F5-TTS Releases

This release updates the Surya library to an upstream fix and includes various documentation, license, and maintenance updates.

Jan 31, 2026

This release introduces a change in file handle semantics to fix an issue with file closure and re-opening. It also refines cache metrics and adds a cache hit rate tracking metric to logs.

Jan 22, 2026

This release implements a workaround for a Gradio UI bug where the Accordion component's default open=False behavior was not functioning correctly, addressing issue #1239.

Dec 21, 2025

This release upgrades the Gradio dependency from version 5.0 to 6.0 and includes several compatibility fixes to ensure proper integration with the f5-tts_infer-gradio component.

Dec 20, 2025

This significant release fixes a critical issue where the last words in shorter sentences were being skipped during batch inference in both TRT-LLM and PyTorch implementations, and updates several dependencies.

Nov 28, 2025

This release focuses on reworking the VAD training paradigm and significantly improving model performance on various edge cases, including unusual, child, and muted voices. It enhances overall stability on out-of-distribution data without a metric update.

Nov 6, 2025

This release addresses compatibility issues with PyTorch 2.9 and refines internal utility functions, including bug fixes in the tuning module and improvements to error handling. It also includes updates to Java examples and argument usage.

Nov 5, 2025

This release adds support for exporting metrics using the OpenTelemetry (OTel) Protocol and improves operational visibility by changing the default log level from WARN to INFO. It also fixes a bug related to concurrent directory listing operations.

Oct 30, 2025

A minor maintenance release fixing typos and enabling more flexible configuration passing.

Sep 30, 2025

Major update introducing a new layout model via surya for performance boosts and a new rendering option for tables.

Sep 24, 2025

This release updates the numpy version constraint to support numpy 2.x for Python versions 3.11 and above, ensuring compatibility with newer numpy releases.

Sep 13, 2025

Release focused on documentation improvements and adding a deployment example for Modal.

Sep 9, 2025

Improvements to table processing logic to reduce hallucinations and fix text cutting issues.

Sep 4, 2025

This release appears to be a maintenance update with no specific changes documented in the release notes.

Aug 28, 2025

Released a new Voice Activity Detection (VAD) model with improved quality, robustness, and reduced error rates on noisy and multi-domain data.

Aug 26, 2025

Feature-rich release adding ASCII rendering, new arrowheads, styling options, and custom font support.

Aug 19, 2025

Bug fix release replacing the VGT model with YOLO for better performance on consumer hardware.

Jul 31, 2025

Service-level update matching the main library, replacing the VGT model with YOLO.

Jul 31, 2025

Simplified the codebase by removing unused components and fixing warnings to create a more stable version.

Jul 30, 2025

Web interface update simplifying the version by removing unused components.

Jul 30, 2025