Pushing Boundaries in AI Video Generation with Skyreels and Hunyuan

0:00

/0:04

Prompt: FPS-24, a unique car in the road, colored building, few people, realistic / Seed: 685308083704661

Skyreels AI video model is a specialized fine-tuning of the Hunyuan base model that excels in generating more consistent, flicker-free footage. By targeting video-specific training data and leveraging user feedback for iterative improvements, Skyreels aims to reduce artifacts and maintain coherent motion across frames.

Many enthusiasts highlight its enhanced prompt controls, smoother transitions, and compatibility with existing workflows as standout advantages over Hunyuan’s more general-purpose foundation.

(Note: Skyreels while high quality in its own right, had the unfortunate luck of launching at around the same time that Wan 2.1 broke out onto the scene, and became the SOTA open source video model for quality. Skyreels still holds a speed advantage.)

Verified to work on ThinkDiffusion Build: Jan 28, 2025

ComfyUI v0.3.12 with Huyuan video model support.
Why do we specify the build date? ComfyUI and custom node versions that are part of this workflow that are updated after this date may change the behavior or outputs of the workflow.

Why choose Skyreels model over Hunyuan native base model?

0:00

/0:17

Prompt: sunset at the sweet waves of sea, realistic / Seed: 1171400988189

💡

I prefer the Hunyuan Skyreels model because it delivers resolutions up to 960x544 in 97 frames. Trained on 10 million videos and clips, it excels in action recognition, scene understanding, and character awareness, resulting in more coherent and fluid videos. In comparison, the Hunyuan native model shows limited movement and lacks in quality.

How to run Hunyuan Skyreels Image2Video in ComfyUI

Installation guide

ThinkDiffusion-StableDiffusion-ComfyUI-Img2Vid-SkyReels-Hunyuan

Download the File Here

ThinkDiffusion-StableDiffusion-ComfyUI-Img2Vid-SkyReels-Hunyuan.json

22 KB

💡

Workflow Updated 4/10/2025: Replaced with new workflow using comfy native nodes

💡

Download the workflow and drag & drop it into your ComfyUI window, whether locally or on ThinkDiffusion. If you're using ThinkDiffusion, it's necessary to use at minimum the Turbo 24gb machine, but we do recommend the Ultra 48gb machine.

Custom Node

If there are red nodes in the workflow, it means that the workflow lacks the certain required nodes. Install the custom nodes in order for the workflow to work.

Go to ComfyUI Manager > Click Install Missing Custom Nodes

Check the list below if there's a list of custom nodes that needs to be installed and click the install.

Models

For this guide you'll need 4 models. There will 3 models that needs to be downloaded, 1 models are pre-loaded and 1 model is optional.

💡

Download the necessary models:
1. hunyuan_video_vae_bf16.safetensors
2. skyreels_hunyuan_i2v_bf16.safetensors
3. clip-vit-large-patch14/model.safetensors (auto-download)
4. llava_llama3_fp8_scaled.safetensors

Go to ComfyUI Manager > Click Model Manager

Search for the model to install and when you find the exact model that you're looking for, click install and make sure to press refresh when you are finished.

Optional Model Path Source

Use the model path source if you prefer to install the models using model's link address and paste into ThinkDiffusion MyFiles using upload URL.

Model Name	Model Link Address	ThinkDiffusion Upload Directory
hunyuan_video_vae_bf16.safetensors	📋 Copy Path	.../comfyui/models/vae/
skyreels_hunyuan_i2v_bf16.safetensors	📋 Copy Path	.../comfyui/models/diffusion_models/
clip-vit-large-patch14/model.safetensors	Auto Download	Auto Dowload
llava_llama3_fp8_scaled.safetensors	📋 Copy Path	.../comfyui/models/text_encoders/

Step-by-step Workflow Guide

This workflow was pretty easy to set up and runs well from the default settings. Here are a few steps where you might want to take extra note.

Steps	Recommended Nodes Setup
1. Load Image
2. Resize Image Set the resolution up to 960x544. Don't set beyond that resolution because it will degrade the quality of the video.
3. Set Frame Length Set the frames up to 49, 73 or 97
4. Set the Models Set the models as seen on the image. Be sure precision is set to bf16.
5. Write a Prompt Include "fps-24" in your prompt and use short prompts. Less than 50 words. Describe the image and add a few movement instructions.
6. Check the Sampling Settings Check the settings as seen on the image.
7. Check the Generated Video

💡

Although Skyreels offers significant capabilities, I find there is still room for optimization as it may not consistently produce ideal results on the first try. To achieve satisfactory outcomes, I often need to generate multiple times.

Examples