From b101deb894bea7ff4bd1c4bb050f8239a8a2a70b Mon Sep 17 00:00:00 2001 From: Adrien Toupet Date: Tue, 9 Dec 2025 01:01:54 -0500 Subject: [PATCH] docs: fix contributor links and formatting in release notes --- README.md | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/README.md b/README.md index f285214..de87611 100644 --- a/README.md +++ b/README.md @@ -38,9 +38,9 @@ We're actively working on improvements and new features. To stay informed: **2025.12.09 - Version 2.5.18** -- **🚀 CLI: Streaming mode for long videos** - New `--chunk_size` flag processes videos in memory-bounded chunks, enabling arbitrarily long videos without RAM limits. Works with model caching (`--cache_dit`/`--cache_vae`) for chunk-to-chunk reuse *(inspired by @disk02 PR contribution)* +- **🚀 CLI: Streaming mode for long videos** - New `--chunk_size` flag processes videos in memory-bounded chunks, enabling arbitrarily long videos without RAM limits. Works with model caching (`--cache_dit`/`--cache_vae`) for chunk-to-chunk reuse *(inspired by [disk02](https://github.com/disk02) PR contribution)* - **⚡ CLI: Multi-GPU streaming** - Each GPU now streams its segment internally with independent model caching, improving memory efficiency and enabling `--temporal_overlap` blending at GPU boundaries -- **🔧 CLI: Fix large video MemoryError** - Shared memory transfer replaces numpy pickling, preventing crashes on high-resolution/long video outputs *(inspired by @FurkanGozukara PR contribution)* +- **🔧 CLI: Fix large video MemoryError** - Shared memory transfer replaces numpy pickling, preventing crashes on high-resolution/long video outputs *(inspired by [FurkanGozukara](https://github.com/FurkanGozukara) PR contribution)* **2025.12.05 - Version 2.5.17**