Text-to-speech models, audio codecs,
training, inference, and evaluation.

_{text-to-speech  ·  audio codecs  ·  open source}

• VyvoTTS — LLM-based TTS training with SNAC and Mimi audio tokens.

• Nar TTS — Speech-token TTS using causal language models and the Mimi audio codec.

• Echo DACVAE — Experimental TTS with EchoDiT, flow matching, and DACVAE audio latents.

• NAR-VAE — Experimental non-autoregressive TTS with flow matching and DACVAE latents.

• VoiceHub — Unified inference for speech synthesis, recognition, and voice activity detection.

• VoicePlus — Native PyTorch inference pipelines for TTS, ASR, VAD, and audio codecs.

• WhisperPlus — Whisper transcription, speaker diarization, summarization, and video captioning.

• Voice Agent Next — Python runtime for speech-to-speech models and STT–LLM–TTS pipelines.

• Fast-DACVAE — DACVAE inference optimization with PyTorch compilation and CUDA graphs.

• Fast-Mimi — Mimi codec inference optimization with fused kernels and CUDA graphs.

• Fast-MOSS-Tokenizer — MOSS Audio Tokenizer v2 inference with CUDA Graph streaming and fused kernels.

• VoiceHub Arena — TTS evaluation with intelligibility metrics, performance reports, and audio comparisons.

• Tracking: Norfair · OC-SORT · ByteTrack · StrongSORT · SORT
• Detection & segmentation: MetaSeg · YOLOv7 · YOLOv6 · YOLOX
• Image restoration: CodeFormer · BSRGAN

• Speech & generative AI: Insanely Fast Whisper · Diffusers #3590 · Diffusers #3586 · ComfyUI
• ML tools: Evaluate · PyTorch Examples · LlamaIndex · AutoLLM
• Computer vision: SAHI #486 · SAHI #322 · SAHI #501 · Norfair · StrongSORT · YOLOv7 · YOLOv6 · Kornia