Three new models for speech recognition and text-to-speech are now available in Amazon SageMaker JumpStart

Amazon Web Services has added three new AI models to its SageMaker JumpStart platform, bringing advanced speech recognition and text-to-speech capabilities to enterprise customers. The Qwen3 model family includes two text-to-speech variants and one automatic speech recognition model, all supporting multiple languages and designed for different enterprise use cases. The Qwen3-TTS-12Hz-1.7B-CustomVoice model offers instruction-driven control over voice characteristics like emotion and prosody across 10 languages, while the Base variant provides rapid 3-second voice cloning capabilities from audio input. The speech recognition component, Qwen3-ASR-1.7B, supports 52 languages and dialects with enhanced accuracy in challenging acoustic environments, making it suitable for transcription services and real-time captioning applications. All three models can be deployed directly through SageMaker Studio with minimal setup, allowing enterprises to integrate voice capabilities into their applications without extensive infrastructure development. The models are optimized for both streaming and offline processing scenarios, addressing common enterprise requirements for voice-powered applications and multilingual customer support systems.