A hit product in 1 day! DeepSeek+comfyUI+KeLing AI video creation process disassembled, hand-in-hand teaching you how to use AI to make short videos

In the era of short videos, how to quickly and efficiently create attractive video content has become a challenge for many creators and brands. This article will reveal how to use cutting-edge AI tools such as DeepSeek, ComfyUI, and Keling AI to automate the entire process of short video creation.

In the era of short videos, AI tools have become a "god assist" for designers and video creators! Especially AI platforms like DeepSeek, with their powerful AI creation capabilities, are changing the face of creation in the entire industry.

Today, we will share how to use DeepSeek + comfy UI + KeLing to efficiently produce AI video content through a real case study - making a short video for Douyin.

From storyboard design to raw image optimization, to video generation and post-production synthesis, more advanced AI tools can help us improve efficiency and free up more creative space.

01 DeepSeek is easy to use and quick to deploy

DeepSeek is an AI creation tool based on deep learning, focusing on intelligent analysis and high-quality text output. Its core advantages are:

● Powerful model: It can understand and adapt to the needs of different design fields, and provide accurate and flexible support whether it is graphic design, UI/UX design or brand design.

● Easy deployment: Whether deployed locally or in the cloud, you can get started quickly.

● Efficient output: From storyboards to spells, put your creativity into practice quickly.

1. Visit the official website: Open your browser and go to DeepSeek’s official website: https://www.deepseek.com.

2. Register an account: If it is your first time using, please register first and then log in to start creating.

3. Select model: Select options such as "Model version" and "Networking mode" according to your needs.

4. Input requirements: Tell AI your ideas and requirements through text description.

5. Generate ideas: Click the "Send" button and AI will generate storyboards, answers, analysis, etc. based on your needs.

02 Practical analysis of financial short video promotion cases

When promoting Internet finance consumer loan businesses on short video platforms such as Douyin, they often face the following problems:

● The content is too boring: users lose interest and may even swipe away;

● High legal risk: If you are not careful, you may be complained or even fined;

● Poor image: The image of micro-loan business has always been not very good, and users always feel that there are many pitfalls.

Our goals are to:

● Attract users: Make the video content interesting enough for users to watch;

● Compliance publicity: ensure that the content is legal and compliant to avoid risks;

● Improve image: Reverse the negative image of microfinance business through vivid stories;

● Enhance the brand: Strengthen brand awareness and make users feel good about 58 Finance.

To achieve this goal, I asked deepseek to output the following strategies:

Solution 1. Hard-core popular science + expert endorsement

Invite financial experts to appear on camera and explain loan knowledge through short videos, such as interest calculation, repayment plan, consequences of overdue payment, etc.

Solution 2. Emotional resonance + real cases

Tell stories using real user cases, such as "Xiao Ming got into trouble because of overdue loan, but later he repaid the loan on time and regained his life."

Solution 3. IP image + popular science animation

Design a cute IP image (such as "Fubao"), and explain loan knowledge through animation, such as "5 things you must know before borrowing" and "correct operation after overdue payment". Final solution: IP image + popular science animation

The advantages of this solution are:

1. The animation is relaxing and interesting, and users are willing to watch it;

2. IP image is highly malleable and can convey brand warmth and personality.

3. Animation content is highly compliant, risk is controllable, and it is easy to output continuously.

03 Intelligent storyboard design: from creativity to visualization

Send the reviewed video copy to deepseek, and after a few simple adjustments, you can get an intelligent storyboard design that includes scene descriptions.

04 Image spell optimization: Let AI better understand your needs

With the storyboard, the next step is to use the storyboard to write the prompt, put it into comfyUI, use the previously trained lora, and use the flux model to generate fixed frames.

In AI video creation, the prompt is the key to determining the quality of the picture. A good prompt should not only describe the content of the picture, but also include details such as style, light, and color.

1. The basic structure of a mantra

A complete Shengtu spell usually includes the following parts:

● Subject description: the main characters or objects in the picture.

● Scene description: background environment, atmosphere, etc.

● Style setting: the artistic style of the picture (such as cartoon, realism, minimalism, etc.).

● Additional details: light, color, movement, etc.

2. Spell Optimization in Examples

Take the first shot in "Will the property be frozen if the loan is overdue?" as an example:

Content of the picture: Fubao is wearing casual pajamas, sitting on the bed in the bedroom which is piled with small dolls, playing with his mobile phone. The phone prompts "loan overdue", his smile freezes, and his movements stop.

Initial spell:

"A white doll is sitting on the bed playing with a mobile phone. There are many small dolls on the bed, and the phone shows 'loan overdue'."

The incantation after optimization with deepseek:

"The cartoon-style white doll Fu Bao, wearing red casual pajamas, sits on the bed in the bedroom piled with dolls and snacks, lowering his head to play with his mobile phone. He is in a good mood. The sun is shining and the atmosphere is warm."

The full spell in English:

fubao,white skin,Fubao is wearing red casual pajamas,sitting on a cozy bedroom bed full of dolls,playing with his mobile phone,with some snacks beside him,sunshine,happy,red carpet,soft lighting and high resolution,masterpiece,c4d,hd

With the optimized spell, the image generation efficiency is higher. By putting it into the comfy UI, all the still frames can be generated in as fast as 1 hour.

3. Tips for spell optimization

● Keyword stratification: Divide the spell into four levels: subject/scene/style/details, and refine them layer by layer.

● Unified style: Make sure the spell style is consistent in each shot to avoid sudden changes in picture style.

● Emotional guidance: By describing details such as light and color, guide AI to generate images that match the emotional tone.

05 Practical skills of AI video generation

1. Multiple rounds of iterations, gradual optimization

‍

If the generated image does not meet your expectations, you can perform multiple rounds of iterations by modifying the spell or adjusting the parameters.

2. Generate in segments and adjust frame by frame

AI tools have limited processing capabilities for long videos. It is recommended to generate the video in segments, generate video clips for each shot separately, and then synthesize them in post-production software.

3. Notes on post-synthesis

● Transition effect: Use appropriate transition effects to enhance the continuity of the video.

● Dubbing and music: Choose dubbing and music that match the emotional tone of the video to enhance the appeal of the video.

● Subtitle design: Ensure that the subtitles are clear and easy to read, and are consistent with the picture style.

06 Summary and presentation

Through the case study "Will overdue loans freeze property?", we can see that AI tools such as DeepSeek have great potential in video creation. From storyboard design to raw image spell optimization, video generation and synthesis, the AI ​​production workflow of [deepseek+comfyUI+KeLing] can help us efficiently produce high-quality short videos to meet the current fast-paced promotion needs!

AI is not a replacement for designers, but a lever to amplify creative potential. When you use DeepSeek to produce three days of content for your competitors in three hours, you can free up more time to polish the core of the story - this is the real moat for designers and other content creators.

Author: Song Qianran