Using DeepSeek to create AI videos is truly an efficiency marvel. Novices can get started in just 10 minutes, and the results are astonishing. DeepSeek is really powerful. During this Spring Festival, it dominated the screens.
On January 27, 2025: The DeepSeek app topped the free APP download rankings in the Chinese and US regions of the Apple App Store, surpassing ChatGPT in the US download list.
DeepSeek aids in video production, making operations simpler and more efficient. Let’s deconstruct the new AI video production process with the following tools: 1. Utilize DeepSeek to create script scripts, as well as copy narration and generate painting prompts. 2. Employ AI tools (MJ or Miaohuà) to draw scenes and characters. 3. Use AI video tools to generate animations. 4. Add sound, special effects, subtitles, and scene transitions during editing.
Step one: DeepSeek generates video scripts and storyboards. Before DeepSeek, we were using AI tools like ChatGPT to generate scripts. However, DeepSeek is an inferential model; AI interaction does not require the fine-tuning of prompts like other AI tools. The art of DeepSeek’s prompts is that there is no art—simply ask in colloquial language. Let’s use our national pride, DeepSeek, to handle scriptwriting and image generation.
Just activate deep thinking mode and ask DeepSeek! I want to create an AI video about the Legend of the White Snake; help me design the script, copy, and images, including a timeline, output in table format, with a video length of approximately 60 seconds. DeepSeek directly generated a table containing storyboard descriptions and subtitles. It also intelligently provided AI painting and AI video shot prompt references on the right, completing the preliminary framework of the entire video content. Below is a script generated by ChatGPT with the same prompts; a comparison will reveal the difference. In addition to storyboard descriptions, DeepSeek also offers suggestions for background music and rhythm control, which can be used for reference.Step two: Generate storyboard image painting prompts. The storyboard descriptions from the previous step are not yet suitable for drawing images. Please optimize the detailed storyboard script according to Midjourney’s syntax rules, including image details, style parameters, and shot language, to generate Midjourney painting prompts.
DeepSeek provided painting prompts: A simple instruction completes the AI video shot prompts, with excellent efficiency and quality.Step three: Generate storyboard images. In the previous step, we obtained 8 MJ painting prompts, and we generate images in Midjourney. The image output is better with MJ, but Midjourney is a paid tool and requires English prompts. For users unaccustomed to using MJ, domestic tools like Miaohuà and Jimeng offer a Chinese interface and are free to use.
Midjourney official website: https://www.midjourney.com/home
Miaohua: https://miaohua.sensetime.com/inspiration?invite_code=w43E5L3nZ Jimeng: https://jimeng.jianying.com/ Take generating pictures with the MJ web version as an example: Copy the painting prompt generated in the first step into the prompt input box at the top: Copy the prompts for each shot one by one and press Enter. The generated pictures will be displayed in the Create area. After generating the pictures, if you are not satisfied with the effect, you can edit the prompt or regenerate it once. Next, let’s generate all the shot pictures: Shot 1: Chinese Ink Painting Opening (0:00 – 0:06) Prompt: Chinese ink painting of West Lake, misty rain over pagoda, white snake coiled on lotus leaf, soft gradient cyan and gray tones, delicate brush strokes, wuxia atmosphere, fog effect, depth of field blur –ar 16:9 –style expressive –v 6.0 Shot 2: Transforming and Crossing the Tribulation (0:07 – 0:12) Prompt: Full body shot of white snake transforming into woman, dynamic lightning strikes, flowing white hanfu with silk texture, green energy aura around hands, cinematic storm lighting, ultra-detailed scales to fabric transition, epic composition –chaos 40 –stylize 800 –v 5Scene 2: Broken Bridge Encounter (0:13-0:18)
Keywords: Two-shot closeup, Chinese beauty in white dress, scholar with oil-paper umbrella, suspended rain drops, snake tail illusion in puddle reflection, Kodak Portra 800 film grain, shallow depth of field, romantic color grading — Aspect Ratio 3:2 – Style raw Scene 3: Warmth of the Marketplace (0:19-0:24) Keywords: Wide angle of ancient Chinese herbal shop, shelves with dried herbs and porcelain jars, warm golden hour light through paper windows, woman in white measuring medicine, soft focus background, Studio Ghibli style pastel colors — Aspect Ratio 16:9 – Style scenic Scene 4: Dragon Boat Festival Surprise (0:25-0:30) Keywords: Dramatic horror scene, woman clutching throat in pain, snake shadow on wall by candlelight, spilled golden wine cup, Dutch angle composition, high contrast chiaroscuro lighting, UHD textures — Stylize 600 — Version 6Shot 0: Water Flooding Jinshan (0:31 – 0:38) Prompt: Epic battle scene, giant tsunami waves crashing over Buddhist temple, two women flying in air with glowing water dragons, monk holding golden staff emitting light beam, dynamic splashes, Greg Rutkowski art style, ultra-wide panorama –chaos 7
Shot 7: Suppressing Demons in Leifeng (0:39 – 0:48) Prompt: Extreme closeup of woman’s face with falling tear, golden light beam pulling her upward, broken jade hairpin in foreground, cinematic bokeh background, hyper-realistic skin texture, Makoto Shinkai color palette –style expressive –stylize 900 Shot 8: Reunion After Reincarnation (0:49 – 0:54) Prompt: West Lake sunset, modern teenager holding glowing jade hairpin, holographic projection of ancient woman smiling, neon reflections on water, blending traditional and futuristic elements, octane render In this way, we completed the picture generation required for the AI short video. During this process, we need to draw cards and modify the prompts to obtain the pictures we are satisfied with.Step 4: Utilize AI tools such as Kling AI to transform generated images into videos. Upload the saved images from Step 4 into the AI video tool and click on the ‘Image to Video’ feature. Currently, there are three mainstream AI video software options: Kling AI: https://klingai.kuaishou.com/, Conch AI: https://hailuoai.com/video, and Vidu: https://www.vidu.cn/create. Among these, Kling AI is recognized for its superior generation effects, Vidu excels in speed, and Conch AI provides excellent emotional portrayal in characters. All three tools support video generation from reference images. To generate a video in Kling AI: select ‘Image to Video’, choose the 1.6 model, upload the images that have been generated, select high quality, and create the first video.
