Hunyuan have released the official image to video model.
it can be trained for image to video lora's.
Workflow Pack: https://civitai.com/models/1013919 (v4 onwards)
You will need:
Checkpoint:
(Native BF16)
V1
https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/resolve/main/split_files/diffusion_models/hunyuan_video_image_to_video_720p_bf16.safetensors
V2
https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/resolve/main/split_files/diffusion_models/hunyuan_video_v2_replace_image_to_video_720p_bf16.safetensors
place inside ComfyUI\models\checkpoints\
(GGUF UNET)
https://huggingface.co/city96/HunyuanVideo-I2V-gguf/tree/main
recommended: (system dependent)
Q8_0 = 14GB
Q5_K_M = 9.43GB
Q4_K_M = 7.86GB
place inside ComfyUI\models\unet\
Clip Vision Model:
https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/resolve/main/split_files/clip_vision/llava_llama3_vision.safetensors
place inside ComfyUI\models\clip_vision\
the text encoder / vae is unchanged, refer to my earlier article on Hunyuan Video (text 2 image)
WORKFLOWS: (GGUF)
