Datasets:
Sundanese Spontaneous Conversation — Free Sample
Unscripted two-speaker Sundanese (su-ID) conversation recorded in West Java, Indonesia. This is a free 30-minute sample of a larger commercially licensable corpus.
Why this exists
Sundanese has approximately 40 million native speakers, yet spontaneous conversational data is almost absent:
| Resource | Type | Volume |
|---|---|---|
| Common Voice Spontaneous 4.0 | Spontaneous, 2 speakers | 0.62 h |
| OpenSLR SLR36 / SLR44 | Read speech | — |
| This corpus (full) | Unscripted conversation | 50+ h |
Regional speech models such as MERaLiON and SEA-LION cover Singapore and mainland Southeast Asian languages. Sundanese is not represented.
What is in this sample
- 30 minutes, unscripted two-speaker conversation
- Priangan dialect, Garut region
- MP3 192 kbps, 48 kHz, downsampled copies available on request
- Measured channel isolation 7.9 dB, SNR 17.4 dB (Grade B)
- Natural overlap, laughter, and Sundanese-Indonesian code-switching
- Per-speaker metadata: gender, age range, dialect, region
Quality grading
All material is measured before release. Isolation is computed by locating passages where only one speaker is active and measuring how much quieter the other channel is; the worst-case figure sets the grade.
- Grade A — above 18 dB isolation, external microphone per speaker
- Grade B — 6 to 18 dB, dominant speaker per channel with cross-talk
- Grade C — below 6 dB, or mixed channels
This sample is Grade B. The production corpus is Grade A.
Consent and licensing
Every participant signed a bilingual written consent before recording, read aloud and explained in Sundanese or Indonesian. It covers commercial AI training, sublicensing, and cross-border transfer under Indonesian Law No. 27/2022. Consent records are held in two parts so that buyers can verify consent without receiving personal identifiers.
This sample is released under CC BY-NC 4.0 — research and evaluation only. Commercial AI training is not permitted under this licence.
Full corpus — commercial licensing
The production corpus is Grade A, recorded with an external microphone per speaker, in sessions of 2 to 6 speakers. Available with transcripts and Sundanese-to-Indonesian and Indonesian-to-English translation. Non-exclusive, term-exclusive, and full buyout terms available.
Contact: torino@indospeech.com · indospeech.com
Citation
If you use this sample in research, please cite:
- Downloads last month
- 42