Orchestrates low-latency, multi-stage pipelines for omni and multimodal models by running each stage with its own scheduler and using zero-copy shared memory for tensor transfer. Emphasizes per-stage bottleneck tuning and OpenAI-compatible streaming endpoints, suitable for TTS and multimodal serving.
[next-mdx-remote-client] error compiling MDX: Unexpected character `-` (U+002D) before name, expected a character that can start a name, such as a letter, `$`, or `_` More information: https://mdxjs.com/docs/troubleshooting-mdx
Zero-shot text-to-speech across 600+ languages with short-reference voice cloning and attribute-based voice design. Uses a diffusion-style discrete non-autoregressive architecture mapping text to multi-codebook acoustic tokens, supports phoneme/pinyin control and very low RTF.
Provides text-to-speech synthesis and one-shot voice cloning via compact ONNX models and a larger KittenTTS 2 speech model. Offers CPU-first inference (25–506 MB models), 47 built-in voices, multilingual output, expression tags, streaming, and a Python SDK.
Generates directed, multi-character, long-form speech from text with natural-language directions and voice cloning. Built on a 35B MoE backbone (3B active per token) with a codebook multi-token-prediction decoder to enable multi-turn revision, infilling edits, streaming, and expressive control.
