
"num_gaussians": 4,
"init_type": "kaiming_normal"
}
Acoustic model: {
    "_target_": "nnsvs.model.RMDNv2",
    "in_dim": 341,
    "out_dim": 206,
    "hidden_dim": 256,
    "num_layers": 2,
    "bidirectional": true,
    "dropout": 0.5,
    "num_gaussians": 4,
    "dim_wise": true,
    "init_type": "kaiming_normal"
}
Post-filter model: {
"_target_": "nnsvs.postfilters.MultistreamPostFilter",
"stream_sizes": [
60,
1,
1,
5,
2,
1
],
"mgc_postfilter": {
"_target_": "nnsvs.postfilters.Conv2dPostFilter",
"in_dim": 58,
"channels": 64,
"kernel_size": [
5,
5
],
"init_type": "kaiming_normal",
"noise_type": "frame_wise",
"noise_scale": 1.0,
"smoothing_width": 100
},
"mgc_offset": 2,
"bap_offset": 0,
"bap_postfilter": null,
"lf0_postfilter": null
}
Vocoder: world

nnsvs_oniku_kurumi_utagoe_db_svs_world_rmdnv2_inaba_kumori_tokoshizume_lead_w_mgc_postfilter

Luna Ascension (UK Hardcore Remix)

Original Composer : flashygoodness 
Jacket Illustration : charle_magnette
Remix : Foilverb x Nichico Twelve
Thanks for flashygoodness & Tower of Heaven
Askiisoft askiisoft.com/games/tower-of-heaven/

Luna Ascension Continue Remix

nnsvs test

idk when i'll finish it LMAO
this is the 10th time i posted too

vb by @kei_amano

natsume_singing nnsvs test2

御丹宮くるみ歌声データべース と NNSVSで童謡の「雪」です。

雪 (NNSVS / oniku-kurumi v1.0)(simpleqst3 285 dim)

I made a cover with an AI-based voice synthesis called NNSVS.
I did a lot of pitch correction on the vocal data, tho.

Singing voice database creation by Amano Kei, 
Raw singing data by Sota Kirino(Twitter:@K_Sdcm)
NNSVS by Ryuichi Yamamoto(Twitter:@r9y9)
Composite recipe by Shirani(Twitter:@taroushirani)
Special Thanks by CrazY(Twitter:@crazy_toho)
inst by 生音風カラオケ(https://youtu.be/GXlq5hbTW4Q)

【AI NatsumeYuuri】WanteD! WanteD!【NNSVS Cover】

moresamplerとオリジナルの原音設定を混合させたタイプ。
先頭音以外かなり良くなってる気がする。

雪 (NNSVS / utau-namineritsu-sinigng v2.2.0)(simpleqst4 285 dim)

hhh she sounds so good
ignnore the clipping issues i'll fix those in vshifter

Lagtrain nnsvs wip

beep beep
ok technically the part before this was tuned but

ust by rewoundclockwork

nnsvs_oniku_kurumi_utagoe_db_svs_world_rmdnv2_inaba_kumori_tokoshizume_lead_w_mgc_postfilter

taroushirani

Related tracks