Amphion Releases
102 releases of open-mmlab/Amphion
- v0.5.1
Nov 15, 2025
- v0.5.0
Introduces DSQ (DeepSeek Quantization) to reduce VRAM usage for DeepSeek-OCR and PaddleOCR-VL models.
Nov 14, 2025
- v0.4.1
Fixes a critical memory crash on Blackwell GPUs caused by RoPE cache expansion.
Nov 10, 2025
- v0.4.0
Adds PaddleOCR-VL backend, lazy loading, and unified sampling controls.
Nov 9, 2025
- v1.24.0
Consolidates binaries into a single `deskflow-core` executable and improves logging.
Sep 11, 2025
- v3.8.3
A maintenance release dedicated to resolving TypeScript compilation issues following previous releases.
Aug 27, 2025
- v3.8.2
A minor release with no specific changes detailed in the provided changelog text.
Aug 27, 2025
- v3.8.1
A minor release with no specific changes detailed in the provided changelog text.
Aug 27, 2025
- v3.8.0
A feature-rich release introducing new build tooling support, removing deprecated polyfills, adding new examples, and fixing various playback and UI bugs.
Aug 27, 2025
- v2.3.0
Adds custom model paths and enforces mandatory model file validation.
Aug 17, 2025
- v2.2.1
Minor maintenance release focusing on internal CI stability.
Aug 14, 2025
- v2.2.0
Project is now available on PyPI for simplified installation.
Aug 14, 2025
- v2.1.1
Adds stdin support for piping text and improves packaging.
Aug 14, 2025
- v0.5.2Release v0.5.2
This release primarily updates the default model to GPT-4.1-mini, which offers quality improvements, though users should be aware of behavioral differences compared to the previous model.
Jul 31, 2025
- v1.23.0
Focuses on code cleanup, Qt updates, and UI improvements.
Jul 23, 2025
- v0.5.1-alphaRelease v0.5.1-alpha (including paper resources)
This alpha release includes the TinyTroupe paper, introduces action correction mechanisms for agents, improves simulation performance with parallel execution, and adds tools for controlled experiments and monitoring.
Jul 16, 2025
- 2.3
This release adds full-text search, fiber goal support, and enhances Docker image capabilities to support exercise videos and fixtures. It also introduces ingredient sync management commands and Prometheus metrics.
Apr 5, 2025
- v0.4.0-alpha
This alpha release introduces deeper persona specifications, JSON-based persona definitions, and the concept of fragments for reuse, with a note that API changes may break existing programs.
Jan 29, 2025
- v0.1.1-alpha
This alpha release adds significant new features including DiffWave vocoder, HifiTTS data preprocessing, Docker installation support, and visualization capabilities. It also includes multiple bug fixes and improvements to speaker similarity evaluation with Resemblyzer and WavLM.
Feb 23, 2024
- v0.1.0
The first stable release of Amphion, a multimodal audio generation platform. This release includes core implementations of VALL-E, VITS, VitsSVC, and NaturalSpeech2, along with improved G2P processing and training recipes.
Feb 23, 2024