How to Make a Fancy Cafe AI Video in 720p
Choose one or two authorized adult photos, sign in, purchase a credit, and generate a downloadable 720p Fancy Cafe AI video.
Last updated: 2026-11-03
To make a Fancy Cafe AI video, choose one or two photos, confirm you have permission to use them, sign in, purchase a video credit, and start generation. The result is a 720p MP4 that you can watch and download from the generator.
Step 1: Understand the two roles
The Left person is the visitor who mistakes the small table for a seat. The Right person is the arriving companion who demonstrates how the furniture works. Assign each photo to the intended role before generating.
You can replace either person or both. With only one photo, the other person stays unchanged from the original video.
Step 2: Prepare suitable photos
Use one clear photo of a consenting adult per role. Good lighting, a visible face, and head-to-waist or head-to-knees framing provide useful identity information. Avoid sunglasses, heavy face coverings, blurry images, group photos, and aggressive beauty filters.
The selector accepts JPG, PNG, or WebP, up to 10 MB per photo. Minors, public figures, sexual content, harassment, and unauthorized photos are prohibited. See the acceptable use policy.
Step 3: Choose your cast locally
Select Choose photo on either role card. The selected image appears in that card without immediately uploading it. Use Change photo to replace it or Remove to clear the role.
Confirm that you have permission to use every selected photo and that each person is an adult.
Step 4: Sign in and purchase a credit
Click Sign in to generate. Anonymous visitors are sent to registration or login and then returned to the generator. If your account has no video credits, continue to the secure checkout.
Current USD prices are $9.99 for one video, $19.99 for three, and $29.99 for five. A checkout redirect alone does not add credits; the signed payment notification must confirm the order. After returning, the generator displays your available balance.
Step 5: Generate the 720p video
Once a paid credit is available, click Generate 720p video. The site uploads the selected photos, performs an image safety check, consumes one credit, and submits the Kling V4 video edit through fal.ai.
The prompt instructs the model to replace only the assigned characters while preserving the source background, camera, actions, subtitles, audio, timing, aspect ratio, and scene continuity. No output-duration override is applied.
Generation can take several minutes. Keep the page open while the status is checked. If processing fails, the consumed credit is returned automatically.
Step 6: Review and download
The completed video appears in place of the reference player. Review identity consistency, role assignment, movement, scene preservation, and audio before publishing. Then select Download 720p MP4 to save the file.
AI output can contain visual errors even when a task completes successfully. Review and disclose generated content appropriately, and only publish photos and videos you have the necessary rights to use.
Frequently asked questions
How do I make a Fancy Cafe AI video?
Select one or two clear adult photos, assign them to the left or right role, sign in, purchase a credit, generate the scene, review it, and download the 720p result.
What photos work best for the café scene?
Choose a sharp, evenly lit photo of one consenting adult, with the face and upper body clearly visible. JPG, PNG, and WebP files up to 10 MB each work in the local selector. Avoid group photos, face coverings, heavy filters, and blur.
Can I replace just the right person?
Yes, the local selector lets you choose a photo for the right person without choosing one for the left. This records your intended cast visually in the browser. Actual single-role replacement still needs verification with the new video model.
Will the finished video be 21 seconds with the original sound?
The source reference is approximately 21 seconds and includes audio. The model receives instructions to preserve timing and audio, but generated output still needs review; the sample is not a guarantee of quality.