Creating Video with AI
At a recent AI video workshop, Jacob Rangel did something useful: he skipped the polished keynote version of AI and showed the messy working version.
Not the “type one sentence and get a Hollywood movie” fantasy.
The real version.
Folders full of images. Repeated generations. Strange audio issues. Characters that almost stay consistent. Tools that are brilliant one minute and maddening the next. In other words, the version business owners, creators, and filmmakers actually need to understand.
AI video is moving fast. But Jacob’s core message was practical: the people who win with these tools will not be the ones who simply chase the newest model. They will be the ones who learn a repeatable creative process.
Here are the biggest takeaways from his workshop on creating video using AI.
1. AI video is not a shortcut around creativity
One of Jacob’s first points was that AI video has changed dramatically in just the last couple of years.
He referenced the moment when Tyler Perry paused an $800 million studio expansion after seeing what text-to-video tools like Sora could do. That story gets repeated a lot because it captures the fear side of AI: jobs, studios, sets, crews, and production economics all being disrupted at once.
But Jacob’s workshop was not about fear.
It was about craft.
The tools can now generate realistic scenes, camera movement, sound, character moments, and cinematic visuals. But they still need taste, direction, and judgment. AI can create a snowy cabin in the woods. It cannot automatically know why that cabin matters to your story.
That distinction matters.
For a small business, this means AI video can help create more content faster. For a filmmaker, it can reduce the cost of experimentation. For a founder, it can help prototype an idea before investing in full production.
But AI is not the director.
You are.
2. Start with the image, not the video
One of Jacob’s strongest recommendations was simple: don’t start by typing a random video prompt and burning credits.
That might have worked when AI video was a novelty. It was fun. It felt magical. But it was also inconsistent.
Now, the better process is to start with a strong image.
Jacob described using sketches, reference images, and detailed prompts to create a base visual before moving into animation. This gives the video model a clear starting point. In animation language, this is called a “keyframe.”
A keyframe is a still image that anchors the scene.
Think of it like giving directions to a contractor. “Build me a cool kitchen” is vague. “Build me a kitchen that looks like this sketch, with these materials, this layout, and this lighting” is much better.
Same idea with AI video.
Jacob’s process often starts with:
A sketch or rough concept
A generated character image
A visual reference for the scene
A specific camera angle
A prompt refined with help from ChatGPT or Claude
A final image that becomes the base for animation
That process gives the AI model less room to wander.
And wandering is where things get expensive.
3. Protect your original ideas early
Jacob also spent time on intellectual property, which is especially important for creators building original characters, shows, or brand assets.
His advice: start with your own work.
Even if the sketch is rough, draw it. Photograph it. Save it. Document the development process.
Why?
Because if a character, concept, or series ever becomes valuable, you want a record showing that it originated with you. That does not mean every creator needs to become a copyright attorney. But it does mean you should be intentional.
For example, Jacob showed how he created characters for a futuristic diner concept where humans, AI, and robots work together. The characters started as his own sketches and ideas, then were developed using AI tools.
That matters.
If you are creating an AI-generated mascot for your business, do not start by asking for “a character like Mickey Mouse” or “a superhero like Iron Man.” That is asking for trouble. Start with your own concept, your own description, and your own visual direction.
My suggestion is this:
Create a simple project folder for every AI video project. Save your sketches, prompts, reference images, outputs, revisions, and final files. It may feel tedious in the moment, but future-you will be grateful.
And future-you is usually the one stuck cleaning up present-you’s downloads folder. Ask me how I know.
4. Use the right tool for the right job
A major theme from the workshop was that there is no single perfect AI video tool.
Jacob talked about using different platforms for different parts of the process. Some tools are better for image generation. Some are better for animation. Some are better for talking scenes. Some produce stronger sound. Some are expensive but worth it when the shot really matters.
The tools mentioned included:
Google AI Studio for image and creative generation workflows
Nano Banana Pro for image creation and visual development
Runway as a place to test different video models and generate clips
Seed Dance as a high-quality video model with strong cinematic and audio capabilities
Google Veo for talking scenes
Suno for music generation
ChatGPT and Claude for prompt writing, shot lists, and creative development
The practical lesson is not “go use all of these tools.”
The lesson is to build a workflow.
For example:
Use ChatGPT or Claude to develop the scene and shot list.
Use an image tool to create the character, setting, and keyframes.
Use a video tool like Runway to test animation models.
Use a higher-quality model only when the shot is worth the cost.
Add music and final polish separately.
That approach saves time, money, and frustration.
5. Generate options before committing
One of Jacob’s best tactical tips was the “2x2 grid” method.
Instead of generating one image at a time, he prompts the tool to create a 2x2 grid with four different options. Then he reviews the four options, selects the best frame, crops it, and upscales it if needed.
That is a smart operator move.
It turns AI generation into a rapid creative selection process instead of a slot machine.
For example, if you need a conference room scene, you can ask for four different camera angles in one generation. Maybe one angle has better lighting. Maybe another has better composition. Maybe one gives you the exact feeling you need.
Then you build from the winner.
This is especially useful for business content. If you are creating a social media video, explainer, ad concept, or product scene, do not burn time trying to get the perfect result on the first try.
Generate options.
Pick the strongest one.
Refine from there.
6. Asset management becomes a real problem fast
This may have been the most underrated business lesson in the entire workshop.
Jacob said one of the biggest problems is managing all the assets you create.
That is not glamorous, but it is very real.
AI tools make it easy to generate dozens, hundreds, or even thousands of images and clips. The creative upside is obvious. The operational downside is also obvious: your downloads folder becomes a crime scene.
Jacob’s recommendation was to organize as you go.
Clean your downloads folder before a new session. Use tags. Create folders. Label characters, scenes, angles, versions, and final selects.
For a serious AI video project, your folder structure might look like this:
01_Script
02_Characters
03_Sets
04_Keyframes
05_Video_Generations
06_Audio
07_Final_Selects
08_Final_Edit
This may feel excessive for a 30-second video.
It is not.
Once you start generating assets, organization becomes part of the creative process. The person with the cleanest workflow will often move faster than the person with the best prompt.
7. Audio is still one of the hardest parts
AI video is improving quickly, but Jacob was clear that audio remains tricky.
Voice consistency is hard. Environmental sound is hard. Matching dialogue to a realistic scene is hard. Tools like ElevenLabs can create high-quality voices, but those voices often sound like they were recorded in a clean studio. That can be a problem if the scene is supposed to take place in a diner, street, desert, office, or crowded room.
That is why Jacob often prefers to separate music from the rest of the audio.
For example, he may prompt a video model to create environmental sound, foley, and voices, but not music. Foley means the everyday sound effects in a scene: footsteps, doors, dishes, clothing, machines, and movement.
Then he adds music later so it remains consistent across shots.
That is a smart distinction.
If you let every clip generate its own music, your final video may feel stitched together. The shots may look connected, but the soundtrack will fight itself.
8. The best way to learn AI video is to make something
Jacob came from a traditional filmmaking background, but his advice for people without that background was refreshingly simple: make something.
You can read books about film language. You can watch tutorials. You can study camera angles, lighting, editing, and sound design.
But at some point, you need to create.
Jacob likes contests because they create a deadline. A deadline forces decisions. It prevents endless tinkering. It teaches you what actually works.
That is probably the best advice for business owners too.
Do not start with a full brand campaign.
Start with a 30-second product scene.
Start with a customer story.
Start with a short event recap.
Start with a founder introduction.
The first one will be rough. That is fine. The goal is not perfection. The goal is to build muscle.
Final takeaway
AI video is not replacing creativity. It is compressing the distance between idea and execution.
Jacob Rangel’s workshop made one thing clear: the future of video belongs to people who can combine taste, process, experimentation, and organization. The tools are powerful, but the workflow is what turns them into something useful.
My recommendation: pick one small video idea, create the keyframes first, organize your assets from day one, and use AI as your production partner—not your creative replacement.
