Aisar is a project on Chuvash speech recognition: datasets, three Whisper models, and an automatic-labeling pipeline. Everything was built and released by one person.

Model weights and datasets are on HuggingFace, under the SpeechCollector organization. For any questions, ideas or feedback, reach out on Telegram — I'd be glad to hear from you.