AUDIO
COMMERCIAL
LABELED
Mixing a studio clip with a noise file does not reproduce the Lombard effect or the way people lean in and slow down in a wet market, which is one reason ASR trained with synthetic noise still fails outdoors. Contributors go somewhere loud, read the phrase on screen, and say what the noise was; quiet rooms, added noise and drowned-out speech are refused. Ten clips per contributor at most, from the streets, markets and transit of the cities where a global base lives. Phrases come from a 600-item bank, so the same text recurs across noise conditions, and each clip carries the prompt and the noise description.
Volume
10,000+
Clips
Contributors
2,000+
Regions
100+ countries
Updated
What’s in the collection and how it’s delivered.
Data modality
Audio
Annotated
Labeled (prompt text, noise description)
Exclusivity
Non-exclusive rights
Capture tool
Phone microphone (in app)
Use cases
Noise-robust ASR evaluation and fine-tuning, speech enhancement with real noise, acoustic scene classification
Format
M4A
Resolution
M4A (AAC) at the phone's native sample rate
Total volume
10,000+ clips (approx. 3 GB)
Regions
100+ countries
Purity level
99%+ passed QA review
Where this data comes from and how it’s licensed.
Every file comes straight from the contributor who provided it and is rights-cleared before it’s listed, so your license traces back to the source.
Contributor
Provides the data and uploads it straight to Kled.
Kled
Clears the rights before anything is listed.
Your team
Licenses it under the terms below.
License terms
Non-exclusive commercial training license.
Consent
Each contributor opted in to this use when they uploaded.
Specs reflect the current collection and may change as new data is added. Final terms are set in your license agreement.
