datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
swamiji-tts-rag10-cfgskip-ab
CFG-skip on the 10 RAG answers — b and d against baseline
⚠️ SYNTHETIC AUDIO — PRIVATE
Every clip is generated by a text-to-speech model. Nothing here is a
recording of any person.
These texts are actual RAG answer content about Sri Ganapathi Sachidananda
Swamiji, spoken in a synthesised version of his voice. That combination is easy
to mistake for a genuine recording, which is why this repo is private.
The same 10 RAG-answer texts, same seeds, three guidance… See the full description on the dataset page: https://huggingface.co/datasets/sw-voice/swamiji-tts-rag10-cfgskip-ab.ragas_trainscarnatic-ragasTamilVoiceCorpus
Tamil Conversational ASR Dataset
This is a dataset for Automatic Speech Recognition (ASR) focused on conversational Tamil, collected from various public sources on the web. Each sample is a short audio clip (averaging 10 seconds) paired with its corresponding transcription.
Dataset Summary
Language: Tamil (ta)
Domain: Conversational speech
Average Duration per Clip: ~10 seconds
Format: Audio (.wav) + text
Sample Rate: 16kHz recommended
Total Examples:
PureVox:
Train: 12… See the full description on the dataset page: https://huggingface.co/datasets/ragunath-ravi/TamilVoiceCorpus.swamiji-tts-rag10-optimised
The 10 RAG-answer samples, regenerated with the current optimised config
⚠️ SYNTHETIC AUDIO — PRIVATE
Every clip is generated by a text-to-speech model. Nothing here is a
recording of any person.
Unlike the public benchmark sets, these texts are actual RAG answer content
about Sri Ganapathi Sachidananda Swamiji, spoken in a synthesised version of
his voice. That combination is easy to mistake for a genuine recording, which
is why this repo is private. Do not make… See the full description on the dataset page: https://huggingface.co/datasets/sw-voice/swamiji-tts-rag10-optimised.sust_banspeech-orpheus-formattedsynth-ragswamiji-rag10-3models
RAG-10 answers, rendered by three fine-tunes
The project's ten standing validation texts — the same ones every earlier arm was
judged on, from
swamiji-rag-answers-10ep-v2-copy-text
— rendered by three models, two independent draws each, side by side on one row.
column prefix
model
trained on
current_*
sw-voice/swamiji-voxcpm2-ft-lr0.0001-10ep-deva
original 958-clip corpus, silence not normalised
src0_*
sw-voice/swamiji-voxcpm2-ft-src0-lr0.0001-10ep-deva… See the full description on the dataset page: https://huggingface.co/datasets/sw-voice/swamiji-rag10-3models.sample_audio_3sample_audiosample_audio_4swamiji-rag-10ep-samplesswamiji-rag-answers-10ep-v2-copy-textsample_audio_1ivan-chygrynau-na-pylnai-daroze-uladzimir-ragautsou
На пыльнай дарозе
Metadata
Author: Іван Чыгрынаў
Title: На пыльнай дарозе
Narrator: Уладзімір Рагаўцоў
Source Group: Аўдыёкнігі
Source:
Notes
The original audio files are preserved as-is:
no conversion;
no re-encoding;
no filename changes inside each split folder, except removing one common top-level archive folder when present.
To avoid Hugging Face Dataset Viewer scan-size errors, the dataset is split into smaller folders.
Target maximum split… See the full description on the dataset page: https://huggingface.co/datasets/archivartaunik/ivan-chygrynau-na-pylnai-daroze-uladzimir-ragautsou.swamiji-rag-answers-10epswamiji-rag-answers-10ep-v2-copytelugu-asr-smallzsy12345-filtered3eou_AudioTextswamiji-rag-answers-ttsswamiji-rag-answers-10ep-v2lucaszsy12345reks-staut-krylaty-revalver-uladzimir-ragautsou
Крылаты рэвальвер
Metadata
Author: Рэкс Стаўт
Title: Крылаты рэвальвер
Narrator: Уладзімір Рагаўцоў
Source Group: Аўдыёкнігі
Source:
Notes
The original audio files are preserved as-is:
no conversion;
no re-encoding;
no filename changes inside each split folder, except removing one common top-level archive folder when present.
To avoid Hugging Face Dataset Viewer scan-size errors, the dataset is split into smaller folders.
Target maximum split… See the full description on the dataset page: https://huggingface.co/datasets/archivartaunik/reks-staut-krylaty-revalver-uladzimir-ragautsou.transcription-datasetswamiji-base-rag-answersmulti-genre-musicembedded_world_2026_rag_tts_clonedembedded_world_2026_rag_tts_INC_names