AI video templates

Start with the material you have. For one clear portrait, compare Bulin 47 and Burning Bridges. For two people appearing together, look at Hotel Lobby, Zombie Love Story or Raindance. The two-role options take individual character images. Use fictional character images or, for real people, adult photos with consent for the AI processing and intended scene. Choose the number of roles first, rather than uploading extra photos and hoping the template will decide for you.

Next, consider where you plan to watch or share the result. Bulin 47 uses a portrait 9:16 scene. Zombie Love Story, Burning Bridges and Raindance have fixed landscape scenes. Hotel Lobby offers landscape, portrait and square; its Normal mode fits the complete source performance with margins when needed. A new aspect ratio is not a promise of a newly directed camera shot, and every template does not offer every crop or frame shape.

Finally, choose the sound you actually want. An original-sound template follows an existing performance. Other Hotel Lobby music styles generate a new song and performance instead. New singing is not voice cloning from a portrait. Treat the examples as references for the type of scene you can make, rather than a guarantee of identical expressions, lip synchronization or motion in every output.

Prepare a clear JPG, PNG or WebP character image for each required role, up to 8 MB per image. A readable face, usable lighting and a clear subject make the reference easier to review. Avoid heavy blur, strong beauty filters, large obstructions and a crowded group picture when the field asks for one person. Hair, glasses, hats and hands across a face can affect the result. That is a reason to inspect the output carefully, not a claim that every accessory makes a photo unusable.

For a two-person scene, read the role labels and check both previews before continuing. Do not rely on file names to decide which portrait belongs on the left or right. Use the swap control when the chosen template supports it. Photos can come from different occasions, but sharply different lighting, angles or framing may make the generated pairing less consistent. Clear inputs help; they do not guarantee a perfect likeness through every turn or occlusion.

Use fictional character images or adult photos that you have the right to process for the intended result. Each real person must consent to the AI processing and intended scene. Selecting an image is not consent to buy or generate. Keep identification documents, unrelated private details and other sensitive material out of the references. Upload timing and retention differ between the account asset library and template flows, so read the privacy policy before submitting material you need handled in a particular way.

First, watch the reference clip. Check where people stand, how they move, whether they turn or overlap, and what the soundtrack contains. A poster cannot show the complete performance. Browsing the directory and playing an existing sample do not submit a model request, so use that stage to decide whether the scene fits your purpose.

Second, open the relevant creator, add the required portraits and review the role assignment. Each template page shows its own photo requirements, permission confirmation and purchase flow. Changing a selection, photo or quality setting does not automatically start a charged generation.

Third, review the current price or account-credit quote and explicitly submit. A payment return restores the process; it does not create a video by itself. While a task is processing, check that task before starting another one. A lost connection or a result still being saved is not automatically a confirmed failure. When the video is ready, review the complete performance and audio, then download the version you want to keep.

The scene determines the available specification. A fixed 15-second template is not an arbitrary-length video editor, and it is not a menu containing every model, resolution or music option. Raindance is fixed at 28 seconds, 720p and 16:9. The other cards list their default length and quality choices. Compare those facts before buying, especially if a project requires a particular frame shape or delivery size.

Hotel Lobby Normal preserves the source performance and soundtrack, while its other styles use the reference movement to generate new audio. Both current paths use the character photos directly and do not require a separate first-frame purchase or approval. Higher resolution does not guarantee perfect identity consistency. Keeping a soundtrack also does not automatically grant every right in the music or the original performance.

If you want to invent motion around your own still photograph, compare the homepage image-to-video workbench. If you already have a source clip, read the Video Recast page. Those workflows have their own input and duration limits; they are not interchangeable with the fixed templates here. Select the task you want to complete before comparing model names or spending credits.