mpfb2 Releases

Introduces configurable Sogou web search, refactors inbound context routing, and improves the CLI experience with a structured terminal UI. Also adds Android build support and various UI improvements.

Apr 22, 2026

A major update adding support for over 20 new social media platforms including YouTube, Wikipedia, Threads, Chess.com, Roblox, and Habr. The release also introduces plugins support, migrates the project configuration to pyproject.toml, and changes the license to MIT.

Apr 9, 2026

This release focuses on enhancing hooks with documentation and a respond action, introducing support for isolation, and implementing a short-term memory engine (LCM). It also includes the addition of a memory benchmark tool and various fixes regarding gateway stability and WebUI connectivity.

Apr 8, 2026

Major version bump triggered by a breaking change, managed via semantic release.

Mar 12, 2026

Maintenance release focused on removing legacy code and bumping the version.

Mar 12, 2026

Updates the Audio2Face-3D sample client configuration to version 2.0, including new diffusion models and tongue blendshapes.

Mar 11, 2026

This is a stable release of MPFB2. Detailed release notes available at the official release page.

Feb 25, 2026

Server update introducing support for experimental message types.

Feb 13, 2026

Jun 16, 2025

Apr 10, 2025

The first stable release of MPFB (MakeHuman for Blender). This marks the transition from alpha/beta to a production-ready version. Installation is now available via the Blender extension platform.

Jan 28, 2025

This release focuses on dependency upgrades, specifically bumping the pytest suite from version 7.x to 8.x and updating related testing tools. It also introduces new SOWEL techniques, fixes a Gravatar username extraction bug, and updates the PyPI publishing workflow.

Dec 12, 2024

MPFB2 beta 1 release. This represents the first beta version of the MakeHuman for Blender 2.0 series, indicating the software is feature-complete and entering testing phase.

Jul 23, 2024

This release updates dependencies and fixes tests for the Socid-extractor tool. It includes updates to pytest and beautifulsoup4 versions and repairs most of the test suite to ensure stability.

Oct 11, 2023

MPFB2 alpha 3 release. Third alpha iteration of the MakeHuman for Blender 2.0 rewrite, focusing on early development and feature prototyping.

Jul 13, 2023

This release introduces significant performance improvements and new deployment options for the FunASR library, including GPU runtime support via nv-triton, CPU quantization, and C++ gRPC services. It adds several new models for streaming speech recognition, punctuation, speaker diarization, and domain-specific tasks like finance and audio-visual processing.

Mar 16, 2023

This version enhances the deployment flexibility of FunASR by introducing features to export models to ONNX and TorchScript formats, allowing deployment without ModelScope. It also adds an ONNXRuntime runtime for CPU inference and a gRPC service capability. Additionally, several new models are released, including contextual hotword support, multi-channel meeting transcription, and various language and dialect support.

Feb 20, 2023

This release introduces a comprehensive long-form audio model that integrates VAD, ASR, punctuation, and timestamp prediction. It adds standalone models for VAD and punctuation, along with a lightweight Paraformer-Tiny model for command words. Notably, the library migrates from TensorFlow to PyTorch for better customization support and adds new audio input type support.

Jan 16, 2023

This is the initial release of the FunASR library, establishing its core capabilities. It introduces batch decoding support for Paraformer models, adds new UniASR models, and includes Transformer and Conformer architectures. The release focuses on making inference and fine-tuning of models within ModelScope more convenient.

Dec 10, 2022

Sep 15, 2022