KIVIDE AIGenerate AI videos & images with premium models. No credit card needed.
Upload two separate portraits or a single couple photo, choose a scene and a camera angle, and render a short, natural kiss clip. The page is built around the one thing that decides quality: how well the two faces are aligned and lit.
Make a kiss scene from photos
Photo 1 — first person
Photo 2 — second person (optional if you upload a couple photo)
Front-facing, well lit, one face per photo. Matching face size and angles align best.
Uploading a single photo of both people together usually distorts less than two separate shots.
Scene
Camera angle
Duration
Costs credits per run — the exact cost is shown in the studio before you generate.
Sign in to generate — your photos and settings are kept.
Quality guide
Almost every disappointing kiss video fails for the same three reasons: the two faces are not aligned, the lighting does not match between them, or the model guesses wrong about who is who. None of these are model bugs you can prompt your way out of — they are input problems, and they are fixable before you ever hit generate.
The strongest predictor of a clean result is how similar the two faces are in the frame. Make each face roughly the same size, at a similar height, and turned toward roughly the same direction. If one photo is a selfie from above and the other is a full-body shot from below, the model has to invent a physically impossible geometry and the result warps.
Two portraits lit differently — one warm and front-lit, one cool and backlit — fight each other once the faces share a frame. Pick photos with similar light direction and colour temperature, or accept that a scene preset with strong ambient light (sunset, night neon) will do some of the harmonising for you.
If a photo contains more than one face, the model may animate the wrong pair or blend them. Crop each upload down to a single, clearly visible face. This single step prevents the most common ‘one person vanished’ failure.
Sunglasses, hands near the face, heavy shadow across the mouth, and extreme side profiles all reduce how accurately the model can move a face. Front-facing, evenly lit, eyes and mouth visible is the target. Glasses in particular tend to warp or reflect oddly during a lean-in.
When both people already appear in one photo, they share lighting, depth and camera angle by definition, so there is far less for the model to reconcile. Two separate photos are more flexible but need the alignment work described above.
Written to keep the two faces aligned, evenly lit and tasteful. Copy one, swap the setting, and generate.
Golden-hour warm backlight, medium close-up.
Two people lean in and share a gentle, romantic kiss at golden hour, faces aligned, soft warm backlight, shallow depth of field, medium close-up, cinematic slow motion, photoreal
Wet streets, cool key light, medium shot.
A couple shares a tender kiss in the rain at night, streetlights reflecting on wet pavement, faces close and aligned, soft cool key light, medium shot, cinematic realism
Neon bokeh, three-quarter angle.
Two people kiss under city lights on a rooftop at night, neon bokeh behind them, faces evenly lit, three-quarter angle, shallow depth of field, photoreal, gentle motion
Soft daylight, clean backdrop, waist-up.
A newly married couple shares a soft kiss at their wedding, warm daylight, clean background, faces well lit and unoccluded, waist-up framing, cinematic documentary style
Studio light, both faces clearly visible.
Two people hold hands and lean in for a gentle kiss, slow natural motion, both faces clearly visible, soft diffused studio lighting, medium close-up, tasteful and romantic
Warm lamp light, calm and natural.
An anniversary moment: a couple shares a warm kiss at home in the evening, lamp light from the side, comfortable framing, natural skin tones, photoreal, calm and romantic
Each row pairs the input that causes the problem with the input that prevents it. Fix the input first; prompting alone rarely rescues a badly matched pair.
Problem input
A low-resolution photo where the face is small and slightly turned, next to a sharp close-up.
Better input
Two sharp, front-facing photos with faces of similar size and resolution.
Fix: Use the highest-resolution photo you have for each face, crop in so the face fills a good part of the frame, and avoid mixing a blurry shot with a sharp one.
Problem input
Two photos shot from very different heights and angles, forcing an impossible lean-in.
Better input
Two photos at a similar height and angle, so the faces can approach each other believably.
Fix: Pick a scene preset with a strong, stable camera angle (medium shot works better than an extreme close-up), and keep the movement in your prompt gentle rather than dramatic.
Problem input
A group photo with several faces, or a couple photo where one face is partly hidden.
Better input
One clearly separated face per upload, or a couple photo where both faces are clearly visible.
Fix: Crop out other people before uploading. If you only have a group shot, crop a single-person segment for each upload rather than handing the model the crowd.
Problem input
One warm front-lit portrait and one cool backlit portrait composited into one frame.
Better input
Two portraits with similar light direction and colour temperature.
Fix: Choose a scene preset with dominant ambient light, or pick photos taken under similar conditions. Avoid combining an indoor flash shot with an outdoor backlit one.
Upload a single couple photo, or two separate portraits — one per person. Crop each to a single, front-facing, well-lit face so the model has no ambiguity about who is who.
Choose sunset, rain, city night or wedding, then set the camera to a close-up, medium shot or three-quarter angle. A medium shot is the most forgiving; extreme close-ups magnify any misalignment.
Check the credit cost, then render a short clip. If it is not right, change one input at a time — usually the photo with the weaker lighting — and re-run rather than stacking new prompt instructions.
Yes. You can upload two separate portraits, one per person, or a single photo that already shows both people together. A single couple photo usually distorts less because both faces share the same lighting, depth and camera angle, so use one when you have it. When you do upload two separate photos, crop each to one clear, front-facing face and try to match their size and lighting.
Change one input at a time and re-run. The most common causes are mismatched lighting, faces of very different sizes, tilted or side-profile photos, low resolution, and group photos with several faces. Crop to a single front-facing face per photo, pick two images with similar light, and choose a medium shot rather than an extreme close-up. If a run fails outright, nothing is charged for a failed generation in most cases, and you can simply try again with the adjusted photos.
Yes. Kivide can render a romantic scene from illustrations, anime portraits and 3D character art as well as photos, and you can say so in your prompt — for example, “anime style” or “3D render”. Keep each uploaded image to one clear character with a visible face so the model can align them, exactly as you would for a photo.
Cost depends on the model, duration and resolution you choose, and the exact price is shown next to the generate button before you confirm — it is never hidden. New accounts get free credits on signup, which is enough to render your first clips. A failed run is not charged in most cases.
You can download and use the clips you generate, including commercially, but you are responsible for how they are used and presented. Only upload photos of yourself or of people who have given you permission, never of public figures or anyone who has not consented, and disclose that the video is AI-generated where that matters. Do not use it to mislead anyone about a real event or relationship.
