Skip to content
Become a Patron Help Build a Better ComfyUI Knowledge Base
NewsSkywork Opensources SkyReels-V1: A Video Generation Model Focused on AI Short Drama Creation

Skywork Opensources SkyReels-V1: A Video Generation Model Focused on AI Short Drama Creation

Skywork has officially open-sourced its latest video generation model, SkyReels-V1. This model is based on the fine-tuning of the Hunyuan video model, and has achieved several breakthroughs in the field of video generation after training on millions of high-quality film and television data.

Key Features

Professional Visual Effects

  • Cinematic Lighting：Achieving professional-level lighting effects through learning from high-quality film and television works
  • Subtle Expressions：Supporting 33 different expression types, capable of achieving over 400 natural motion combinations
  • Scene Consistency：Exhibiting excellent consistency in the representation of people and scenes

Technical Innovations

  1. Self-Developed Inference Framework

    • Developing the SkyReels-Infer inference framework
    • Significantly improving inference efficiency
    • Optimizing resource usage

  2. Dual Modality Support

    • Supporting text-to-video generation (T2V)
    • Supporting image-to-video generation (I2V)
    • Flexibly adapting to different creative needs

Hardware Requirements and Deployment

Basic Configuration Requirements

  • Recommended video memory: 16GB or above
  • Model size: approximately 24GB
  • Supporting multi-GPU parallel inference

Performance Comparison

Performance on an RTX 4090:

  • Single GPU: generating a 4-second video takes approximately 889 seconds
  • 4 GPUs in parallel: reducing the generation time to 293 seconds
  • 8 GPUs in parallel: further optimizing to 159 seconds

Practical Application Effects

Image-to-Video Generation Effects

  • Character Consistency：Able to well maintain the characteristics of the original image characters
  • Motion Continuity：Generated motions are natural and smooth, with smooth transitions
  • Scene Adaptability：Applicable to a variety of scenarios, including real-world and two-dimensional styles

Application Scenarios

  1. Short Video Creation

    • AI short drama production
    • Creative video content generation
    • Marketing material creation

  2. Content Enhancement

    • Dynamicizing static images
    • Expanding video content
    • Adding special effects

Developer Access

ComfyUI Plugin Support

Model Acquisition

Usage Suggestions

  1. Resolution Selection

    • Default recommendation: using 512 resolution
    • Adjustable based on video memory conditions
    • Supporting post-processing video scaling

  2. Prompt Word Optimization

    • Recommending detailed prompt words
    • Prompt word quality directly affecting generation effects
    • Supporting flexible style adjustments

Open Source and Commercial Use

  • Model has been open-sourced on GitHub
  • Supporting commercial use
  • Providing detailed deployment documentation and examples

Future Outlook

The release of SkyReels-V1 marks a new stage in AI video generation technology. This model not only simplifies the creative process but also provides creators with more possibilities. The team will continue to optimize model performance and look forward to community feedback and contributions.

Project Address