Getting the Best Results: Photo & Video Tips

The quality of your AI-generated UGC video depends heavily on the inputs you provide. A great photo and well-chosen reference video can make the difference between a video that looks incredible and one that falls flat. Here's everything you need to know.

The Perfect Character Photo

Do's

• Clear, well-lit face. Natural light or soft studio lighting produces the best results. The AI needs to clearly see facial features to generate accurate multi-angle views.
• Neutral background. A plain white or light-colored background helps the AI focus on the character rather than complex surroundings.
• High resolution. Aim for at least 512x512 pixels. Higher resolution means more detail for the AI to work with.
• Natural pose. A relaxed, natural standing pose with arms visible works best for full-body character generation.
• Single person. Make sure only one person is in the frame.
• Full body when possible. While close-up photos work, full-body shots give the AI more information about clothing and body proportions.

Don'ts

• Avoid heavy filters or editing. Instagram filters, heavy retouching, or artistic effects can confuse the AI's understanding of the character.
• Avoid hats or sunglasses. Anything that obscures the face reduces quality. The AI needs to see hair, forehead, and eyes clearly.
• Avoid group photos. Even if you crop to one person, residual artifacts from other people can affect results.
• Avoid motion blur. Sharp, in-focus images only.
• Avoid extreme angles. A photo taken from directly above or below will produce worse results than a standard eye-level shot.

Choosing the Right Reference Video

What works well

• Single person clearly visible. The reference video should feature one person performing clear, distinct movements.
• Steady camera. Minimal camera shake or movement produces better motion transfer.
• Good lighting. Well-lit scenes with consistent lighting throughout the video.
• Clear body movements. Dancing, gesturing, walking, talking — movements where the body is clearly visible.
• Consistent scene. Videos that stay in one location without cuts or scene changes.

What to avoid

• Multiple people. The AI may get confused about which person's motion to follow.
• Rapid scene changes. Videos with cuts, transitions, or fast camera movements don't work well.
• Very dark scenes. Low-light videos produce inconsistent results.
• Extreme close-ups. If only a face is visible with no body, the motion control has limited data to work with.
• Text overlays or watermarks. These can interfere with motion extraction.

Duration and Credits

Each 5-second block of video costs 1 credit:

Duration	Credits
1-5 seconds	1 credit
6-10 seconds	2 credits
11-15 seconds	3 credits
16-20 seconds	4 credits
21-25 seconds	5 credits
26-30 seconds	6 credits

Start short. We recommend starting with 5-second videos to test photo/video combinations before committing credits to longer durations.

Troubleshooting Common Issues

"My character doesn't look like the photo"

This usually happens with low-quality or heavily filtered photos. Try uploading a clearer, unedited photo with good lighting.

"The video was blocked by safety filters"

Our safety systems may occasionally flag legitimate content. Try a different photo (well-lit, neutral background) or a different reference video. If the issue persists, contact support.

"The movements don't match the reference"

This can happen with very complex or fast movements. Try a reference video with slower, more deliberate movements.

"The generation is taking too long"

Most videos complete in 3-8 minutes. During peak hours, it may take slightly longer. Pro and Agency plans have priority queue access for faster processing.

Ready to create? Upload your photo and video now.