Free, open-source AI video generator for Mac. Run LTX-2.5 and MiniMax H3 (Hailuo) locally on Apple Silicon: text-to-video and image-to-video with sound, lip-sync, character LoRA training, songs from lyrics (YuE2) and image editing (Qwen-Image-Edit). One-click install via Pinokio. Offline: no cloud, no API key, no NVIDIA GPU.
Python ai-filmmakingai-videoai-video-generatorapple-siliconhailuoimage-to-videolip-synclocal-ailora-trainingltx-2ltx-2-5ltx-videomacosminimax-h3mlxmusic-generationoffline-aipinokiotext-to-videovideo-generation
docs(STATE): record the v0.14.8 ltx-2-mlx catch-up + pin standing rule
mrbizarroadd7cd34d271
diff --git a/docs/STATE.md b/docs/STATE.mdindex 16e2469..3202857 100644--- a/docs/STATE.md+++ b/docs/STATE.md@@ -4,6 +4,8 @@Current `dev` head: see `git log -1` for the live SHA. `dev` tracks `beta/main` (private repo — see §1).+> **Latest dev (2026-06-01, on `beta/main`, NOT yet public):** ltx-2-mlx pin bumped **v0.14.0 → v0.14.8** (commit `aaf1ba8`). Dropped 6 of 7 runtime patches — they're native upstream now, so `patch_ltx_codec.py` is **codec-only** and dev==prod parity is real. This is the proper fix path for the I2V "mosaic" (#17): it's a macOS **Metal command-buffer watchdog** issue (NOT a RAM-tier issue — dgrauet's `_pre_denoise_flush` commit reproduces on a 64 GB M2 Max), now native and wired into the Q4 one-stage path. Vendored tree stays **PINNED/detached at v0.14.8** (standing rule: always re-pin at the end). Full modality matrix validated on dev (M4 Max 64 GB). **Awaiting Mr Bizarro's Dev-panel test + an affected small-Mac user's confirmation before promoting to public `main`.** Full detail: memory `phosphene_ltx_pin_v0148.md`.+Live URL: `https://github.com/mrbizarro/phosphene` · Linear project: `https://linear.app/hairstylemojo/project/phosphene-9c11240704bb`This doc is the **session-start handoff**. A new Claude window entering this project should read this first, then `CLAUDE.md` (architecture), then the relevant Linear issues.