I taught myself AI filmmaking, and this article describes the technology as of June 18, 2025. AI technology undergoes significant changes every few months, with the possibility of completely overturning previous methods. So, the first key to self-learning AI filmmaking is always keeping up with the changes.
How do you keep up with changes? The fastest way is to follow a group of content creators on Bilibili and YouTube focusing on AI filmmaking. Browse whenever you have time, and bookmark critical new technologies when you encounter them. These creators usually stay on top of trends and won’t let you fall behind.
Watch the latest technology on Bilibili and YouTube, then view the newest AI films on WeChat Video and YouTube. You can see the progress of the entire community. This is like scanning rankings when writing web novels – AI film creation isn’t entirely about working behind closed doors.
AI filmmaking currently uses the approach of generating images first and then converting them to video. Of course, there are also multi-reference direct video generation methods. Therefore, image generation is the foundation of everything.
The first thing to learn is Midjourney. Although there are other options for image generation now, Midjourney’s aesthetic quality remains the best. You can get started by looking for a video that introduces Midjourney parameters. Sometimes, you can also use ChatGPT’s GPTs to generate prompts.
Next is mastering multimodal image editing. You can use Flux Kontext on Liblib or Doubao. Both can modify images based on natural language. This can be used to create multi-angle storyboards and make characters interact.
After completing the simple parts, you must understand how to use Flux. There are plenty of workflows on Liblib and RunningHub – you don’t need to learn to write workflows initially. LoRA training on Liblib is also convenient and fast, requiring no parameter adjustments and direct operation.
Flux’s most significant use case is LoRA. Currently, you can use wan2.1 for multi-angle character generation when training LoRA – you can find this by searching on Bilibili. Flux can use Krea’s optimized version, which is quick and convenient and supports LoRA. Flux in Liblib’s WebUI can also work if you just need simple image generation.
Other image generation tools, such as Google’s Whisk, can be helpful, but Whisk isn’t quite mature yet. Midjourney and Flux are mature, so I recommend beginners master these two to perfection first. They can basically meet AI filmmaking needs.
After image generation comes the inevitable upscaling process. Here, I recommend Magnific—though it costs money, it’s effective. Color grading is where everyone shows their skills—remember you can grade twice or more: once for images and once for video.
Next comes video production. Veo3’s text-to-video has the most cinematic feel currently, but if you need long narrative storytelling, you’ll need image-to-video. I currently use a combination of Kling 2.1 and Dreamina 3.0. In the Kling 1.6 era, I used Kling for all medium shots and above featuring people.
But now Dreamina 3.0 can also render detailed characters. You must try multiple times to understand how Kling and Dreamina perform with different shots – inevitably, there will be some failed attempts. But after you’ve made films, you’ll increasingly sense each model’s strengths and weaknesses. For example, Kling 2.1 is good at fast camera movements.
After producing the video, you can use Topaz’s video enhancement for final processing.
Next comes soundtrack creation. I recommend Suno 4.5, which can use natural language prompts and responds to prompts much better than previous models. You can use GPTs to generate Suno prompts, giving you relatively professional soundtracks.
AI filmmaking is a process that combines multiple software tools and various workflows. You must gradually sense each tool’s pros and cons and mix them. It’s easy to get started, but there are many details. You’ll need time to let it settle. After learning AI tools, don’t forget to study directing techniques and the fundamentals of audiovisual language to strengthen the cinematic feel.
I taught myself AI filmmaking, and this process has been enjoyable because creation brings me joy. Please also find your expressive voice through AI.
Follow me, and let’s explore AI filmmaking together.
