Video generation method and device, electronic equipment and medium

By generating watermarked images from user-input text through structured processing and then parsing the watermarked information using a video generation model, the problem of low video generation efficiency in existing technologies is solved, achieving efficient video generation and resource utilization.

CN122053930APending Publication Date: 2026-05-15MIGU CO LTD +1
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
MIGU CO LTD
Filing Date
2025-12-19
Publication Date
2026-05-15

AI Technical Summary

Technical Problem

Existing technologies lack structured tags in the video generation process, resulting in wasted computing resources and low generation efficiency. Furthermore, the adaptation of prompt words between different models requires manual intervention, leading to a fragmented process.

Method used

By structuring user-input text information, a target image containing watermark information is generated. Then, a preset video generation model is used to parse the watermark information and directly generate a target video that matches the target image, avoiding repeated analysis and manual intervention.

Benefits of technology

It improves the efficiency of video generation and resource utilization, reduces computing power waste, and achieves unified prompt word extraction and cross-model adaptation across different models.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122053930A_ABST
    Figure CN122053930A_ABST
Patent Text Reader

Abstract

The embodiment of the invention discloses a video generation method and device, electronic equipment and a medium, and relates to the technical field of artificial intelligence, one specific implementation mode of the method comprises the steps that a target image containing watermark information is generated based on text information input by a user and a preset image generation model, the watermark information is obtained by performing structured processing on text information input by a user; and performing dynamic parameter processing on the target image based on the watermark information and a preset video generation model, and generating a target video matched with the target image. In the process, the watermark information is embedded into the target image, in the video generation process, prompt words do not need to be input repeatedly, the target video meeting the requirement can be generated only according to the watermark information of the target image, the computing power of analyzing the image by a video model is reduced, and the video generation efficiency is improved.
Need to check novelty before this filing date? Find Prior Art