Datasetpublic

aoxo/t2a-mommy

t2a-mommy Female-voice ASMR corpus for the text2asmr project. Previously published as aoxo/audios2. Companion repos: aoxo/t2a-daddy (male voice), aoxo/t2a-audios-v1 (the original v1 corpus). Layout path what <creator>/<title>.m4a source audio, 48 kHz AAC, one folder per creator <creator>/<title>.json word-level Whisper large-v3 alignment ([] = skipped: near-silent or undecodable) labels/qwen3omni.jsonl non-speech ontology labels for gap clips… See the full description on the dataset page: https://huggingface.co/datasets/aoxo/t2a-mommy.

sourceHugging Faceapache-2.0updated 12m agoView on Hugging Face

0likes14kdownloads

Dataset Card

t2a-mommy

Female-voice ASMR corpus for the text2asmr project. Previously published as aoxo/audios2.

Companion repos: `aoxo/t2a-daddy` (male voice), `aoxo/t2a-audios-v1` (the original v1 corpus).

Layout

path	what
`<creator>/<title>.m4a`	source audio, 48 kHz AAC, one folder per creator
`<creator>/<title>.json`	word-level Whisper `large-v3` alignment (`[]` = skipped: near-silent or undecodable)
`labels/qwen3omni.jsonl`	non-speech ontology labels for gap clips, `Qwen3-Omni-30B-A3B-Instruct` (1.93 M rows)
`labels/qwen3omni_expansion.jsonl`	same, for creators added in the 2026-09 expansion
`labels/clap_gate_index.jsonl`	gap-clip index (uid → source, start, duration)
`labels/gemini_pro_pilot.jsonl`	8.6 k audio-only Gemini-3.1-Pro labels — the ground-truth/eval set
`labels/legacy_gemini/`	superseded Gemini-flash label ledgers, kept for provenance. Not learnable from audio — see the repo README before using them
`labels/pending_candidates.jsonl`	gap clips cut but not yet labeled

Label rows are {uid, raw, label, labeler}, where uid is <source>.m4a_<start_ms>.

Ontology: whispering, normal speech, breathing, mouth sounds, moaning, kissing, silence, plus the physical tail (tapping, scratching, crinkling, brushing, liquid, page turning, other sound). A label covers at most 3–4 s of audio.

Audio is sourced from soundgasm; each creator keeps their own folder, so attribution and creator-split evaluation are preserved.