Higgsfield + Seedance 2.5 Edit

Create Realistic AI TikTok Shop Videos

Turn a motion reference + character image into a realistic AI video using the Titans workflow.

Video = MotionMovement, camera, framing, timing, and performance.
Image = IdentityFace, hair, body, clothing, product, and character look.
Prompt = RulesWhat must stay locked and what gets replaced.
@Video1 Motion
@Image1
Character reference image used as Image1
Result Locked
Prompt hierarchy
START HERE

The 5-step Titans workflow.

Follow the steps in order so the motion reference, character image, Higgsfield setup, prompt, and final review all work together.

01

Get Motion

Record your own reference video or use an existing reference clip you have permission to use.

02

Get Character

Use an existing character/reference photo or create a new character with the Titans Character Builder.

03

Open Higgsfield

Open Higgsfield and select the recommended Seedance workflow.

Open Higgsfield
04

Attach + Prompt

Attach the motion video as @Video1, attach the character image as @Image1, then generate your prompt below.

05

Generate

Create the video, review the result, make adjustments if necessary, then export.

Follow the order Motion video → Character image → Higgsfield → Prompt → Generate
Walkthrough video

Watch the workflow once.

See how the Titans process is used from the motion reference to the character image, prompt, and final video setup.

Step 1

STEP 1 - Choose Your Motion Reference

The reference video controls the movement blueprint. It does not need to look like your final character.

Movement Camera movement Framing Timing Body positioning Performance
Option A

Record It Yourself

Best for maximum control and originality. Record the exact motion, timing, camera behavior, and framing you want the AI character to perform.

Option B

Existing Reference Video

Use an existing reference clip that you have permission or rights to use as the motion reference.

Download a TikTok Reference Video

Paste a public TikTok video link to get the highest-quality clean version available.

Option B Tool

The person in the motion video does not need to look like your final character. The motion video primarily controls movement and camera behavior.
Good reference video

Clean & Stable

The camera stays still, the subject is sharp and well lit, and the background is fixed.

GOOD ✓ Stable camera Good lighting Sharp subject Fixed background
Bad reference video

Shaky & Unclear

Shaky camera, blur, changing lighting, and shifting framing give the model too much unstable information to track.

BAD × Shaky camera Poor lighting Motion blur Changing framing
Step 2 / Character image

STEP 2 - Choose or Create Your Character

Your image controls who appears in the final video. Use a clean existing reference image or create a new Image1 character below.

Option A

Use an Existing Image

Upload a clean reference image of the person or character you want used.

Option B

Create a New Character

Use the Titans Character Builder below to generate a new character reference.

Basic

Use adults only. The prompt will always say adult.

This controls compatible hair, presentation, and outfit wording.

If the product is a hat, shirt, jacket, glasses, jewelry, bag, or anything worn/held, describe it here.

Advanced Options

Use avatar mode when you already have a character image and want to match another reference image's lighting, scale, and phone angle.

Only used when Custom identifier is selected.

Generated image prompt
Body-
Face-
Hair-
Eyes-
Skin-
Outfit-

          

            Product sellers: be specific before you copy.
            If the image is meant to sell or show a product, put the exact product/outfit requirement in the field on the left. For hats, shirts, jackets, glasses, jewelry, bags, or uniforms, the product needs to be named directly or the image model may invent the wrong item.
          

          

            
Use the start frame.This gives the image model the exact original pose and angle.

            
Image first.This prompt creates Image1. The Seedance prompt below handles the video replacement.

            
Reroll details.Same age, gender, and profession can produce different realistic combinations without leaving that character context.

          

          

            age
            gender
            profession
            weighted body
            compatible face
            age-compatible hair
            random eyes
            outfit
            product detail
          

        

      


    


    

      

        
Character examples

        

          
Good vs Bad Character References

          

            Strong Image1 references are simple, realistic, and easy for the
            model to keep consistent across frames.
          

        

      


      

        

          

            Adult male mechanic-style worker character example on a clean studio background
          

          

            Good reference
            
Mechanic-style worker

            

              Clear subject
              Simple readable outfit
              Realistic lighting
            

          

        


        

          

            Adult woman in a light-blue outfit character example on a clean studio background
          

          

            Good reference
            
Light-blue outfit

            

              Clean silhouette
              Easy clothing shape
              Natural skin texture
            

          

        


        

          

            Adult blonde college woman character example on a clean studio background
          

          

            Good reference
            
Blonde college woman

            

              Distinct identity
              Consistent studio lighting
              Usable outfit detail
            

          

        

      


      

        
Bad character examples

        

          These can look impressive as still images, but they are harder for
          video models to keep stable because they contain too many tiny
          details, complex hands, layered clothing, and elements that can
          flicker or change frame-to-frame.
        

      


      

        

          

            Low-quality adult blonde woman character example with off proportions and obvious AI artifacts
          

          

            Bad reference
            
Low-res and AI-looking

            

              Low-quality source image
              Poor proportions
              Obviously AI-generated appearance
            

          

        


        

          

            Over-detailed adult male character example with heavy dirt, straps, pockets, zippers, and difficult hands
          

          

            Bad reference
            
Texture overload

            

              Too many straps and zippers
              Complicated hands
              Heavy texture can flicker
            

          

        


        

          

            Over-detailed adult blonde woman character example with layered clothing, jewelry, hair details, and difficult hands
          

          

            Bad reference
            
Hard to keep consistent

            

              Excess jewelry and layers
              Tiny lace and hair details
              Hands/fingers are difficult
            

          

        

      

    


    

      

        
Step 3 / Higgsfield setup

        

          
STEP 3 - Set Up Higgsfield / Seedance

          

            Load the motion video and character image before you paste the
            Titans prompt. Beginners should match this setup first.
          

          Open Higgsfield / Seedance
        

      


      

        

          

            1
            

              Open Higgsfield
              
Start in the video editing flow.

            

          

          

            2
            

              Select Seedance / recommended editing model
              
Use the recommended Seedance edit workflow for this helper.

            

          

          

            3
            

              Add motion reference as @Video1
              
This controls movement, camera, framing, timing, and performance.

            

          

          

            4
            

              Add replacement reference as @Image1
              
This controls the defining appearance of the selected person, outfit, vehicle, product, or environment.

            

          

          

            5
            

              Keep sound/audio generation off during testing
              
Focus first on clean motion, face, hands, product, and clothing.

            

          

          

            6
            

              Paste the generated Titans prompt
              
Use the Step 4 builder below after both references are attached.

            

          

          

            7
            

              Generate
              
Review the output before exporting or editing.

            

          

        


        

          

            Higgsfield
            →
            Edit Video
            →
            Seedance 2.5 Edit
          

          

            

              Higgsfield Seedance 2.5 Edit setup showing prompt mode, uploaded video, uploaded image, 1080p, high bitrate, and sound off
            

            

              Model: Seedance 2.5 Edit
              Mode: Prompt
              Resolution: 1080p
              Bitrate: High
              Sound: Off
            

          

        

      

    


    

      

        
Step 4 / Video prompt

        

          
STEP 4 - Build Your Video Prompt

          

            The references provide the motion and identity. The prompt tells
            the model what must remain consistent and what should change.
          

        

      


      

        

          

            
            
          


          

            
            
          


          

            
            
            
Add a concrete detail only when the replacement needs it.

          


          

            
            
          

        


        
      

    


    

      

        
Step 5 / Generate + review

        

          
STEP 5 - Generate & Review

          

            Run the generation, then check the failure points before exporting
            or posting.
          

        

      


      

        

          
Before export, check this.

          

            Identity stayed consistent
            Hands and face look natural
            Clothing/product stayed consistent
            Motion matches the reference
            Camera movement stayed accurate
            No flickering or warping
            Background stayed believable
            Product details are correct
          

        


        

          
Final Edit

          

            Run the finished video through your normal editing workflow such as
            CapCut to trim the clip, add captions, sound, overlays, branding,
            or other final edits.
          

          

            Download result
            Edit in CapCut
            Prepare for posting
          

        

      

    


    

      

        
Examples / Reference to result

        

          
Reference → Image → Final Result

          

            The point is simple: this movement plus this character image becomes
            the final AI video when the prompt keeps the hierarchy locked.
          

        

      


      

        

          

            
Person Replacement Case Study

            Complete example
          

          

            

              
Motion Reference

              
            

            
            

              
Character Image

              Character image used for the case study
            

            
            

              
Final AI Video

              
            

          

        

      


      

        

          More results made with this workflow.
          
These clips stay secondary so the page teaches the process first and shows extra proof after.

        

        

          Same core idea.
          
Preserve the motion video, replace only the selected subject or object, and keep the final clip stable.

        

      


      

        

          
          
Result01

        

        

          
          
Result02

        

        

          
          
Result03

        

        

          
          
Result04

        

        

          
          
Result05

        

        

          
          
Result06

        

      

    


    

      

        
Final quality checklist

        

          
Use it only when the reference hierarchy holds.

          

            The final export should look like the same reference clip with a
            clean replacement. If identity, hands, face, clothing, or background
            drift frame-to-frame, adjust the references before over-prompting.
          

        

      


      

        

          Reject
          
Run again when the output changes the scene.

          
Background shifts, camera angle changes, clothing morphs, face flickers, hands melt, or timing no longer matches the original video.

        

        

          Approve
          
Use it when the reference hierarchy holds.

          
The motion, camera, timing, framing, and scene match the reference video while the replacement identity stays stable from Image 1.

        

      

    


    

      

        
Advanced / Reference rules

        

          
Understanding Your References

          

            This is the deeper reason the workflow works. The uploaded
            references already carry most of the information. Your prompt should
            enforce the hierarchy and stop the model from changing the wrong
            things.
          

        

      


      

        

          
Video controls

          
Motion

          
Body movement, gestures, posture, head turns, facial timing, lip movement, pacing, and scene behavior.

        

        

          
Video controls

          
Camera

          
Camera angle, framing, movement, perspective, subject scale, composition, background, and timing.

        

        

          
Image controls

          
Identity

          
Face, hair, body proportions, skin tone, clothing, colors, products, and overall person appearance.

        

        

          
Prompt controls

          
Boundaries

          
Only replace the intended subject. Do not restage, redesign, improve, reinterpret, or change the scene.