Make videos on the computer you already own. FreeVideo runs MiniMax H3 in as little as 8 GB of VRAM and 16 GB of RAM, and adapts its acceleration path to your hardware.
Keep pictures and prompt rewrites going when the ComfyUI unload fails
Bowen Xuefd0d907e2afd
diff --git a/freevideo_engine/image_gen/service.py b/freevideo_engine/image_gen/service.pyindex 034baec..3ff148f 100644--- a/freevideo_engine/image_gen/service.py+++ b/freevideo_engine/image_gen/service.py@@ -892,7 +892,11 @@ class Service:resident_process.OWNER.stop_prewarm()IDLE.stop()from ..comfy_residency import release_native- released = release_native(comfy_server())+ try:+ released = release_native(comfy_server())+ except Exception: # only frees memory early; the picture's own budget still decides+ logging.warning('[FreeVideo] Could not unload ComfyUI models before the picture', exc_info=True)+ released = Noneif released is not None:logging.info('[FreeVideo] Unloaded ComfyUI models before the picture: %s', released)with self.mutex:diff --git a/freevideo_engine/prompt_vlm/service.py b/freevideo_engine/prompt_vlm/service.pyindex c5cbec5..423248f 100644--- a/freevideo_engine/prompt_vlm/service.py+++ b/freevideo_engine/prompt_vlm/service.py@@ -1,6 +1,7 @@"""Optional rewrite and ask jobs; prompts, questions and answers stay in memory, never in job reports."""import atexitimport json+import loggingimport osfrom pathlib import Pathimport queue@@ -404,7 +405,10 @@ class Service:IDLE.stop()from ..comfy_residency import release_nativefrom ..image_gen.service import comfy_server- release_native(comfy_server())+ try:+ release_native(comfy_server())+ except Exception: # only frees memory early; the load's own free-memory check still decides+ logging.warning('[FreeVideo] Could not unload ComfyUI models before the prompt model', exc_info=True)with runtime_lock(Path(root) / 'engine.lock', inherit=False) as descriptor:self.check()if action == 'install':