Wan 3.0 Reference NSFW prompt guide

Start with one body you want to keep

Choose Wan 3.0 Reference, upload one explicit image of an adult performer and check the labels shown in the editor. Start with a single image and one short act. Add a second source only when it contributes a separate performer, room, motion or sound.

The current Reference workflow requires at least one image or video; audio alone is not enough. Video and audio references each have a combined fifteen-second limit, and video inputs can change the quoted credit cost. Review the model's available settings before submitting. Opening this guide or filling a prompt does not start a generation. On XXXCut, Reference uploads continue in MeltVibe, where you use a separate account and balance. Open Creator takes you to the Reference editor; choose your files and review the quote before generating.

Keep the first test narrow: one act, one camera angle and one source. Add explicit body detail only after the identity holds.

Use @Image1 for the woman's face, body, skin tone and hairstyle. She sits on the edge of a bed in a quiet room, pushes her underwear aside with two fingers and spreads her thighs. The camera holds a low medium close-up on her exposed vulva, arousal fluid shining between her labia, as she circles her clitoris and then slides two fingers into her vagina. Keep her body and face consistent with @Image1 for the whole shot. Soft window light; one continuous shot; no cuts. Wet friction and her breathing are the only sounds.

Tell the model what each source is for

Images, videos and audio are numbered separately: @Image1 is the first image, @Video1 the first video and @Audio1 the first audio. Use the labels actually shown after uploading; they are not one combined sequence.

Write the role of each source before the act. One image can define a performer's face and body; a second can define the room. With two performers, bind each person separately and state positions before contact: who is on top, whose legs are spread, where the hands go. Do not ask one image to define two identities or a body that contradicts the text.

The current catalog allows up to ten reference images, five reference videos and five reference audio clips. Video references have a combined fifteen-second limit; audio references also have a combined fifteen-second limit. Follow the uploader's format and duration validation. These are limits, not an invitation to fill every slot.

Use @Image1 for the adult woman's appearance and @Image2 only for the bedroom: bed, headboard and lamp. She kneels between a seated man's legs and takes his erect penis into her mouth, lips sealed around the shaft, one hand stroking the base, saliva coating her lips and his shaft. The camera holds a medium close-up from the side; his hand rests on the back of her head without forcing. Keep her identity, clothing state and the room from the references. Warm lamp light, one continuous shot. Wet oral sounds and her soft breathing.

Consensual roleplay between two adult step-siblings in their twenties. Person A uses @Image1 for appearance and lies on her back on the bed, thighs spread, knees pulled back. Person B uses @Image2 for appearance and kneels between her thighs. He guides his erect penis to her labia and pushes inside her in one slow stroke, then thrusts with deep, steady strokes while his hands grip her hips. Her labia stretch around his shaft as it slides in and out; his hips meet hers with a wet slap. Keep both identities, faces and body shapes distinct; do not blend their features. The camera holds a side medium shot at mattress height. Warm bedroom light, one continuous shot. Wet thrusting sounds, her moans and his breath.

Build an explicit prompt in a useful order

Write in this order: who, what act, which anatomy and contact, camera, setting and light, sound, then constraints. In Reference mode, put the source-to-role bindings first.

Use direct nouns and verbs: vulva, labia, clitoris, vagina, anus, penis, glans, shaft, testicles, nipples, breasts, semen, arousal fluid, sweat. 'Her sex' or 'his member' leaves the model guessing; 'his glans presses against her labia' does not. Name the act — vaginal, anal or oral penetration, handjob, fingering, cunnilingus, anilingus, masturbation — and state who does what to whom.

For longer shots, use numbered shot blocks with approximate time ranges:

[Shot 1] 0.00-1.99 seconds: ...

[Shot 2] 2.00-4.99 seconds: ...

Time ranges communicate order and pacing, not frame accuracy. Add one new observable change per block: a garment moves, a hand grips, hips meet, the camera drops. Keep the same bodies, clothing state and room across blocks.

Use clinical precision with sensory detail: the wet sound of insertion, the stretch of lips around the shaft, the grip of hands on hips during thrusting, sweat on skin, fluid running down a thigh. Sensory words guide texture and motion; they do not replace naming the body part.

Relationship framing is content too. For incest roleplay, name the relationship and the fact that both people are adults in the first sentence — 'two adult step-siblings in their twenties' — then direct the act. That sentence closes the two ambiguity gaps the model cannot infer: a minor reading and a real-relative reading. Write it as consensual roleplay, never as coercion.

Match the act to the shot length

Begin with one explicit beat. Undressing, oral sex, penetration and climax are four beats; put at most two in a ten-second shot and one in a five-second shot. A short clip that shows one act clearly is more useful than a rushed summary of a whole scene.

Write the order of beats, not a story: 'first she strokes him to full erection, then he lifts her onto the desk and enters her'. Use time ranges to keep the blocks in sequence. Describe the rhythm — slow deep strokes, short hard thrusts, a pause — because the model reads motion words as pacing.

Reserve the final beat for the ending you want to see: eye contact, a held pose, ejaculation, or an empty frame. If the result feels rushed, remove a beat instead of shortening every action. If it feels static, add a small physical change or a camera move, not more adjectives.

Direct the camera, contact and sound

Say what moves and what stays still. A fixed camera on a medium close-up is the clearest way to show anatomy; a slow push-in adds emphasis without changing the subject. Choose the angle that makes the act visible: POV from the performer's eye line, a low angle between spread thighs, a side close-up of insertion, an over-shoulder view of hands and hips. Do not stack an orbit, a pan and a zoom into one short shot.

Describe body mechanics, not just desire: whose hand grips whose hip, how the thighs are held open, where the weight rests, how the skin moves on impact. That is what keeps penetration and contact readable across frames.

For dialogue, name the speaker and write one short line in quotation marks. Put wet sounds, breathing and room tone in a separate sentence. Enable the sound control when you need audio; wording alone does not enable sound or guarantee the exact voice, words or synchronization. Ask for a clean image; overlays and captions are not reliably suppressed by prose.

Copy a binding pattern, then fit it to your media

Each example lists the inputs it expects and the labels it uses. Copy only after checking what @Image1 and the other labels mean in your editor. Filling an example changes prompt text; you still choose the media, mode, settings and duration.

Start from the binding pattern that matches your sources, then change one element: the act, the room, the angle or the ending. Keep each source's job stable across edits, and remove a reference rather than adding a conflicting role. These are authored suggestions, not verified generations, and no phrasing guarantees a specific body, identity or outcome.

Use @Image1 for the adult woman's appearance. Use @Video1 only for the rhythm of the hips and the handheld camera sway; do not copy the people or the room from it. She straddles a partner on a couch, his erect penis inside her, and rides him at the same pace as @Video1. His hands grip her ass; her breasts move with each bounce and his glistening shaft disappears inside her on every stroke. The camera stays close and handheld, swaying with the rhythm. Keep her face, hair and body from @Image1. One continuous shot, warm lamp light. Wet skin sounds, her moans and the couch creaking.

Use @Image1 for the adult woman's appearance. Use @Audio1 as a reference for the pace of her breathing and moans, not for words. She lies on her back, thighs spread, sliding two fingers in and out of her vagina while the camera holds a low close-up on her wet vulva. Her breath and soft moans follow the rhythm of @Audio1; her hips rise as she gets closer. Keep the room, light and her body from @Image1. Enable sound. No dialogue and no music; only her breath, wet friction and the sheets.

Use @Image1 for Person A's appearance and @Image2 only for the bedroom. Person A lies back on the bed while Person B kneels between her thighs and enters her; her legs wrap around his waist and his hands press her knees open as his shaft pushes deep into her. Keep Person A's identity from @Image1 and the room from @Image2; take the act, pace and slow arcing camera move from @Video1. One continuous shot, warm lamp light. Wet thrusting, her moans and his low breath.

When a reference result goes off track

If identities mix, simplify the cast and bind each performer to one image. If the act is unclear, name the anatomy and one physical verb instead of describing desire. If the room bleeds into the performers, re-state that @Image2 is the setting only. If penetration is hidden, move the camera to the side or lower and say where the contact happens in the first block.

If the result copies the video reference's people, repeat that @Video1 supplies motion or camera only. If a label becomes invalid after removing an upload, check the updated labels and edit the prompt. Prompt instructions can guide a model, but cannot guarantee exact identity, anatomy, continuity or timing.

If the scene reads as coercive, check that the prompt shows willing participation and an explicit adult framing; remove any age-ambiguous or non-consensual wording before retrying.

Frequently asked questions