Ditto, a free and open-source AI that can edit videos using text prompts.Developed by Ant Group, an affiliate of Alibaba, Ditto is a holistic framework designed for instruction-based video editing. It utilizes a novel data generation pipeline to create a large-scale, high-quality dataset of video editing examples.
Tested on 10GB GPU + 64GB RAM + 'Sage Attention' , 3 120GHZ monitors, Firefox (21 Tabs open 4 actively used) , background music , Ubuntu 24.04.3 LTS: works like a charm
Base model: https://civarchive.com/models/1651125?modelVersionId=1868891 or any wan2.1 T2V (Text to video)
Wan21_CausVid_14B_T2V_lora_rank32_v2.safetensors to
loras/for inference accelerationWan2_1_VAE_bf16.safetensors to
vae/wan/umt5-xxl-enc-bf16.safetensors to
text_encoders/fp8 is not supported yet
Workflow in training data file or search civitai for different variations
More info HuggingFace | GitHub
These models are redistributed here for the sake of convenience.