Hotel Lobby AI photo and character image requirements

Choose one authorized image for each performer, with a clear face and enough clothing detail for the result you want. Keep the images separate and decide who's on the left and who's on the right. Open each file before adding it. A file can pass a format check even when the face, outfit or intended character is hard to make out.

Video Hotel Lobby is developing routes for real-person photos and anime characters. The real-person API integration is in progress. Seedance 2.0, Seedance 2.0 Mini and Seedance 2.5 are still labeled Anime only. Personalized generation and checkout are unavailable. You can prepare your files now and watch the real six-second anime demo.

Choose photos or anime references

For a video of you and a friend, use a separate photo of each person and agree on how you'll use the clip. Will it stay private or be shared? Make sure you have permission for the material you plan to submit. Preparing these photos doesn't make them eligible for the currently labeled anime models.

For an anime performance, use one reference for each character. With original artwork, you can choose the design yourself. Get permission for artwork someone else created; recognizing a character or finding an image online doesn't give you permission to process or publish it.

Choose the guide that fits your images. The own-photos guide covers a video of you and a friend; the anime character guide covers original character performances. Each route still needs confirmed model requirements before submissions open.

Choose a crop that shows the details you need

Open the source without zooming in too far. Can you make out the face, eyes, hair and clothing? A tight headshot gives you little to check below the neckline. A full-body image can show shoes and accessories, but the face may be too small to inspect easily. Choose the crop according to the details you need.

Use the table to decide what the references need to show, then check the selected model's input requirements. We haven't established a crop that works best across models. In our anime demo, the camera moved closer and cropped the legs despite the full-body instruction. Check the source composition before adding a file, then watch for cropping in the output.

Look for hidden or unclear features

Check the features that make the person or character recognizable. Hair over an eye, a hand across the face or a dark shadow can hide a detail you need to judge. Look for another authorized image if it gives you a clearer view. Keep a mask or eye patch when it's part of the design, and note which features need to stay recognizable.

For a photograph, compare the image with how the person wants to appear. A strong filter or heavily edited face may already differ from that. For anime, compare the face, hair, accessories and clothing with the intended character design.

Save the untouched source with any prepared copy. After cropping or re-exporting, compare the two. If you've enlarged or sharpened the image, judge the visible details rather than the file size. A clear reference helps you judge the later result, although it can't promise an exact likeness.

Label the left and right performers

Video Hotel Lobby's local preparation interface labels the first slot @Image1 for the left performer and the second @Image2 for the right performer. These are the intended roles; generated positions aren't guaranteed. Filenames such as left-character-original.png and right-character-original.png can help you keep them straight. Check the previews too.

Write a short note about each character's hair, outfit and distinctive accessory, using details that match the image. If the characters look similar, choose references that show their differences clearly before trying to explain them in a long instruction.

Look at the images side by side. If one is a close face crop and the other a distant figure, work out which details each one shows. Keep one separate image for each role. You don't need a combined scene image for this workflow, and support for that input method hasn't been established.

Check the local file picker limits

The local image picker accepts JPG, PNG and WebP, up to 10 MiB and 50 megapixels per image. These checks apply to file preparation in the browser. They aren't confirmed requirements for the real-person API or proof that a provider will accept the files.

If your phone image is HEIC, export a JPG, PNG or WebP copy. Renaming the extension won't convert the contents. Open the exported file to check its orientation, crop and detail, and keep the original in case you need to convert it again.

If the picker can't read a file, try a valid export or another authorized source. If the file exceeds a local limit, make a smaller copy and check that the face is still clear. We haven't verified a minimum input size or required portrait aspect ratio. Selected character images stay on your device; seeing a local preview doesn't mean the file was uploaded or a video was made.

Compare two example image choices

As a photo example, compare a tiny crop from a group photograph with a clear, separate portrait your friend has authorized. The portrait lets you inspect their face without another person in the frame. This is an illustrative choice, not a tested real-person generation result.

For an anime example, compare a sheet with three views of a character against a file showing one clear figure in the chosen outfit. Prepare the single figure for that role and keep the sheet for your own design checks. Neither image has been tested in a model for this example.

These examples help you choose a readable reference; they don't compare generation quality. For an actual output and the problems found in its review, see the documented anime case study.

Check both images before you finish

• Each image shows the intended person or character, and you have permission to use it.
• You can make out the faces, hair and important outfit details in the actual files.
• The filenames and previews match your left and right assignments.
• Both files pass the current local format and size checks.
• Your brief states who the performers are, how they should move and what the camera should show.
• You've checked whether your chosen route is available and which inputs it supports.

Keep the two originals, prepared copies and role notes together. Follow how to make a Hotel Lobby AI video for the full preparation sequence, then use the identity, framing and audio checks to decide what to watch for in a result. You can prepare files now. Personalized generation, live task quotes and checkout are still unavailable.