Datasetpublic

ghananlpcommunity/voxcpm-ghana-latents

VoxCPM Ghana — Precomputed AudioVAE Latents The exact training-ready data used to fine-tune ghananlpcommunity/voxcpm-ghana: precomputed VoxCPM-0.5B AudioVAE latents (16 kHz) for 42 Ghanaian languages + filtered Ghanaian English, with language-tagged transcripts. Drop-in for VoxCPM fine-tuning — no audio decoding or VAE encoding needed at train time. 1,756,157 clips · ~3,400 h · 16 kHz 42 Ghanaian languages (incl. Twi split: twi-asante, twi-akuapem) + en AudioVAE from… See the full description on the dataset page: https://huggingface.co/datasets/ghananlpcommunity/voxcpm-ghana-latents.

sourceHugging Facecc-by-4.0updated 3mo agoView on Hugging Face

0likes713downloads

Dataset Card

VoxCPM Ghana — Precomputed AudioVAE Latents

The exact training-ready data used to fine-tune `ghananlpcommunity/voxcpm-ghana`: precomputed VoxCPM-0.5B AudioVAE latents (16 kHz) for 42 Ghanaian languages + filtered Ghanaian English, with language-tagged transcripts. Drop-in for VoxCPM fine-tuning — no audio decoding or VAE encoding needed at train time.

—1,756,157 clips · ~3,400 h · 16 kHz
—42 Ghanaian languages (incl. Twi split: twi-asante, twi-akuapem) + en
—AudioVAE from `openbmb/VoxCPM-0.5B`: 64-dim, hop 640 (~25 fps)

Format (parquet shards, `CODE__*.parquet`)

column	type	notes
`feat`	binary	fp16 AudioVAE latent, shape `[feat_t, 64]`. Reconstruct: `np.frombuffer(feat, np.float16).reshape(feat_t, 64)`
`feat_t`	int32	number of latent frames (~25 per second)
`text`	string	transcript prefixed with the language tag `<	lang:CODE	> `
`dataset_id`	int32	per-language id (table below)
`split`	string	`train` / `dev` (balanced: ~40 dev clips per language)

Language tags

Each transcript starts with <|lang:CODE|> , where CODE is the ISO-639-3 code (e.g. ewe, dag, hau, fat), with Twi split into twi-asante / twi-akuapem, and en for English. The model learns the tag as text (VoxCPM is tokenizer-free).

dataset_id map

code	id
`acd`	0
`ada`	1
`akp`	2
`any`	3
`avn`	4
`bib`	5
`bim`	6
`biv`	7
`bov`	8
`bud`	9
`bwu`	10
`dag`	11
`dga`	12
`en`	13
`ewe`	14
`fat`	15
`ffm`	16
`gjn`	17
`gur`	18
`hau`	19
`kbp`	20
`kdh`	21
`kma`	22
`kus`	23
`lef`	24
`lip`	25
`maw`	26
`mzw`	27
`naw`	28
`ncu`	29
`nko`	30
`ntr`	31
`nzi`	32
`sfw`	33
`sig`	34
`sil`	35
`snw`	36
`tpm`	37
`twi-akuapem`	38
`twi-asante`	39
`vag`	40
`xon`	41
`xsm`	42

Source

Derived from `ghananlpcommunity/ghana-speech` and `ghananlpcommunity/ghana-english-tts-filtered`.

VoxCPM Ghana — Precomputed AudioVAE Latents

Format (parquet shards, CODE__*.parquet)

Language tags

dataset_id map

Source

Format (parquet shards, `CODE__*.parquet`)