Want to turn a simple horror idea into an AI-generated short film using your own face? In this complete beginner-friendly guide, I’ll break down the exact workflow from the video tutorial: idea → script → character reference sheets → AI video → Filmora editing → final horror reveal.
The goal is not just to generate a random AI clip. The important part is keeping the characters consistent, using the script as the video direction, testing cheaply before spending credits, and then finishing the result with color and sound in Filmora.
🎬 Watch the Full Tutorial
What We’re Creating
The concept is a short K-drama-inspired horror scene: a character stands completely still, another character watches, and then a disturbing smile reveals a red mark on the face.
For the tutorial, the workflow is built around Filmora.tv, character reference sheets, Seedream 5.0 Pro for the character-image generation step, and Seedance 2.5 for the 30-second AI video generation. The final polish is done in Filmora 16.
The Complete Workflow
- Develop the horror idea
- Create the script
- Build consistent character reference sheets
- Generate the AI video
- Edit the result with color, film grain, and sound
Before You Start
You’ll want a few things ready:
- Your own reference photo for the main character.
- A reference image for the second character.
- Access to Filmora.tv for the node-based workflow.
- Filmora 16 for the final edit.
- Your script idea or scene concept.
Important: The tutorial uses a normal Korean girl photo found on Pinterest as the reference for the second character. Make sure you have the appropriate rights or permission to use any reference image you choose. Also, the tutorial avoids using clips from the movie itself.
Step 1: Start With the Idea and Script
Before generating anything, decide exactly what you want the scene to feel like.
The tutorial starts from a horror idea inspired by a scene the creator remembered from The 8th Night. Instead of using footage from the movie, the concept is recreated with original AI-generated visuals.
The first practical step happens inside filmora.tv:
- Open Filmora.tv.
- Create a new project.
- Name the project “K Drama Horror.”
- Click the + icon.
- Add a text node.
- Write the scene you have in your head.
- Use GPT 5.6 to turn that rough idea into a complete script.
If you are completely new to node-based workflows, think of each node as one building block. You connect the blocks together and the final result comes from the complete chain.
Step 2: Create Character Reference Sheets
This is one of the most important parts of the entire workflow.
If you take a random selfie and immediately send it into an AI video generator, the character can change from frame to frame. Facial features, hair, clothing, body proportions, and even the overall identity may drift.
That is why the tutorial creates a character reference sheet first.
Why a Character Sheet Helps
A reference sheet gives the image and video model multiple views of the same character. Instead of asking the AI to guess what the person looks like from one photograph, you give it a clearer visual reference.
The supplied references demonstrate this approach with multiple views, including full-body and portrait/profile views. The goal is simple: lock in the character before generating the movie scene.
Main Character Reference Prompt
Use the following prompt exactly as provided:
Character Reference sheet: Create a professional character reference sheet based strictly on the uploaded reference image. Use a clean, neutral plain background and present the sheet as a technical model turnaround while matching the exact realistic visual style of the reference. Arrange the composition. two full-body standing views – front, left profile. one close-up portrait – front. Maintain perfect identity consistency across every panel. Keep the subject in a relaxed A-pose with consistent scale and alignment, accurate anatomy, and clear silhouette. Lighting should be consistent across all panels. Output a crisp, ultra-realistic, print-ready reference sheet.
How to Use the Character Sheet in Filmora.tv
- Place your reference photo on the Filmora.tv canvas.
- Connect it to an image generation node.
- Use Seedream 5.0 Pro for the character-image generation step.
- Use the character reference prompt above.
- Generate the reference sheet.
- Keep the resulting sheet available for the later video-generation step.
Note: The Seedream prompt/settings are generated within Filmora.tv in this workflow, so there is no separate Seedream prompt that needs to be copied into this article.
Create the Second Character
The tutorial follows the same basic process for the girl character. A normal Korean girl photo is used as the reference, and a character sheet is generated from it.
The result is two locked character references: your main character and the girl character.
Step 3: Generate the AI Video
Now we move from still images to video.
This is where the tutorial uses a slightly different approach from many AI-video tutorials: you do not need to write a huge video prompt.
The Script Becomes the Video Direction
Remember the script created in Step 1? That script becomes the main scene direction.
Inside Filmora.tv:
- Add a video node.
- Connect the first character reference sheet.
- Connect the second character reference sheet.
- Connect the text node containing the script.
- Choose Seedance 2.5.
- Use the simple instruction below.
Create a horror short film scene using @image1 and @image2 with the script and scene direction from @textnode.
In the tutorial, @image1 and @image3 are the character reference sheets, while @image2 is the thumbnail/poster reference.
Make sure you connect the correct assets to the correct nodes in your own project. The exact node references may depend on how your Filmora.tv canvas is arranged.
Why Seedance 2.5?
The tutorial uses Seedance 2.5 because the script runs for about 30 seconds, and the workflow is designed around a video generation length of up to 30 seconds.
Do NOT Start With the Expensive Render
This is one of the most useful tips in the tutorial.
AI video generation can produce something completely different from what you imagined. If you immediately render at high quality, you can waste credits on a result you do not even want.
Instead, the tutorial uses:
- 21:9 aspect ratio
- 480p for the initial test
Generate the cheap test first. Watch it carefully. Check the character identity, scene direction, timing, camera behavior, and overall mood.
If the result is good, make small prompt adjustments if needed and then render the final version at 720p or 1080p.
Step 4: Edit the AI Video in Filmora 16
Getting a good AI video is not the end. The raw generation can still look flat, overly clean, or obviously AI-generated.
This is where the final edit makes a major difference.
1. Add a Horror Color Look
The tutorial uses Filmora 16's Halloween filters. There are multiple scary-style presets, so choose the one that fits your footage.
The goal is to move the footage away from a plain AI-generated look and toward a darker cinematic mood.
2. Add Film Grain
After applying the color treatment, add film grain. This helps the footage feel more like a finished film rather than a perfectly clean AI render.
3. Build the Horror With Sound
Sound is a huge part of the scare. A visually creepy shot can feel weak if it is completely silent.
The tutorial adds:
- Cricket sounds at the beginning, helping the silence feel natural.
- A mild jumpscare hit when the character is first revealed.
- A riser that builds tension until the red mark appears.
Optional: You can also add trending Instagram audio if it fits your content strategy and you have the appropriate rights to use it.
🎧 Try This Editing Test
Mute your finished video and watch the scare without sound. Then play the same section with the sound effects enabled. The difference makes it easy to understand why audio is so important in horror editing.
The Final Reveal
Once the edit is complete, play the final short film from beginning to end without interruptions.
The final reveal is built around the character's smile and the red mark. The creator's final line is:
How to Make the Ending Loop
The tutorial also uses a simple looping trick.
End on the frozen smile with the red mark, then hard-cut back to the same frame used for the opening hook. Because the beginning and ending visually match, the video can replay more seamlessly.
Complete Workflow Recap
If you are a beginner, here is the entire process in simple terms:
- Start with an idea. Decide what the horror scene should be about.
- Write the script. Use GPT 5.6 to turn the rough idea into a complete scene.
- Build character reference sheets. Give the AI a strong visual identity for each character before generating video.
- Connect everything in Filmora.tv. Use the character sheets and the script together in the video workflow.
- Test at low cost. Generate the first version at 480p and 21:9 before spending credits on the final render.
- Render the final video. Once the test works, move to 720p or 1080p.
- Finish in Filmora 16. Add the Halloween look, film grain, and sound effects.
- Build the reveal. Let the sound and visual timing lead into the red-mark reveal.
- Loop the ending. Match the final frame to the opening hook for a smoother replay.
The Most Important Lessons From This Workflow
1. Don’t rush into AI video
The character sheet is not an unnecessary extra step. It gives the video generator a much stronger reference to work from.
2. Your script can do more than tell a story
In this workflow, the script is also used as the main scene direction for the video generation step. That is the key idea behind the “five steps, zero prompts” approach.
3. Test before you spend
AI generation is unpredictable. A cheap 480p test can reveal problems before you spend more credits on a high-resolution render.
4. Editing still matters
AI generation gives you the raw material. Color, film grain, timing, and sound help turn that raw material into something that feels more like a finished short film.
5. Horror depends on timing
The scare is not only the red mark. It is the silence before it, the sound buildup, the reveal, and the exact moment everything changes.
Final Thoughts
You do not need to be an experienced filmmaker to experiment with this workflow. The process is surprisingly straightforward once you separate it into stages:
Idea → Script → Character Sheets → Video → Edit
The biggest takeaway is that the AI video itself is only one part of the process. A strong reference sheet helps maintain the character, the script provides the scene direction, low-resolution testing saves credits, and the final Filmora edit gives the result its cinematic and horror-focused finish.
And when everything comes together, you can reach the final punchline:
Note: This guide follows the workflow and prompts supplied for the tutorial. Tool interfaces, model availability, generation limits, credits, and settings can change over time.
0 Comments