Step-by-Step Guide to Creating Realistic Video Ads with Higgsfield
This article is a continuation of our AI model series. At this stage, we assume that your AI character (face/identity) has already been created and is visually consistent.
Why Video Matters for AI Characters in Marketing
Static AI images are useful for visuals, but they are not enough for product promotion.
Affiliate marketing, especially with platforms like Amazon Associates, relies on:
- believable interaction with products
- natural hand movement
- short, calm lifestyle videos
To achieve this, you need controlled motion, not imaginative animation.
Higgsfield enables this by copying real human motion from reference videos.
Higgsfield Tools Used in This Workflow (Exact Names)
| Goal | Tool Name |
|---|---|
| Lock character identity | Soul ID Character |
| Add product to hands | Inpaint |
| Copy real motion | Kling Motion Control |
| Fix face drift (optional) | Face Swap |
| Make character speak (optional) | LipSync Studio / Talking Clips* |
* Availability depends on plan and model version.
Step 1: Confirm Character Identity (One-Time Setup)
Path:Image → Soul ID Character
Upload 1–3 images of your AI character and create a Soul ID.
This anchors facial identity across images and videos.
If your Soul ID already exists, skip this step.
Step 2: Add a Product to the Hands (Image Stage)
This step is mandatory.
If the object looks fake in a still image, video generation will fail.
Path:Image → Inpaint
Actions:
- Upload a clean image of your AI character
- Brush-select both hands and the space between them
Prompt example:
a plain ceramic coffee mug held naturally with both hands
Rules:
- describe only the object
- no emotions, no marketing language
Your goal is a realistic static image with correct hand-object interaction.
Step 3: Generate Video Using Real Motion
Path:Video → Kling Motion Control
Do not use:
- Create Video
- Cinema Studio Video
Those tools invent motion and often break realism.
Inside Kling Motion Control
Base Image:
Upload the AI image created in Step 2 (with the product already in hands).
Reference Video:
Upload a real video of a person holding or using a similar object.
Prompt (minimal):
natural realistic movement, casual product interaction
Enable motion from reference.
Avoid camera motion options.
Where to Get Good Reference Videos (Critical Section)
The quality of your output depends more on the reference than the prompt.
1. Stock Video Platforms (Recommended)
Free:
- Pexels
- Pixabay
Search examples:
- “woman holding coffee mug b roll”
- “skincare routine hands”
- “lifestyle hands close up”
Choose videos where:
- the person is silent
- the mouth does not move
- movement is slow and simple
2. Pinterest (Surprisingly Effective)
Pinterest is excellent for lifestyle reference, especially:
- morning routines
- coffee moments
- skincare application
How to use:
- search for short video pins
- choose clips with no speech
- use only 3–6 seconds as reference
Avoid:
- influencer talking pins
- branded tutorials
Pinterest works well because movements are often calm and repetitive.
3. YouTube (With Caution)
Look for:
- silent vlogs
- routine videos without talking heads
Trim a short fragment and remove audio.
Do not use recognizable creators or branded content.
4. Recording Your Own Reference (Best Control)
If possible, record your own reference video:
- phone camera is enough
- neutral background
- slow movement
- closed mouth
This gives the cleanest motion and avoids lip movement issues entirely.
Common Problem: The Character Moves Their Lips (Even Without Audio)
This is a known issue and not a prompt mistake.
Why It Happens
Video models often assume:
face in frame = speaking person
Especially if:
- the reference video has lip movement
- the face is centered
- the head subtly nods
How to Stop Unwanted Lip Movement (Proven Fixes)
1. Fix the Reference (Most Important)
Use only reference videos where:
- the mouth stays closed
- no speech occurs
- no expressive facial movement
Even minimal lip motion will be copied.
2. Add Explicit Restrictions to the Prompt
Add this line in Kling Motion Control:
no speaking, no lip movement, closed mouth
Remove words like:
- expressive
- engaging
- friendly
3. Avoid Audio-Based Models
Do not use:
- Kling 3.0 (with audio)
- any model labeled “with sound”
These models are optimized for speech.
4. Occupy the Mouth (Practical Trick)
Have the character:
- take a sip
- hold the product close to the mouth
This significantly reduces lip animation.
What If You WANT the Character to Speak?
This requires a different workflow.
When Lip Sync Makes Sense
- explainer videos
- UGC-style ads
- voice-over content
Tools Used
- LipSync Studio
- Talking Clips (if available)
Basic Workflow
- Generate the base video first (without speech)
- Open:
Video → LipSync Studio - Upload:
- the generated video
- audio file (voice-over)
Use short scripts and neutral tone.
Overacting breaks realism quickly.
⚠️ Important:
Do not mix lip sync and motion control at the same time in the first generation.
Always add speech after motion is finalized.
Amazon Affiliate Use Case (Lifestyle Positioning)
Your AI character should not claim personal experience.
Correct positioning:
- lifestyle inspiration
- visual usage context
Example caption:
Morning routine essentials ☕
Link in bio
Affiliate links go through:
- Linktree
- Beacons
- personal landing page
Final Workflow Summary
- Confirm Soul ID
- Add product with Inpaint
- Generate motion via Kling Motion Control
- Fix lips / face only if needed
- Add lip sync only when intentionally required
Key Takeaway
AI characters work best when:
- motion is copied, not imagined
- prompts stay minimal
- realism is prioritized over “cinematic” style
They are scaling tools, not shortcuts.
