This Setup and workflow have been inspired from basic Stable video infinity (SVI pro 2.0) already explained in other tutorial. The main aim is to generate more consistent videos without color degrading and glitches.
 |
| Stable video infinity architecture |
Installation
1. Update ComfyUI to its latest version from the Manager.
2. Make sure you have already installed the Kijais' ComfyUI WanVideoWrapper custom node and have the basic Wan 2.2 I2V
High and Low noise, Lightx2v Lora, Vae, text encoders etc models
already downloaded and setup. If already done, then update this custom
node to its latest version from the Manager by selecting Custom nodes
manager option.
3. People having VRAM minimum 16GB can use this:
(a) Download Smooth Mix Wan 2.2 I2V 14B High Noise from CivitAI platform.
(b) Download Smooth Mix Wan 2.2 I2V 14B Low Noise from CivitAI platform.
4. If you have low Vrams use GGUF variant instead:
(a) Download Smooth Mix Wan 2.2 (High Noise & Low Noise) GGUF
(b) Download Smooth Mix Wan 2.2 V2.0 GGUF
5. Download Wan2.2 Distill Loras if you want to use for faster generation having low VRams but quality degradation is there.
(a) wan2.2_t2v_A14b_high_noise_lora_rank64_lightx2v_4step_1217.safetensors
and
(b) wan2.2_t2v_A14b_low_noise_lora_rank64_lightx2v_4step_1217.safetensors
6. Next, download SVI Wan2.1 I2V 14B lora v2.0 from hugging face repository.
(a) SVI_Wan2.2-I2V-A14B_high_noise_lora_v2.0_pro.safetensors and
(b) SVI_Wan2.2-I2V-A14B_low_noise_lora_v2.0_pro.safetensors)
Workflow
1. Download workflow(Wan2.2-SVI-V2.0-Pro-SmoothMix.json) from our Hugging Face repository.
2.Here, its need not to change anything. It uses the basic Wan 2.2 I2V workflow.
Shift- 5
Sampler- Euler
Steps-4.0
CFG-1-2
Donot use higher shift value. As the moment you add higher the shift value, the more inconsistent your character will be.
3. There are default 8 extended nodes for generating longer videos. You can add more individual nodes into it to make it further longer. Copy the groups and add to the end.
Then, connect the extended image node to the extended groups created image input.Now connect the preview sampling of older node to preview sampling to new node.
Now repeat this task to get infinite length videos.
