How to Create a Waterfall Kiss Video With AI

A lush rainforest romance that balances dramatic waterfall atmosphere with safe body support, intimate close-ups, realistic wet hair, and two readable kisses.

Example image

Waterfall Kiss two-person example

This is the visual reference for the effect. The generated video follows its two-person composition, wardrobe direction, lighting, body language, transitions, and romantic atmosphere.

What you need

One clear photo of an adult man
One clear photo of an adult woman
Front-facing or three-quarter faces with natural lighting
No heavy beauty filters, sunglasses, face obstruction, or tiny faces

Upload order may vary. The prompt identifies the man and woman by the people visible in the reference images rather than by image number.

Step 1: Open the Waterfall Kiss effect

Choose Waterfall Kiss in AI Video. This effect is designed for two-person reference-to-video generation in a vertical cinematic format.

Step 2: Upload the two portraits

Upload one portrait for each person. The source photos provide identity only; the effect creates the wardrobe, environment, camera movement, transitions, and actions described below.

Step 3: Generate the video

Start generation and allow the effect to build the complete sequence. If a result has identity drift, merged hands, unstable lips, or an incomplete ending, use clearer portraits before generating again.

Key details to protect

Keep footing and support believable on wet ground
Preserve yellow, rose, and mint swimwear colors
Match the wider waterfall turn and final close cut

Full English prompt

Design the first 10.9 seconds as a complete tropical-waterfall romance; after 10.9 seconds hold the final close embrace without repeating. A tall multi-stream waterfall drops through dense emerald rainforest, fine mist sparkling in soft diffused daylight. The adult man is shirtless and wears bright butter-yellow swim shorts. The adult woman wears a dusty-rose triangle bikini top and pale mint-green fitted high-waisted swim shorts. Both skin and hair are lightly wet. 0.0-3.0: medium close vertical shot; he stands securely with one arm around her lower waist and the other hand cupping the back of her head. She arches backward gradually, eyes closed, one hand around his side, while he fully supports her balance. 3.0-5.3: he brings her upright, gently frames her jaw with both hands and brushes wet hair away; she places one hand on his ribs and they hold intense affectionate eye contact. 5.3-8.2: they lean in and share one slow stable closed-mouth kiss, bodies close, waterfall mist and loose hair moving naturally. 8.2-10.3: clean cut to a slightly wider rear three-quarter view as they turn together to admire the waterfall, still joined at the waist; she rests her head near his shoulder and he kisses her temple. 10.3-10.9: quick clean cut to a tighter face-to-face angle; she turns back, he cups her cheek and they finish with a brief clear kiss. Rich wet greens, soft gray mist, natural skin sheen, subtle handheld push-in, realistic footing on damp ground and no extra people. Preserve the adult man and adult woman identities from the two reference photographs throughout, regardless of upload order. Use the man visible in the reference images as the male lead and the woman visible in the reference images as the female lead; never refer to image one or image two. Keep facial structure, skin tone, hair identity, body proportions and apparent adult age stable in every shot. Natural anatomy, hands, lips, eye lines, balance, weight transfer, fabric movement and contact physics. Romantic, affectionate and tasteful. Vertical 9:16 photorealistic cinematic lifestyle footage. No captions, text, logos, watermarks, interface, extra people, duplicate bodies, face swaps, morphing, fused limbs, broken fingers, floating props or abrupt identity changes. Audio: a gentle romantic instrumental music bed only; no dialogue, speech, laughter, breathing, moaning, kiss sounds or any other human vocalization.

Quality checklist

Exactly one adult man and one adult woman remain recognizable
Faces, ages, skin tones, hairstyles, and body proportions stay stable
Wardrobe and scene changes follow the prompt timeline
Hands, embraces, turns, lifts, and kisses have believable contact physics
The final kiss is natural, closed-mouth, tasteful, and free of facial distortion
Romantic instrumental music only; no human voices
No duplicate people, extra limbs, text, subtitles, logos, interface, or watermark

Pro tip

Use portraits with neutral expressions and clean lighting. The effect defines the emotion and choreography, so simple source photos usually preserve identity more reliably than heavily edited selfies.