How I Created a Fantasy TV Show (Complete AI Video Workflow)
- AIartkingdom

- Jun 24
- 5 min read
Updated: Jun 26
The AI Video Workflow Using Grok, Seedance 2.0, ChatGPT, and CapCut

AI video generation is evolving at an incredible pace, but one of the biggest challenges creators still face is consistency.
Anyone can generate a beautiful AI video clip.
Creating multiple clips that look like they belong in the same world is an entirely different challenge.
Recently I set out to create a short fantasy film called The Wizard's Plug, a deliberately campy, practical-effects-inspired fantasy short designed to look like a forgotten television special from the late 1970s.
The project was created using a combination of:
Grok
Captain HaHaa's AI video workflow tutorials
ChatGPT
Seedance 2.0
CapCut
What follows is the complete workflow I used and the lessons I learned along the way.
Step 1: Start With Style Before Story
One of the biggest mistakes AI creators make is beginning with a story before establishing visual rules.
Most AI generators naturally drift toward:
modern fantasy art
cinematic blockbuster lighting
glossy CGI
concept art aesthetics
None of those matched the project.
Instead, the entire production was locked to a specific style. After coming across Captain HaHaa's X account and seeing all of the amazing work he was doing in AI video I was awestruck. His account was a wealth of information as well as inspiration, so I set out to make my own Captain HaHaa style video :
Visual Rules
1970s television fantasy
practical creature suits
practical costumes
practical smoke effects
miniature-level production values
slightly theatrical performances
slightly campy tone
handmade practical sets
35mm film grain
Explicit Negative Rules
no blockbuster cinematography
no epic fantasy
no glossy CGI
no concept art aesthetics
no modern fantasy illustration
This style guide became the foundation of every image and every video prompt throughout the entire project.
Once the visual language was established, the story became much easier to develop.
Step 2: Design Consistent Characters
Before generating any video, I created character reference images.
Zephyrion

An eccentric old wizard with:
long white beard
blue robes
pink glasses
classic wizard hat
Rather than looking like Gandalf from a modern fantasy film, the goal was to make him look like a character from an old television production.
Mosswald

A swamp creature inspired by:
moss
foliage
practical monster suits
mascot-style performances
The key was avoiding realistic monster design.
Instead, Mosswald needed to feel like a physical creature suit that an actor could actually wear. My biggest visual inspiration for this character was literally a moss covered version of McDonald's Grimace.
This decision ended up becoming one of the most important creative choices in the project.
Step 3: Build the World First
Before generating any animation, I created a dedicated environment image.

The setting was:
mystical swamp clearing
dense foliage
hanging moss
shallow swamp water
practical fog effects
fallen log
The environment remained visually consistent throughout the production.
This helped every generated shot feel like it belonged to the same physical location.
Step 4: Learn From Existing AI Video Creators
One of the most useful parts of the process was studying the workflow shared by creator Captain HaHaa.

Rather than trying to generate an entire film in a single prompt, the workflow focused on building scenes from individual shots.
This is much closer to how real filmmaking works.
Instead of generating:
"Make me a movie."
The process becomes:
"Create this shot."
Then:
"Create the next shot."
Then:
"Create the reaction shot."
This mindset completely changes the quality of the final result.
Step 5: Plan the Film Like a Real Production
The final film was broken into seven individual video shots.
Shot 1
Wizard enters swamp.
Discovers sleeping Mosswald.
Sits beside him.
Shot 2
Close-up of Zephyrion.
He appears to have an idea.
Shot 3
Close-up of Mosswald waking up.
Shot 4
Wizard plucks moss.
Packs pipe.
Smokes pipe.
Shot 5
Wizard exhales smoke.
Smoke rolls across swamp water.
Shot 6
Water nymph appears from smoke.
Shot 7
Wizard laughs.
Mosswald claps happily.
Title card.
Breaking the project into individual scenes made the entire production manageable.
Step 6: Why Seedance 2.0 Was the Right Tool
For this project I used the Seedance 2.0 video model.
Seedance offers several advantages:
strong image-to-video generation
start frame support
end frame support
excellent consistency
up to 15-second generations
Most importantly, Seedance allows creators to think like filmmakers.
Instead of describing an entire sequence, you can define:
Start Frame
Where the shot begins.
End Frame
Where the shot ends.
The model then creates the motion between them.
This dramatically improves control.
Step 7: Use Start and End Frames
This was probably the single most important lesson from the project.
Whenever possible:
Create a start image.
Create an end image.
Then generate the motion between them.
For example:
Start Frame
Wizard preparing to exhale.

End Frame
Smoke drifting across water.

Seedance handled the transition naturally.
The results were significantly more consistent than relying on text prompts alone.
Step 8: Use Close-Ups to Hide AI Limitations
One of the challenges with AI video is that characters can sometimes jump between positions.
The solution came from traditional filmmaking.
Insert close-up shots.

Instead of forcing a complex movement sequence:
Cut to a character reaction.
Cut to a facial expression.
Cut to an object.
Then return to the main scene.
These inserts dramatically improved pacing while hiding generation inconsistencies.
Step 9: Edit Everything Together in CapCut
After generating the seven video clips, everything was assembled inside CapCut.
The editing process was intentionally simple.

Cuts
Straight cuts between scenes.
VHS Effects
Light analog degradation.
Subtle tape noise.
Minor image softness.
Nothing excessive.
The goal was authenticity rather than parody.
Step 10: Add Music Last
The final addition was music.
This ended up transforming the project.
Music:
smooths transitions
establishes tone
hides minor imperfections
creates emotional continuity
Without music the film felt like a collection of clips.
With music it felt like a complete short film.
What Worked Best
Several discoveries stood out throughout the process.
Style Consistency Beats Visual Quality
Many creators chase higher resolution.
Consistency matters far more.
A perfectly coherent 1970s fantasy world is more convincing than seven technically superior clips that all look different.
Practical Effects Are Easier For AI
AI models often struggle with realistic cinematic imagery.
Interestingly, they perform extremely well when asked to imitate:
creature suits
miniatures
practical effects
stage productions
Imperfections actually enhance the illusion.
Small Stories Work Better
Instead of creating a ten-minute epic, focus on a one-minute story.
The Wizard's Plug succeeds because it tells one simple joke from beginning to end.
Final Thoughts

The most important lesson from this project is that AI video creation is becoming less about prompting and more about filmmaking.
The tools are improving rapidly.
But the creators producing the most interesting work aren't necessarily writing better prompts.
They're thinking more like directors.
They're planning shots.
Building worlds.
Creating characters.
Designing visual rules.
And editing everything together into a cohesive story.
Seedance, ChatGPT, Grok, and CapCut were simply the tools.
The real breakthrough came from treating AI video generation like a film production rather than a prompt generation exercise.
If you're looking to create more convincing AI films, start small.
Pick a style.
Create a world.
Build memorable characters.
Then generate one shot at a time.
You might be surprised how quickly a handful of clips turns into a complete film.
Explore, create, and escape today.* Ready to create? Join our AI Art Kingdom Club or read our Guide to Mastering AI Art Prompts.
If you'd like to know more you can head over to AIArtKingdom.com for a curated collection of today's most popular, most liked AI artwork from across the internet. Plus explore an extensive array of AI tools, complemented by comprehensive guides and reviews, on our AI blog.






