pocket-tts Releases

This release focuses on optimizing training efficiency, adding new language support (Hindi and Korean), and improving the user experience with CPU-only installation documentation and CLI enhancements. It also includes significant refactoring of the core model architecture and dependency updates.

Sep 3, 2026

v3.0.2 focuses on documentation and model tuning improvements. It adds the Czech model to the README and optimizes the `lsd_scratch` training configuration with specific batch sizes and schedules. Additionally, the default voice is updated to `alba` to streamline usage with voice cloning.

Aug 25, 2026

This release is a maintenance patch designed to fix a broken build issue found in v3.0.0. It provides a corrected installation package for users who encountered errors with the initial v3.0.0 wheel.

Aug 25, 2026

v3.0.0 introduces the training code for the project and performs significant internal refactoring of the Mimi audio encoder modules. It includes improvements to documentation, Docker support, and the addition of a new English model variant.

Aug 25, 2026

This release introduces WebGPU support for Three.js, enhances the Studio editor with custom editor integrations, and improves the UI with fixes for Safari, pointer sessions, and modal positioning.

Jul 31, 2026

Rspack is no longer experimental, and the release introduces the Studio Protocol for Element installation, along with CLI improvements for rendering multiple stills.

Jul 30, 2026

The `AbsoluteFill` component is refactored to behave like a sequence, the Skills system is rewritten, and the Studio UI is improved with collapsed sections and quick actions.

Jul 29, 2026

Focuses on UI refinements such as timeline improvements, mobile layout updates, and new cropping capabilities, while deprecating legacy effects packages.

Jul 27, 2026

This release focuses on significant improvements to the Remotion Studio user interface, including better drag-and-drop handling, visual refinements for keyframes and rulers, and a fix for opacity layer leaks. It also introduces new drag-and-drop capabilities and updates documentation.

Jul 24, 2026

This release introduces support for new AI models like MiniMax M3 and Huawei vLLM, alongside significant enhancements to container orchestration, file management, and system security. It also adds global accessibility features such as Persian language support and non-interactive installation.

Jul 7, 2026

This release introduces a turbovec ANN backend and LiteParse text extraction, while migrating from ggml-py to ggml-python and improving various pipeline capabilities.

Jul 1, 2026

This beta release enhances the trading platform with v2 Python visualization support, Unix SIGTERM handling, and expanded exchange integrations for Bybit, OKX, and Betfair, alongside numerous bug fixes.

Jun 29, 2026

A major beta release introducing extensive breaking changes across the Python and Rust v2 APIs, including typed lookups, new caching strategies, and expanded exchange support.

Jun 26, 2026

Patch release focusing on local and open-source model routing, hosted-tool compatibility on non-OpenAI routes, and safer remote file downloads.

Jun 25, 2026

Patch release containing bugfixes, including a behavior change where server functions now respect Axum/Actix request body size limits.

Jun 25, 2026

Optimized app upgrade logic, added snapshot recovery rollback prompts, and fixed multiple issues across certificates, containers, monitoring, SSH, and system components.

Jun 16, 2026

Fixed a critical disk spilling issue in DuckLake to prevent Out of Memory (OOM) errors during long-range searches.

Jun 12, 2026

Added significant new features including SIPREC support, VQRTCP protocol, HttpOnly cookie sessions, and an event log UI.

Jun 12, 2026

Bug fix release addressing a file upload send button issue and updating the Android app icon.

Jun 12, 2026

Compatibility patch for Agency Swarm that restores support for OpenSwarm projects using the old two-value parse_agent_flows return shape.

Jun 11, 2026