Prepare Photos for an AI Video: A Practical Guide

Start with the finished scene

A photo-based AI video is easier to plan when you decide what the final clip must communicate before choosing a picture. A birthday greeting needs the recipient to recognize the person and understand the message. A pet introduction needs a readable animal face and a scene that suits its character. A two-person performance needs two distinct subjects and a clear assignment of roles. Write a single sentence describing the result you want. If that sentence depends on exact spoken words, an exact gesture, or an unchanged face throughout, check whether the template actually promises that behavior before buying.

This guide describes practical preparation and inspection steps. It does not report a measured success rate, a controlled photo experiment, or a guarantee that any particular portrait will work. The examples on template pages show individual outputs; use them to understand the style and its limitations.

Choose a photo with visible facial structure

Start with an image where you can see the subject's eyes, nose, mouth, and outline without zooming into a tiny part of a group picture. Prefer an evenly lit face with natural detail over a heavily retouched selfie. A near-front view is usually easier to interpret than a face almost completely turned away. A hat, sunglasses, hair across the eyes, or a hand over the mouth removes information the model would otherwise use.

There is no universal pixel threshold that guarantees likeness across every model and template. Use the upload control's actual file limits rather than inventing a minimum resolution. If you have two candidate photos, compare the visible detail at the crop you will upload. Upscaling a blurred face creates a larger file, but does not recover the missing features. Choose a clearer original when possible. Avoid screenshots that add interface buttons or text over the face.

Prepare people and pets differently

For a human portrait, check whether the template requests one subject or separate photos for two roles. Do not upload an entire family photograph to a slot intended for one person and expect the system to know whom you mean. Crop to the intended subject while keeping enough head and shoulder context. Use photos only of consenting adults; do not upload children, public figures, or another person's portrait without their permission.

For a pet, choose a picture showing the face and distinctive markings. A dark animal photographed against a dark background may be harder to distinguish than the same animal in softer daylight. A costume can hide the ears or markings that make the animal recognizable. Let the template create its own costume unless its instructions say otherwise. Use an animal photo you are entitled to process and avoid images showing distress or cruelty.

Assign two-person roles deliberately

If a template uses two photo slots, read the role labels and compare them with the sample. Put the intended first subject in the first slot and the second subject in the second. Similar clothing or similar faces can make it difficult to judge whether roles have been preserved. Record your intended assignment before starting, then inspect both subjects in the result.

Do not treat a successful single-person portrait as proof that the same image will work equally well in a duet. A two-person scene adds interactions, occlusion, and camera composition. If the sample repeatedly turns one performer away from the camera, that role may offer fewer moments of recognizable face detail. Choose the scene based on how the complete performance uses each subject.

Watch the whole sample with sound

A poster frame cannot show how faces change during movement. Play the full sample, including moments when a head turns, a hand crosses the face, or a performer moves toward the edge. Listen to whether the soundtrack contains vocals, whether the performance alternates between people, and whether the sound continues across cuts. A template with a fixed reference soundtrack is different from one that generates a new musical performance.

Photographs guide appearance; they do not imply voice cloning. Do not promise a recipient that the result will sing in their own voice unless the product explicitly offers and authorizes that capability. Likewise, the presence of music in a sample does not establish permission to use it in a paid advertisement. Confirm the necessary rights before planning public or commercial distribution.

Budget for one decision at a time

On this site, the 480p option is $4.90 for an opening 8-second clip. The 720p option is $19.90 for the full performance: 10 seconds for AI Rap Duo and 15 seconds for the other templates. A $15 upgrade unlocks the saved full-length HD version of a purchased 480p video. That upgrade does not generate a different face, expression, or performance. Check the current template and checkout for the exact offer before payment.

If you already like the generated performance and need the saved HD version, an upgrade serves that purpose. If you dislike the expression or pose, higher resolution will not change that creative result. Do not buy an upgrade expecting a new attempt. For planning, two separate $4.90 requests cost $9.80 before any applicable checkout adjustments; neither purchase guarantees that the second result will be better.

Build a longer greeting around a short clip

For a roughly 30-second message, plan a short opening title, the generated clip, one or two still-photo moments, and a closing message. An 8-second video can serve as the reveal rather than filling the whole running time. Put the recipient's name and your own message in editing software after generation instead of expecting the model to render perfect lettering in the scene.

Match the output aspect ratio to the place you will share it. Preview cropping before placing text near the edges. If you add a new soundtrack, check both the rights and the balance between that music and the original audio. Keep an unedited downloaded copy so you can redo the edit without another purchase. Label the result as AI-generated where a viewer might otherwise mistake it for a real recording.

Inspect before retrying or upgrading

Watch once for the overall scene and once for identity. Pause at the beginning, middle, and end to inspect facial features, subject assignment, extra limbs, and unexpected changes in clothing. Listen on a device with working sound. A muted player or phone volume setting can look like a missing soundtrack even when the file includes audio.

If the file is missing or broken, save the error message and order reference and contact support. If generation failed, the paid request can be retried without another charge while its inputs remain available. Card refunds are a separate process explained in the Refund Policy. If generation succeeded but you prefer a different pose, that preference is different from a technical failure; describe the issue accurately so support can assess it.

Save files and keep a useful record

Download the finished file before the retention period ends: input and intermediate images are kept for up to 7 days and completed videos for 30 days. Keep the order reference, template, selected quality, and your own notes about what you expected. Do not store passwords or card details in that record. If you need help without signing in, use the email on our Contact page.

A useful final check is simple: the right people or pet are represented, the full clip plays with the intended sound, the duration and quality match the purchase, you have permission for the inputs and intended use, and you have saved the file. If any of those points is unresolved, address that specific issue before spending on another version.