Up to 5 reference images for guided edits
Supply reference images to guide the edit — a photo of a specific outfit, a target background, or a character face. HappyHorse maps each reference to the relevant element in the source video.
HappyHorse Video-to-Video Editing
Video-to-video (v2v) editing takes an existing video clip as input and applies targeted modifications guided by text instructions and optional reference images. Unlike text-to-video (which generates from scratch), v2v preserves the original clip's motion, timing, and spatial layout while changing specified elements — a character's clothing, the background environment, or the visual style. The source video acts as a motion scaffold that constrains the generation.
What you can do
Local edits: change specific objects
Target individual elements without affecting the rest of the frame. Change a character's shirt color, replace a prop on a desk, or swap a logo on a wall — the surrounding scene stays untouched.
Global edits: style and environment transfer
Apply scene-wide transformations: convert daytime footage to night, apply an anime or oil-painting style, or shift the entire color palette. The motion path and character actions remain identical.
Motion preservation from source video
The source clip's motion trajectory, camera movement, and character poses are extracted and used as constraints. Your edit changes the appearance but not the choreography of the original footage.
Natural language edit instructions
Describe edits in plain English: "Change the red dress to a blue business suit" or "Replace the office background with a tropical beach." No mask drawing, keyframing, or timeline editing required.
Get started
How to use
1
Open PonPon Video and select HappyHorse
Go to PonPon Video and select HappyHorse from the model dropdown. Switch to the video-to-video editing mode.
2
Upload your source video
Upload the video clip you want to edit. Supported formats include MP4 and MOV. Shorter clips (3–10 seconds) produce the most consistent results.
3
Add reference images (optional, up to 5)
If your edit involves specific visual targets — a particular outfit, a character's face, a background photo — upload them as reference images. Each reference guides the model toward your intended result.
4
Write your edit instruction
Describe what you want changed. Be specific about which elements to modify and which to keep: *"Replace the character's casual T-shirt with a formal navy blazer. Keep the background and lighting unchanged."*
5
Generate and compare with the original
Click Generate and compare the edited output with your source clip side by side. Check that the motion matches and the edit was applied consistently across all frames.
Built for creators
Whether you're a solo creator, an agency, or a brand — every model adapts to how you work.
Copy & use
Prompt templates
Wardrobe swap
Change the character's white T-shirt and jeans to a formal black tuxedo with a bow tie. Keep the same walking motion, background, and lighting. Maintain the character's face and hairstyle exactly.
Model: HappyHorse · Mode: Video-to-video · References: 1 (tuxedo photo) · Source: 6s clip
Background replacement
Replace the indoor office background with an outdoor rooftop terrace overlooking a city skyline at sunset. Keep the character's appearance, position, and gestures identical. Match the lighting to golden hour.
Model: HappyHorse · Mode: Video-to-video · References: 1 (rooftop photo) · Source: 8s clip
Style transfer to anime
Convert this live-action clip to detailed anime style — Studio Ghibli-inspired with soft watercolor backgrounds and cel-shaded characters. Preserve all motion, facial expressions, and timing exactly.
Model: HappyHorse · Mode: Video-to-video · References: 0 · Source: 5s clip
Character swap with reference
Replace the person in this clip with the character from [person1] reference image. Keep the exact same body movement, gestures, and scene. Match the new character's clothing to the original scene's style.
Model: HappyHorse · Mode: Video-to-video · References: 1 (face photo) · Source: 6s clip
Who it's for
Use cases
Wardrobe changes without reshooting
Fashion brands and content creators can show the same scene with different outfits. Film once, then use v2v editing to generate variations — a model walking in three different dresses from a single source clip.
Background replacement for location flexibility
Swap a plain studio background for a beach, a cityscape, or a branded set. Useful for real estate virtual staging videos, travel content previews, or shooting on a budget without location access.
Style transfer for creative campaigns
Convert live-action footage to anime, watercolor, or cyberpunk aesthetics for music videos, social campaigns, or pitch decks. The original performance and timing stay intact.
Character face replacement with references
Swap a character's face using a reference photo while preserving their body movement and scene context. Useful for personalizing template videos for different clients or audiences.
Compare
HappyHorse Video Editing vs Alternatives
| HappyHorse V2V Editing | Other V2V Tools | |
|---|---|---|
| Reference images | Up to 5 reference images to guide the edit — outfit photos, face refs, background targets | Kling O3: v2v editing with reference support but fewer simultaneous refs |
| Edit precision | Both local (single object) and global (full scene) edits via natural language | Many tools require mask drawing for local edits — more precise but slower |
| Motion preservation | Source video motion fully preserved — edits change appearance only | Some tools partially re-generate motion, causing timing drift on longer clips |
| Mask requirement | No masks needed — describe the edit target in text | Photoshop/After Effects plugins require manual mask drawing per frame or region |
| Best for | Quick creative variations: wardrobe swaps, background changes, style transfers | Kling O3: finer control for complex compositing. Manual tools: pixel-precise corrections |
Get the best results
Tips & best practices
Keep source clips under 10 seconds for best consistency
Longer source videos increase the chance of temporal inconsistencies in the edit. For best results, use 3–10 second clips. For longer sequences, split into segments and edit each separately.
Specify what NOT to change
Explicitly state which elements should stay the same: "Keep the character's face, hairstyle, and background unchanged." Without this, the model may apply broader changes than intended.
Complex edits may require multiple passes
Changing both the wardrobe and background in one pass can reduce quality. For complex transformations, do one edit at a time — swap the outfit first, then use the output as input for the background change.
V2V editing cannot fix source video problems
If the source clip has motion blur, compression artifacts, or low lighting, the edit will inherit those issues. Start with clean, well-lit source footage for the best edited output.
Community
Loved by creators worldwide
Join thousands of creators, agencies, and brands who use PonPon every day.
FAQ
Questions & answers
What is video-to-video editing?
Video-to-video (v2v) editing takes an existing video clip as input and modifies specific elements — clothing, backgrounds, visual style — while preserving the original motion and timing. It's different from text-to-video, which generates entirely new footage from a text prompt.
How many reference images can I use for v2v editing?
HappyHorse v2v editing supports up to 5 reference images. Use them to show the model what you want — a target outfit, a character face, or a background environment. For text-to-video with more references, HappyHorse supports up to 9 reference images.
Does the motion change when I edit a video?
No. HappyHorse preserves the source video's motion trajectory, camera movement, and character poses. Only the visual appearance changes according to your edit instructions.
Can I change just one object in the scene?
Yes. Describe the specific element to change: "Replace the red backpack with a leather briefcase." HappyHorse applies the change locally without affecting the rest of the frame. No mask drawing required.
How does HappyHorse v2v compare to Kling O3?
Kling O3 also offers video-to-video editing capabilities. HappyHorse supports more simultaneous reference images (5 vs fewer) and uses natural language instructions without masks. Kling O3 may offer finer compositing control for complex multi-layer edits.
What video formats can I upload as source?
MP4 and MOV are supported. For best results, use clips with stable exposure, minimal compression artifacts, and a resolution of at least 720p. The output resolution matches the generation settings, not the source resolution.
Is HappyHorse video editing free?
Free daily credits on PonPon cover HappyHorse video editing. V2v editing uses the same credit rate as standard HappyHorse generation. See pricing for details.
Can I combine v2v editing with multi-reference generation?
V2v editing supports up to 5 references. For scenes requiring more character references (up to 9), use HappyHorse's text-to-video multi-reference mode to generate the initial clip, then edit it with v2v if needed.