Synthid-Bypass Releases

This beta release focuses on UI improvements, structured logging, and advanced model routing capabilities. It introduces support for non-stream fallback, WebSocket sessions, and various bug fixes including crash protection and error handling improvements.

Jun 2, 2026

Updates PyVideoTrans by integrating the Chatterbox TTS engine as a built-in dependency, upgrading Torch to version 2.7, and removing the gTTS audio channel. Includes crash protection fixes and bug fixes for specific TTS models.

Jun 2, 2026

Introduces PaddleOCR-VL-1.6, a significant upgrade to the document parsing model achieving new SOTA accuracy (96.33%) on OmniDocBench. It includes enhanced capabilities for table and ancient text recognition, official SDKs, and multi-page TIFF support.

May 28, 2026

A minor maintenance release with no detailed changelog provided in the snippet, only a link to the full changelog.

May 23, 2026

This is a minor patch release focused on fixing a bug related to upload proxy public URLs and updating deployment documentation for the Nx build system.

May 22, 2026

This release introduces a new OpenAI Responses API action, enhances results export with time filters, and addresses numerous security vulnerabilities including SSRF protections, credential access control, and dependency upgrades.

May 21, 2026

A minor feature update adding a new backend component for the audio-lab module.

May 15, 2026

Focuses on the smart-turn pipeline, adding publishing capabilities for fine-tunes and fixing quantization issues.

May 11, 2026

Enhances the smart-turn evaluation and CLI pipeline, specifically focusing on persisting evaluation metrics.

May 10, 2026

This release features a major refresh of the Terminal User Interface (TUI), including redesigned tabs and a stats panel. It adds visual improvements like per-field colors and address scope labels, alongside enhancements to the Reverse DNS resolution and parser stability.

May 5, 2026

Updates models and introduces mining upload capabilities for the smart-turn component.

May 4, 2026

Initial release of the smart-turn training and evaluation pipeline for the ZH (Chinese) module.

Apr 30, 2026

This release introduces deep integration with the Hugging Face ecosystem, supporting 20 major models with flexible inference engine switching. It also enhances document processing by adding Markdown conversion and DOCX export capabilities for parsing results.

Apr 21, 2026

This patch release focuses on PaddleOCR-VL improvements, adding new hardware support and resolving dependency issues for specific NPUs.

Apr 14, 2026

This release focuses heavily on security sandboxing, CPU efficiency, and platform support improvements, including Windows/macOS/Linux hardening and VLAN parsing.

Apr 9, 2026

Apr 9, 2026

This release introduces new AI models, onboarding workflows, and significant security patches, alongside a migration to NX and oRPC.

Apr 8, 2026

A major update introducing significant TUI improvements, deep packet inspection for multiple protocols (MQTT, STUN, BitTorrent), and native multi-architecture Android support.

Mar 17, 2026

This release introduces a new complex document parsing model (PaddleOCR-VL-1.5) with high accuracy and adds support for MLX-VLM inference services.

Jan 29, 2026

This release focuses on fixing internal build errors and logging issues, specifically addressing header injection in backend requests and improving development tooling.

Jan 11, 2026