CPU3DAI tools for 3D, video and audio

SkyReels: video generator — what it does and what you need to run it

SkyReels is an open-source video generator from Skywork AI. The authors position it as the first open model focused on creating videos featuring people. The tool can work both from a text description and from an image.

What it does

  • Generates videos from text (t2v) and from images (i2v).
  • Creates clips with a focus on people — the authors call the model a "human-centric video foundation model".
  • According to the repository data, on a single RTX 4090 with full VRAM optimization, peak consumption is 18.5 GB. In this mode, you can generate a 4-second video at 544x960 pixels and 97 frames.
  • At maximum VRAM load, generation of a 12-second video at 544x960 pixels and 289 frames is available. This uses the --sequence_batch parameter. Generation time on a single RTX 4090 is about an hour and a half; adding additional GPUs, as stated by the developers, significantly speeds up the process.

What you need to run it

  • Platform: CUDA. The authors recommend version 12.2.
  • Language: Python, version 3.10 (developer recommendation).
  • VRAM: as described by the author, peak consumption is 18.5 GB. For maximum video length, the full available memory on an RTX 4090-class accelerator will be required.
  • Code license: non-standard. The authors do not specify a particular type in the repository; a manual review of the terms in the license file will be required.
  • Repository: github.com/SkyworkAI/SkyReels-V1. Last code update — March 10, 2025.
  • Project website: www.skyreels.ai.

Who it suits

The tool is designed for those who work with video on local hardware and are willing to deal with manual installation. It suits researchers and developers who need an open model focused on filming people. The availability of open source allows studying the architecture and fine-tuning the model for your own tasks.

It will not suit those who do not have a GPU with 18.5 GB of VRAM or more. Generating long clips requires significant time even on a powerful accelerator, so for continuous work without a cluster of several cards, the tool may prove impractical. Also, consider the non-standard license — before commercial use, you need to carefully review the terms.

SkyReels is a narrowly specialized open-source video generator with high VRAM requirements. The authors focus on scenes with people and provide the code for self-assembly, but licensing details and actual quality metrics on third-party data remain the user's responsibility.

Input data Text description or image T2V / I2V text-to-video image-to-video Model SkyReels V1 Human-centric Video Diffusion Transformers open source Processing CUDA 12.2 Python 3.10 VRAM 18.5 GB peak load RTX 4090 Result Video file 544×960 px 4 sec / 97 frames 12 sec / 289 frames ~1.5 h on GPU T2V I2V SkyReels video generator pipeline Open model focused on humans · Skywork AI · 2025
How the SkyReels pipeline works. The diagram is drawn from the tool’s fact sheet.

Fact sheet

Code licensenon-standard (check manually) source
PlatformCUDA per the author’s description source
VRAM18.5 GB per the author’s description source
Python3.10 per the author’s description source
LanguagePython source
Last code change2025-03-10 source
Repository created2025-02-12 source
Changes often — as of 2026-08-19
GitHub stars2694 source
Forks309 source

Values are collected automatically from official sources and were checked on 2026-08-19. Each one links to its source, and values taken from the developer’s pages also carry a verbatim quote — hover over the note. Pricing and versions are shown as of the check date and change most often; verify on the vendor’s site before buying.

See also