# How to Choose Photos for AI Video Templates Prepare clear photos for human, pet and object roles, check each upload slot and avoid common reference-photo mistakes. By GenjutsuLite Team | Published 2026-10-25 | 5 min read HTML source: https://genjutsulite.com/hub/photos-for-ai-video-templates Canonical HTML: https://genjutsulite.com/hub/photos-for-ai-video-templates Category: Blog The useful reference photo is the one that makes the intended subject easy to identify. A high file size cannot compensate for a hidden face, a tiny pet or several people competing for the same role. Before uploading anything, open the template's detail window in [All Templates](/templates). Read the role names, required photo count and upload guidance. Those instructions are specific to the selected video; another template in the same collection may ask for different subjects. ## Start with the role, not the filename Photo slots correspond to the replaceable roles in a template. A human role needs a photo of the intended person. A pet role needs the intended animal. An object role needs the kind of object described by that slot. Do not assume that every subject visible in a preview is an upload slot. Some elements belong to the original scene. Likewise, two visible performers do not make the second upload a pet photo: read the actual role labels and check them against the full preview. If a requirement seems inconsistent with the video, pause before paying and ask [support@genjutsulite.com](mailto:support@genjutsulite.com) for help identifying the role. Choosing an unrelated photo to fill a slot is not a useful workaround. ## Choose a clear photo for a person Use a picture where the intended person is easy to see and matches the slot's instructions. A clear, evenly lit face is a useful starting point for a face-led role. If a role specifically asks for a wider body view, follow that requirement instead of uploading a close crop. - Prefer one identifiable person in the image. A crowded group photo makes the intended subject less clear. - Keep important facial features visible. Heavy shadows, hands over the face and strong blur remove useful information. - Avoid strong beauty filters and edits that substantially change appearance. - Check the image itself after export. A large screenshot can still contain a very small face. - Use photos with the person's permission, and respect the site's consent requirements. For two-person scenes, prepare each requested role separately and compare the upload previews before checkout. The position of an upload box is a role assignment, so take care not to swap the two people accidentally. Collections such as [AI Hotel Lobby](/templates/ai-hotel-lobby) and [AI Rap](/templates/ai-rap) can contain templates with different casting requirements. ## Choose a photo for a pet For a pet role, choose a photo that clearly shows the intended animal. Keep its face and recognizable markings visible, with enough of the body to meet the slot's instructions. A relaxed, well-lit photograph is generally easier to read than a blurred action shot. Avoid a frame full of similar animals when only one should fill the role. Strong cropping, clothing that covers the pet's features and a distracting foreground can hide useful details. If the template calls for multiple pets, supply each labeled role separately. Explore [AI Dancing Pet](/templates/ai-dancing-pet) to inspect pet-focused scenes. The preview shows the intended action; it does not guarantee that every animal photo will produce identical motion or proportions. ## Choose a photo for an object An object slot may ask for a car or another clearly specified subject. Use an image of the requested object, with its outline and identifying details visible. Try to avoid heavy obstruction, extreme blur or a crop that cuts off the parts the role needs. Do not put a person into an object slot simply because the video also contains a person. Read which element the template is designed to replace. For scenes in [AI Flying Car](/templates/ai-flying-car), check the individual template instructions before deciding what to upload. ## Check file format, size and role coverage GenjutsuLite accepts JPG, PNG and WebP reference photos, up to **4 MiB each**. If a photo is in another format, export a supported copy before uploading. If it exceeds the size limit, reduce the file size while keeping the subject clear. | Check | What to look for | What to change | | --- | --- | --- | | Role match | The photo shows the requested person, pet or object | Replace a photo intended for another role | | Visibility | Important identifying features are visible | Choose a clearer image or a more useful crop | | Subject count | The intended subject is unambiguous | Use a separate photo rather than a crowded group | | File | JPG, PNG or WebP, no larger than 4 MiB | Export a supported, smaller copy | | Coverage | Every required slot has the correct image | Complete each labeled slot before checkout | The number of required photos is determined by the selected template. Do not upload extra people or assume one image will automatically fill multiple roles. Review every preview in the generator; uploaded photos can be replaced or removed before you submit the form. ## Know what resets when you select Recreate Clicking **Recreate** starts a fresh local form after the template is checked. It does not bring example cast photos into the generator. Your upload slots should begin empty, with 480p selected and the template's default format and complete duration. Ordinary template selection can preserve a draft separately for each template and version. An already created order is different: resume it from **My orders** using its original saved link. If an upload or order request has an uncertain result, recover that flow before starting again. ## Good inputs help, but do not promise an identical result Clear references give a better starting point for identifying the intended cast. They do not guarantee exact likeness, flawless anatomy, identical motion or the same audio as a preview. Read the full template information before deciding whether the scene suits your purpose. New videos retain the selected template's complete duration, up to 21 seconds. Quality is USD 7.99 for 480p or USD 11.99 for 720p, with the final total shown at checkout. Videos are manually prepared after confirmed payment, without a fixed delivery-time promise. When your photos are ready, return to [the template workbench](/templates) or follow the [step-by-step creation guide](/hub/how-to-make-ai-trend-videos).