OTOEARTH

OTOEARTH (oto/inc) builds the datasets, models, and benchmarks that voice AI labs use to ship full-duplex speech-to-speech systems. The company produces two-channel, time-aligned conversational audio datasets recorded with channel separation at source rather than diarized after the fact, covering diverse speakers across accents, ages, and backgrounds. Beyond datasets, oto develops speaker diarization and speech enhancement models tuned for real overlapping conversation, as well as evaluation benchmarks for human-like speech-to-speech behavior co-developed with industry and academic partners. Their datasets have reached top-3 trending on HuggingFace and are downloaded by major AI labs. Backed by the Inception Program accelerator.