Releases | Docs

Healthcare and insurance Q&A over medical records, labs, checkups, and policy documents—summarize, interpret, and surface key points. Multi-turn dialogue keeps context for follow-up questions.

Recognition accuracy in complex noise and dialect scenarios is the first in the industry to exceed 90%, with multilingual and full-system dialect coverage, structured long-audio transcription, and fast, reliable deployment.

Breakthroughs in semantic understanding and emotion for natural, expressive speech

Clone voice timbre in seconds from a single-sentence sample, with emotion transfer and cross-lingual Chinese-English synthesis, enabling rapid accumulation of exclusive brand/character voice assets.

Compact yet highly accurate, with SOTA performance on tables, formulas, and complex layouts; supports all formats and multiple languages, outputs native structured results, and enables fast, deployment-friendly inference.

Supports document classification, general /Schema-oriented extraction, and coordinate-based trace-back auditing, showing stronger advantages in document-heavy, business-critical scenarios such as healthcare.