Datasets
Start training sooner with our off-the-shelf Khmer data, or use it to benchmark before a custom project. Every dataset is consented, checked by linguists and licensed for commercial use.
6 datasets
Natural two-person phone conversations on everyday topics like banking, shopping, health and travel, with speaker-separated transcripts.
Phonetically balanced sentences read by speakers from Phnom Penh, Battambang, Siem Reap, Takeo and more, recorded on everyday phones.
Short commands for apps, cars and smart devices, such as "call", "pay", "open", numbers and dates, in quiet and noisy places.
Studio recordings from professional male and female Khmer voice talent, in neutral, news and friendly styles, with aligned scripts. Every voice comes with a signed voice-licence consent.
News, social media and official text with linguist-checked word boundaries and labelled names, places, organisations, dates and amounts.
Human-translated sentence pairs across banking, health, technology and everyday chat, including Romanized Khmer.
Any dataset can be extended or customised with other accents, domains, devices or labels. Or we build one from scratch for you.
Request a custom projectNo datasets match your search. Try another word, or ask us for a custom dataset.
Volumes are current collection targets and grow every month. Request a sample for the latest numbers and a full datasheet.
Request a sample
Tell us which datasets you want to evaluate. We'll reply within one business day with a sample and a datasheet.