Genjutsu AI Video Generator for reference-led video ideas

Explore the Genjutsu AI Video Generator on Try Nano Banana through reference image experiments. Start with a source clip and a clear visual reference, browse this site’s published examples, and build a focused transformation brief. A fixed video model powers the workflow, with source-video guidance and the supported inputs shown in the generator.

Two Ways to Transform Video

Choose a transformation for Reference-led video ideas on Try Nano Banana

Build a new visual world around an existing performance, or focus your transformation on a specific subject. Start with the option that matches your idea. Prepare a clean reference image before writing the video brief. Decide which visual details should guide the scene, then inspect how clearly those details survive movement.

Motion Transfer

Keep the performance. Reimagine the scene.

Use your source clip to guide movement, timing and camera direction. Add references for a new character or setting, then describe how you want the whole scene to look. Choose this for a dance, an action sequence or a performance in a different world. For this page’s creative direction: Explore how a carefully prepared image reference guides a moving scene, reviewing reference details, image composition and moving-frame clarity before choosing the next creative direction.

Object Swap

Pick a target. Give it a new look.

Focus on a person, outfit, product or object in your source clip. Provide a reference for its replacement and name the target in your prompt. Choose this when you want to change a specific subject while keeping the surrounding scene as your starting point. For Reference-led video ideas on Try Nano Banana, name the exact subject before describing a replacement, then review the whole clip against your brief.

Genjutsu video examples

Compare the source materials and result. Use loads the complete recipe into the generator; Generate submits your new request.

Desert Duo

Object Swap
Source video
Reference result
Desert Duo, reference image 1
Reference 1
Desert Duo, reference image 2
Reference 2
View prompt

Replace the main characters with the characters from my references.

Ceiling Fan Transition

Motion Transfer
Source video
Reference result
Ceiling Fan Transition, reference image 1
Reference 1
Ceiling Fan Transition, reference image 2
Reference 2
Ceiling Fan Transition, reference image 3
Reference 3
Ceiling Fan Transition, reference image 4
Reference 4
View prompt

Create a photorealistic 12-second vertical 9:16 fashion video featuring ONE FEMALE CHARACTER wearing four outfits, using @image1 to @image4. SCENE CONTEXT Overhead view of one woman on a muted blue carpet beside a curved blush-pink sofa, directly beneath a large chrome ceiling fan. Each passing fan blade reveals her in a new outfit and a new pose. ACTIVE REFERENCES @image1 to @image4 all show the same woman. Preserve her face, skin tone, physique and hairstyle identically throughout. @image1 supplies Outfit 1, @image2 supplies Outfit 2, @image3 supplies Outfit 3, @image4 supplies Outfit 4. Each reference image is the exact visual source of truth for its corresponding outfit. Preserve each garment exactly as shown in its reference image, including its original design, silhouette, construction, cut, proportions, print, print placement, colours, fabric appearance, texture, trims, seams, details, footwear and visible accessories. Do not redesign, reinterpret, simplify, embellish, alter, add to, remove from, substitute or invent any part of the garments. Do not change the garment’s length, shape, neckline, sleeves, waistline, proportions, print, print placement, fabric appearance or construction. Do not generate a similar or inspired version of the clothing. The actual garments shown in @image1 to @image4 must remain visually faithful to their respective references. Use these images as identity and wardrobe references only; recreate the setting and actions described below. CAMERA AND COMPOSITION One continuous shot from a completely stationary camera pointing vertically downward at exactly 90 degrees. The camera looks past a ceiling fan toward the woman lying on the floor below. Locked-off framing, identical from the first frame to the last. A muted blue carpet fills the floor. A curved blush-pink sofa is partially cropped along the left edge. The woman settles on the carpet beside it, head toward the top of the frame and feet toward the bottom. Keep her complete body comfortably inside the frame after the opening roll. CHROME FAN A large polished CHROME ceiling fan occupies the extreme foreground, suspended between the camera and the woman. Its reflective circular motor housing is partially cropped at the top centre. Broad metallic blades pass very close to the lens and appear large enough to sweep across her body. Preserve realistic silver reflections, blade brackets and rounded blade ends. The fan begins stationary, then rotates clockwise after she presses a remote. Its blades remain visually readable, with restrained natural motion blur. The rotating blade itself creates each outfit transition. TIMING AND ACTION 0.0 to 1.5s: Wearing Outfit 1, the woman lies fully on top of the pink sofa, her entire body supported by its cushions, then rolls sideways off the edge toward screen-right and naturally drops onto the carpet, landing on her back with her head toward the top of the frame and feet toward the bottom.

Handbag Slide

Motion Transfer
Source video
Reference result
Handbag Slide, reference image 1
Reference 1
Handbag Slide, reference image 2
Reference 2
View prompt

Replace the video's main character

Commercial Product Orbit

Motion Transfer
Source video
Reference result
Commercial Product Orbit, reference image 1
Reference 1
Commercial Product Orbit, reference image 2
Reference 2
Commercial Product Orbit, reference image 3
Reference 3
View prompt

Motion transfer: replicate the exact camera movement, timing, and rhythm of the reference video, applied to a new subject: a kraft-paper burger bag with a folded top and clean logo, next to a paper cup with lid, straw, and condensation. Both items move together as one group with natural weight. Clean restaurant counter, warm blurred background, soft key light with warm rim light. Commercial food ad look, 4K, shallow depth of field, smooth stabilized motion, consistent shape, colors, and logo placement in every frame. Negative: distorted logo, unreadable text, warped packaging, morphing objects, flickering textures, color shifts between frames, jitter, jerky motion, motion not matching the reference, extra or duplicate objects, floating objects, deformed lid, bent straw, low resolution, blurry, overexposed, harsh shadows, cartoon, CGI look, watermark, subtitles, hands, people, cropped subjects

Selfie Walk

Motion Transfer
Source video
Reference result
Selfie Walk, reference image 1
Reference 1
View prompt

Replace the video's main character

Influencer Rhythm Recap

Motion Transfer
Source video
Reference result
Influencer Rhythm Recap, reference image 1
Reference 1
Influencer Rhythm Recap, reference image 2
Reference 2
View prompt

Transfer the single female performer's exact motion in @Video1 to our original adult AI influencer shown in @Image1 and @Image2. These images depict ONE identical woman: @Image1 is the face identity anchor and @Image2 establishes her appearance and body proportions, not the background or wardrobe. Apply her recognizable facial geometry, natural green eyes, full lips, freckles and long chestnut brown hair consistently throughout the video. Preserve the source video's indoor setting, original outfit (black graphic T-shirt and light-colored bottoms), composition, camera, timing, gestures and dance motion. Match the source illumination and natural motion blur. Keep the original video audio if supported. Output the exact 8-second source segment; no introduction, no additional people or cuts, no studio background. Photorealistic natural skin detail with temporally consistent identity.

Orange Jacket

Object Swap
Source video
Reference result
Orange Jacket, reference image 1
Reference 1
View prompt

Replace the main character with the character from my references.

Fantasy Rider

Motion Transfer
Source video
Reference result
Fantasy Rider, reference image 1
Reference 1
View prompt

TASK Edit @video1. Replace the horseback archer with @image1, using her complete face, black hair in a high bun, skin tone, body build, proportions, black high-collared tunic and flowing black skirt panels, dark boots, purple sleeve accents, brown belt, and silver scale armor exactly as shown in @image1. Keep the source rider’s mounted saddle, reins, leather quiver with spare arrows, traditional wooden recurve bow, and arrow. The bow shown in @image1 does not replace the source bow. Keep the dark brown horse unchanged, including its leather bridle, reins, saddle, and dark blue saddle pad. Everything else stays exactly as in @video1: every galloping stride, rider balance, jump, bow draw and arrow release, character position, camera move, framing change, hard cut, and timing. ACTIVE REFERENCES @image1 is the complete appearance reference for the horseback archer: use her face, high black bun, windblown loose hair, skin tone, physique, silver scale armor, black and purple clothing, brown belt, skirt panels, and dark boots. Do not transfer the bow held in the reference image; retain the source wooden bow, arrow, quiver, saddle, and reins. SOURCE VIDEO @video1 is the source video being edited. Preserve the mounted archer and dark brown horse racing toward the camera along the dirt riding track, the horse clearing the wooden jumping rail and landing into a gallop, then the rider raising the bow, nocking and drawing an arrow, aiming while mounted, and releasing the arrow as the horse continues across the grassland. Preserve both hard cuts exactly in place. SHOT-BY-SHOT GUIDE Shot 1 (0-2.3s): Keep the wide, dynamic high-to-low tracking view skimming backward over the foreground wooden posts. @image1 and the unchanged dark brown horse approach from distant center toward the right midground along the dirt track. @image1 sits forward in the saddle in an equestrian riding posture, her black skirt panels and loose hair streaming behind her in the wind. She looks straight ahead with focused determination, keeps the reins in hand, and holds the traditional wooden bow low in her left hand beside the saddle. The horse gallops at full speed with its hooves striking the dirt and throwing dust behind it. End with horse and rider passing close into the right midground. Shot 2 (2.3-6s): Keep the low-angle tracking shot following closely behind horse and rider as they move away through the center midground. @image1 folds forward over the horse’s neck in a compact jumping seat, holding the reins and keeping the wooden bow secure alongside the horse’s flank. The dark brown horse launches over the chest-high wooden rail, tucks its forelegs, clears the rail without contact, then lands on the dirt and continues at a sprint. @image1 rises back toward an upright saddle posture after the landing, still facing forward with a concentrated expression. Keep the tall wooden posts, hurdle, dirt contact, dust plume, and open hills exactly as in the source shot. Shot 3 (6-10s): Keep the medium profile tracking shot pacing the horse and rider from the right side as they travel across the center frame. The dark brown horse spans the lower center frame in a steady gallop over the grass and dirt. @image1 remains firmly seated in the saddle with her torso upright and slightly turned toward the forward-right target direction. She raises the source wooden recurve bow in her left hand, nocks the source arrow, pulls the bowstring back to her cheek with her right hand and elbow extended behind her, squints along the arrow shaft, then releases the arrow. Keep the quiver fixed at her hip or saddle area with spare fletched arrows visible. End with the released arrow moving forward while the horse maintains its gallop. SETTING Keep the source open grassland throughout: the dirt riding tracks, flat grassy ground, rolling green hills, clear blue sky, wooden obstacle posts, and wooden jumping rail. Keep horse, rider, bow, quiver, arrow, and saddle at correct scale and with secure physical contact. The horse’s hooves must strike the dirt and grass naturally, the rider must remain properly supported in the saddle, and the rail must remain chest-high relative to the horse. EXPRESSION NOTES @image1 maintains a determined, composed expression during the gallop and jump. During the final shot, her gaze becomes sharply focused along the arrow toward the forward-right target before the release. PHYSICS Keep @image1’s armor fitted and stable over her torso while the black skirt panels, loose hair, and purple sleeves respond naturally to the speed and wind. Keep both hands correctly attached to the reins, bow grip, and bowstring at each moment. Keep the bow flexible under full draw, the arrow nocked securely before release, the string recoiling after release, and the arrow continuing forward from the bow. Keep the quiver and saddle firmly attached throughout the gallop and jump. LIGHTING Match @image1’s face, scale armor, black clothing, purple fabric accents, an

Genjutsu AI Video Generator for reference-led video ideas

This workspace is organized for creators moving from reference images to video briefs. A useful starting point is a concrete scene: a carefully prepared image reference guides a moving scene. The aim is to explore a visual direction with a source video and supporting imagery, while keeping the important creative decisions easy to inspect. Begin with the subject and the intended change. A focused brief helps you decide whether the output serves your idea, rather than simply producing a surprising frame that does not hold together as a video.

For reference image experiments, judge the sequence through reference details, image composition and moving-frame clarity. These are review priorities, not promises that every feature will remain untouched. Generated video can introduce changes in identity, geometry, lighting or motion. Keep the original clip nearby and compare the beginning, middle and ending. If a variation misses your intention, identify the specific mismatch before revising the reference or prompt. Clear notes make the next attempt more useful than adding several unrelated requests.

Use examples to prepare reference image experiments

The examples beside the generator help creators moving from reference images to video briefs understand a complete creative recipe. Look at the source video, reference imagery, descriptive caption and result where those materials are available. The caption explains the idea; the prompt supplies instructions for the model. They serve different purposes. Watch the entire result before choosing a recipe, paying particular attention to reference details, image composition and moving-frame clarity. A cover image is useful for browsing, but it cannot show changes between frames or reveal whether the ending remains coherent.

An example with a complete reusable recipe includes a Use action. Clicking Use replaces the current source material, reference images, prompt, mode and supported parameters in the generator. It does not start generation or spend credits. You can inspect the loaded recipe, change the brief to fit reference image experiments, and review the current quote before clicking Generate. A display-only example may lack reusable inputs; viewing its result does not imply that its original materials are available to copy.

Build a brief for reference-led video ideas

Prepare a clean reference image before writing the video brief. Decide which visual details should guide the scene, then inspect how clearly those details survive movement. Name the subject, describe the desired visual change and explain any important relationship between the subject and the setting. For a concept in which a carefully prepared image reference guides a moving scene, keep the prompt readable and specific. Use observable details such as surface texture, clothing, position or light direction. Avoid asking for several contradictory styles at the same time, because that makes it harder to understand which instruction influenced the final sequence.

Reference images should support the same intention as the text. Check that the subject is readable, the image is not obscured by unrelated elements, and you have permission to use the material. When multiple images are supported, their roles and order matter. A portrait reference and a product reference are not interchangeable. Keep the reference set focused on reference image experiments, and remove images that introduce a different subject, conflicting wardrobe or a lighting setup that fights the scene you described.

Prepare and review your reference image experiments

Choose an authorized source clip with clear movement and a readable subject. Inspect the available input requirements in the generator before uploading. File size, duration, format and the number of references depend on the actual mode and service configuration. The fixed video model keeps this workspace focused; it does not offer a model picker. If generation is unavailable, do not assume that uploading a different image will enable it. Keep your creative brief so you can continue when the required service is available.

After loading an example or your own materials, check every input before submission. Confirm the selected operation matches the change you want, remove unintended references and read the complete prompt. Review the displayed settings and quote when available. Generate is the deliberate submission action, separate from Use. For creators moving from reference images to video briefs, this separation provides a useful review point: you can prepare reference image experiments while browsing examples without accidentally submitting a task each time you explore a different recipe.

Review reference details, image composition and moving-frame clarity

Assess the full output at the size and pace of its intended presentation. For a direction where a carefully prepared image reference guides a moving scene, pause at moments when the subject turns, crosses the frame or overlaps another object. Inspect reference details, image composition and moving-frame clarity in these more demanding frames. Check the opening and closing as well as the most attractive moment. Small inconsistencies can become distracting once the sequence is placed beside another shot, cropped for a different frame or replayed several times.

Keep creative evaluation separate from factual accuracy. A convincing visual treatment can still alter a face, a product label or the shape of an object. Review details manually and obtain any necessary permission before publishing. Save the brief and the chosen references with your project notes, so the intended direction remains understandable later. Compare variations by changing one important input at a time. This makes reference image experiments easier to refine and gives each additional attempt a clear purpose.

Try Nano Banana Genjutsu AI Video Generator FAQ

Can I choose a different model here?

No. This Genjutsu workspace uses a built-in video model for reference image experiments. It has no model selector. The controls reflect the supported operation and current service availability. Use the wider AI video tools if you need to explore a different model or workflow.

What happens when I click Use?

A complete example replaces the generator’s source video, references, prompt, mode and parameters immediately. No confirmation dialog appears. It does not generate a video or deduct credits. Review the loaded recipe for your reference-led video ideas brief, then submit separately with Generate.

What should I inspect in the generated video?

Watch the full clip and compare reference details, image composition and moving-frame clarity. Pay attention to transitions and occluded details. A strong thumbnail does not establish continuity, accuracy or suitability for publication. Compare the result against your original brief and source material.

Does every example include reusable source material?

No. Some examples are useful for viewing an outcome but do not contain all the original inputs. Only complete, authorized recipes can provide the full Use workflow. Choose your own source material when developing reference image experiments from a display-only example.

Does this page guarantee identical motion or identity?

No. The result depends on the inputs, operation and actual model behavior. Inspect the whole sequence, especially reference details, image composition and moving-frame clarity. References provide direction rather than a guarantee of exact preservation. Revise the specific element that misses your intention.

How do I check generation costs?

Review the current quote and available controls before clicking Generate. Use does not spend credits. Account history and Pricing provide related information, but a reusable example’s saved parameters do not establish the current cost of your new request on Try Nano Banana.

Can I request a refund for my purchase?

You can request a refund within 10 days of purchase if you have unused credits from that purchase and provide a clear reason, subject to our Refund Policy. If eligible, an unused purchase can be fully refunded; a partly used purchase is refunded in proportion to its remaining unused credits. Fully used purchases and free or promotional credits are not refundable.

Do credits expire or carry over?

Credit validity depends on the pack, subscription or promotion that issued them. Review the purchase terms and expiration dates in Settings > Credits. A renewal or new purchase does not extend older credits. Use credits before their stated expiration; do not assume unused subscription credits roll over indefinitely.

What happens to credits if generation fails?

Credits charged for a task that is confirmed as failed due to a system error are returned automatically. A task that is still processing has not yet been confirmed as failed. Check its status and your credit history before submitting again. If a confirmed failure still shows an unresolved charge, email support@try-nanobanana.com with the task ID.

Can I get credits back if I do not like the result?

AI results can vary, so a completed result that differs from your expectations is not automatically treated as a technical failure. Eligible completed image tasks can be submitted for credit refund review using the refund option where available; approval is required. For a video or another issue, contact support@try-nanobanana.com with the task ID and details. A credit refund returns credits, not money to your payment method.

How can I get help with billing or a generation?

Email support@try-nanobanana.com. For billing, include your account email and order ID; for generation, include the task ID, issue description and relevant screenshots. If you purchased as a guest, keep the original browser data and payment receipt so we can help locate the purchase.

Reference-led video ideas on Try Nano Banana

Prepare your reference image experiments

· Updated

What's new on this page

  • — Updated page copy and content.
  • — Page content published.