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.
Commits on main
LoRA check: say when a file with another model's layer names is used on an int8 install
ac5a7e0 Bowen Xue committed 1 month ago
Preserve reference video early steps in exported cache identities
47f76d4 Bowen Xue committed 1 month ago
Run workflows saved by older versions: reset unusable optional inputs to defaults
b88c158 Bowen Xue committed 1 month ago
Image generation: a failed fall-back to streaming keeps the card able to retry it
3a84876 Bowen Xue committed 1 month ago
Image generation: falling back to streaming no longer stops half-way when locking pages fails
aea443e Bowen Xue committed 1 month ago
Image generation: a card that holds the whole model keeps no host copy of the weights
196e700 Bowen Xue committed 1 month ago
Image generation on 6-8 GB cards: lighter decoding and more DiT blocks kept on the card
c88f309 Bowen Xue committed 1 month ago
Background compiling after updates and a more capable Repair (0.3.24) (#180)
49ed463 Bowen Xue committed 1 month ago
One overall progress percentage and a full-window workspace (0.3.23) (#164)
0af617e Bowen Xue committed 1 month ago
Community results: the RTX 5080 Laptop with 0.3.21 [skip ci] (#158)
c026f3e Bowen Xue committed 1 month ago
Community results: the 4060 Laptop and 4070 Laptop with 0.3.21, and an RTX 4060 Ti [skip ci] (#157)
44f41ba Bowen Xue committed 1 month ago
README: link the community results again and keep the intro short (#154)
7f2398f Bowen Xue committed 1 month ago
Add community results for seven more machines: an Apple M4 Max and RTX 5060, 4070 Laptop, 5090 Laptop, 5070 Ti Laptop, 5060 Ti and 4070 SUPER GPUs [skip ci] (#153)
bcca627 Bowen Xue committed 1 month ago
The image editor opens again: its fonts ship in the launcher packages (0.3.22) (#151)
6be4564 Bowen Xue committed 1 month ago
FreeVideo for Linux desktops, an image editor in Media, and faster Windows generation when memory is tight (0.3.21) (#150)
502bcd6 Bowen Xue committed 1 month ago
Contributing: pull requests for substantial new capabilities, small fixes as issues (#149)
70b4f3c Bowen Xue committed 1 month ago
Load FreeVideo for offline environment packages; wider attention head groups on Linux GPUs with memory to spare (0.3.20) (#137)
ebc94ec Bowen Xue committed 1 month ago
Keep 8 attention heads in Windows first passes with reference images or keyframes (0.3.19) (#134)
1a78acf Bowen Xue committed 1 month ago
Keep the release notes to twelve earlier versions (0.3.18) (#133)
cad17a5 Bowen Xue committed 1 month ago
Read less from disk when loading the model on Windows with little RAM (0.3.18) (#132)
5815b23 Bowen Xue committed 1 month ago
Run 16 attention heads at a time on Windows when memory allows (0.3.17) (#128)
e86ef6f Bowen Xue committed 1 month ago