exo Releases

Added blend shape based lip-sync animation support to the speakAudio function and fixed a bug regarding number-to-word conversion.

Mar 3, 2025

This is the initial release of diffusion-rs, a Rust library for diffusion models. It provides optimized kernels for SDPA and LayerNorm, supports loading from DDUF files, and includes a CLI tool and Python bindings.

Jan 7, 2025

Introduced dynamic bones for custom models, improved the morph target animation system, and added audio-based head movement features.

Nov 18, 2024

Migrated preview commands to 'tea' (likely a CLI tool) and refactored the TUI code to be shared, along with updating the encryption method.

Oct 24, 2024

Integrated Tea into the launch and deploy commands, working towards a full migration of the UI to a BubbleTea-based TUI.

Oct 19, 2024

Minimal release notes: "## Changelog * 3ecc864 Fix: issues with env file encryption leading to failed launch/depoloy"

Oct 13, 2024

Fixed an idempotency issue in the init command, allowing users to run it multiple times safely on the same server.

Oct 9, 2024

Oct 7, 2024

This release focuses on packaging, testing infrastructure, and documentation improvements. It introduces official packages on PyPI and DockerHub, adopts Poetry for dependency management, and deprecates legacy Python versions and Tor support.

Jul 8, 2024

This release focuses on performance optimization and flexible deployment, introducing GPU runtime support via nv-triton and CPU quantization solutions. It also adds streaming inference capabilities and various new domain-specific models.

Mar 16, 2023

This release enhances model deployment flexibility by adding export capabilities for Paraformer models into ONNX and TorchScript formats. It also introduces standalone ONNXRuntime deployment and gRPC service support, alongside new language models.

Feb 20, 2023

This release introduces long-form speech recognition with the Paraformer-large-long model and standalone VAD and Punctuation models. It also migrates core models from TensorFlow to PyTorch and adds the Data2vec pretraining model.

Jan 16, 2023

This is the initial release introducing core ASR capabilities, including batch decoding support for Paraformer models and the introduction of UniASR models and Transformer/Conformer architectures.

Dec 10, 2022

Jun 9, 2021

Sep 27, 2018