Stars
System-One-controlled full-duplex voice interaction: turn-taking as plain-language decisions judged by Jev.
DuplexJev-32B and DuplexJev-4B: speech decision models for full-duplex voice agents. Typed single-token readout from a frozen speech LLM, served with vLLM. Try it: api.adventists.cn/duplexjev
M5-TTS is a lightweight 50M-parameter TTS language model supporting real-time CPU inference, multilingual and multi-dialect synthesis, zero-shot voice cloning, and custom voices.
SearchJev: a fast and calibrated System-1 model for search agents (training code)
An open, programmable decision layer for models and compute.
Open decision models from 0.8B to 27B: a calibrated probability for every option in one pass, and a short thought only when unsure. Engine for vLLM, llama.cpp and MLX.
Residual emotion steering for emotional TTS generation. EmoRES-TTS decomposes a mixed-emotion steering vector into a shared axis and an emotion-carrying residual and weight the two independently.
Open-source realtime voice agent server in Go with WebRTC (WHIP), barge-in, streaming STT/LLM/TTS pipelines, plugin system, multi-language SDKs, SIP telephony, ESP32 support & fully local mode.
X2Streaming-ASR is a streaming Automatic Speech Recognition (ASR) system characterized by ultra-low latency and high accuracy.
Confucius4-R2T2: A Low Latency and High Accuracy Real-Time Speech Recognition Model
X-ASR is a series of automatic speech recognition models based on the icefall framework, focusing on streaming ASR and low-latency deployment.
Lean streaming engine + realtime server for Edge0's Audio8 ASR Infinite: same WER, 12 ms per 80 ms step in 3 GB (int4), independent streams, semantic end-of-turn
A nano replica of Jev: parallel decisions, dynamic candidates, and an end-to-end training pipeline.
Open-source end-to-end omni interaction agent for natural full-duplex voice collaboration, continuous audio-visual perception, and asynchronous long-horizon agentic execution.
This repository contains the official implementation and pretrained weights for the paper "ReDimNet2: Scaling Speaker Verification via Time-Pooled Dimension Reshaping".
A 0.9B model for long-form transcription in 50+ languages with speaker diarization, timestamps, and acoustic event awareness
Frame-Synchronous Dual-Head Modeling for Joint Streaming ASR and Turn State Prediction
Official repository of the paper "On-Policy Self-Distillation for Multi-Dialect ASR: Mastering Dialects, Retaining Mandarin".
Standalone causal streaming runtime for Qwen3-ASR
AmphionASR: Personalized Context-Aware Speech Recognition
A vector index built on TurboQuant, written in Rust with Python bindings
Native full-duplex speech dialogue inference for BayLing-Duplex.