Skip to content

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.

fix(install): fail fast on macOS <14 + name the Metal GPU-timeout abort (#44)

mrbizarrofbfa56b5b204
diff --git a/install.js b/install.js
index d730850..52576d7 100644
--- a/install.js
+++ b/install.js
@@ -142,6 +142,15 @@ module.exports = {
// had no log evidence to confirm. These echoes change nothing
// operationally; they just leave a trail.
"echo '=== install diagnostics: venv create ==='",
+ // SHIP-BLOCKER GUARD (2026-07-23, from @hottboytank's Pinokio report):
+ // mlx 0.31.1 publishes NO macOS-13 wheel — its builds start at
+ // macosx_14_0_arm64. On Ventura the mlx step dies deep inside uv's
+ // resolver ("No solution found ... no wheels with a matching platform
+ // tag"), ltx_core_mlx never installs, and the user only sees a cryptic
+ // ModuleNotFoundError at the very end — under an auto-generated title
+ // about python3.11 that points at completely the wrong thing.
+ // Fail fast, up front, with the actual fix.
+ "MACOS_VER=$(sw_vers -productVersion 2>/dev/null || echo 0); MACOS_MAJOR=$(echo $MACOS_VER | cut -d. -f1); if [ \"$MACOS_MAJOR\" -lt 14 ] 2>/dev/null; then echo '=================================================================='; echo \"PHOSPHENE CANNOT INSTALL ON macOS $MACOS_VER\"; echo 'Phosphene needs macOS 14 (Sonoma) or newer.'; echo 'Why: mlx 0.31.1 ships no macOS 13 build, so the engine cannot install'; echo 'and every generation would fail.'; echo 'Fix: System Settings > General > Software Update -> macOS 14 or 15,'; echo 'then reinstall Phosphene.'; echo '=================================================================='; exit 1; else echo \"macOS $MACOS_VER OK (needs >= 14)\"; fi",
"which uv && uv --version || echo 'uv NOT FOUND'",
"which python3.11 && python3.11 --version || echo 'system python3.11 NOT FOUND (uv will try to fetch)'",
"uname -a",
diff --git a/mlx_ltx_panel.py b/mlx_ltx_panel.py
index 77736cf..fa3b510 100644
--- a/mlx_ltx_panel.py
+++ b/mlx_ltx_panel.py
@@ -4822,7 +4822,24 @@ class WarmHelper:
hint = {
"SIGKILL": "out of memory (jetsam) or external kill — close memory-heavy apps and retry",
"SIGSEGV": "native segfault inside MLX/Metal — share the crashlog at ~/Library/Logs/DiagnosticReports/python3.11_*.crash",
- "SIGABRT": "C-level assertion failed — share the crashlog as above",
+ # 2026-07-23 (#44, @BobDixon58, M1 Max): the dominant SIGABRT
+ # in the wild is a Metal GPU-watchdog timeout during Gemma
+ # prompt encoding — "[METAL] Command buffer execution failed:
+ # Caused GPU Timeout Error". Name it, because the generic
+ # "assertion failed" sends people hunting the wrong thing.
+ # Note LTX2_GEMMA_EVAL_EVERY is already 1 (the smallest
+ # command-buffer setting) so there is no knob left to turn —
+ # this is an engine/driver issue worth reporting upstream.
+ "SIGABRT": (
+ "C-level abort. If the log shows '[METAL] Command buffer "
+ "execution failed: Caused GPU Timeout Error', this is the "
+ "macOS GPU watchdog killing an over-long Metal command "
+ "buffer (seen on M1/M2-class GPUs during prompt encoding). "
+ "Phosphene already runs the smallest command buffers it "
+ "can (LTX2_GEMMA_EVAL_EVERY=1), so please open an issue "
+ "with your chip + macOS version and the crashlog at "
+ "~/Library/Logs/DiagnosticReports/python3.11_*.ips"
+ ),
"SIGBUS": "memory access fault — could indicate a Metal driver issue or a bad weight file",
}.get(sig_name, "external kill")
raise RuntimeError(