Sole Alchemy
Find an accomplice
Prompt 2
Five stages inside one continuous shot, each closing with an explicit End state. Paste it as written.
Other prompts in this tutorial
Prompt 1
NANO BANANA PRO · VERTICAL · TWO REFERENCES: IMAGE A + IMAGE B Use Image A as the strict identity lock for the person: preserve their exact facial features, hairstyle, hair color and texture, facial hair if present, skin tone with no idealization, and any eyewear they are wearing in the reference — if Image A shows the person wearing glasses or sunglasses, keep them on the face unchanged. Composition — a vertical editorial product shot, described precisely: the person stands centered, facing the camera against the backdrop. Their RIGHT hand enters the frame from the upper area, raised to roughly face height, fingers spread downward in an elegant relaxed claw — as if they have just released or are delicately suspending the product in mid-air with invisible control. Their LEFT hand waits below at roughly chest-stomach height, palm open and turned upward in a soft receiving gesture. The product from Image B floats in mid-air in the vertical gap between the two hands, positioned in the exact center-foreground of the frame, closer to the camera than the person, tilted at a slight dynamic three-quarter angle that shows its most recognizable side. CRITICAL FOCUS LOGIC: the product is the sharpest element in the image, rendered in crisp macro-level detail — material, texture, edges, and branding exactly as in Image B with no redesign; the person's face and body sit slightly behind in soft focus, recognizable but gently blurred, eyes looking straight into the camera through the gap between the product and the hands. Scale and physics: render the product at a believable real-world size relative to the hands — a sneaker spans roughly two hand-lengths, sunglasses roughly one palm-width, a watch or ring smaller than the palm — floating with a subtle sense of suspended weight, no strings, no stand, no support. Both hands show natural skin texture, knuckle creases, and realistic finger spacing. Wardrobe: an elevated minimal outfit in soft neutral tones — an oversized cream-ivory wool overshirt or coat with visible fabric texture, worn over a simple tonal base layer — unless Image A clearly establishes a different signature style, in which case echo that style in neutral tones so the product remains the only strong color statement in the frame. Background: a seamless luxury white studio backdrop with a very subtle vertical gradient — soft warm white falling into a faint cool grey-white, premium and minimal, with gentle studio depth behind the person. Editorial photographic treatment: shot like a premium eyewear campaign, 85mm f/2 look — shallow depth of field with the focal plane locked on the floating product, soft even studio light with one broad diffused key from the front-left, delicate highlight roll-off on the product's surfaces, natural skin texture in the blurred zones, no digital smoothing. Constraints: identity locked to Image A with no drift; the product locked to Image B with no redesign, no color shift, no invented branding, no duplicated or morphed parts; exactly two hands, anatomically correct with no extra fingers; the product floats free with no visible support; background stays clean luxury white gradient with no props, no text, no watermark.
Tutorial
DENIZ AKKABAK · AI VIDEO TUTORIAL · @DENIZ.AKKABAK
MADE IN INVIDEO · invideo.io/i/DENIZ-ig
Sole Alchemy
A sneaker transformation spell in two generations — Nano Banana Pro conjures the editorial opening frame from your own photo, then Seedance 2.5 transmutes the shoe through three colorways in one continuous 20-second shot.
A TUTORIAL BY DENIZ AKKABAK · @DENIZ.AKKABAK
PIPELINE NANO BANANA PRO → SEEDANCE 2.5 FORMAT 9:16 · 20S · ONE SHOT PLATFORM INVIDEO
The trick reads like magic but is built like watchmaking: a shoe floats between two hands, dismantles into its real materials — suede panels, laces, threads — and reweaves itself into the next colorway, twice, before duplicating into a proper left-and-right pair that lands on the wearer's feet. No glow, no sparks, no smoke; the ban on fantasy effects is precisely what makes it feel supernatural. Both generations run inside invideo: Nano Banana Pro for the opening still, Seedance 2.5 for the one-shot film.
THE PIPELINE AT A GLANCE
- 01
Conjure the frame
Nano Banana Pro fuses your portrait (Image A) with the product sheet (Image B) into one editorial still: the shoe razor-sharp in mid-air, you soft-focused behind it. This becomes @Image 1 of the video.
- 02
Cast the spell
Seedance 2.5 — selected from invideo's model picker — takes that still plus three sneaker sheets and runs a five-stage transformation in a single continuous 20-second shot.
STEP 01
Nano Banana Pro — Conjuring the Opening Frame
Two inputs go in: Image A is your own photo, and it is treated as a strict identity lock — face, hair, skin, even the sunglasses stay untouched. Image B is the product, here the green sneaker's multi-panel sheet. The prompt then does something unusual: it directs the photograph like a set of blocking notes — which hand enters from where, where the product floats, and a CRITICAL FOCUS LOGIC block that puts the shoe on the focal plane and the face in soft blur. That inversion — product sharp, human soft — is what makes the frame read as a premium campaign instead of a portrait holding a shoe.
INPUTS — IMAGE A AND IMAGE B
Prompt 1
NANO BANANA PRO · VERTICAL · TWO REFERENCES: IMAGE A + IMAGE B Use Image A as the strict identity lock for the person: preserve their exact facial features, hairstyle, hair color and texture, facial hair if present, skin tone with no idealization, and any eyewear they are wearing in the reference — if Image A shows the person wearing glasses or sunglasses, keep them on the face unchanged. Composition — a vertical editorial product shot, described precisely: the person stands centered, facing the camera against the backdrop. Their RIGHT hand enters the frame from the upper area, raised to roughly face height, fingers spread downward in an elegant relaxed claw — as if they have just released or are delicately suspending the product in mid-air with invisible control. Their LEFT hand waits below at roughly chest-stomach height, palm open and turned upward in a soft receiving gesture. The product from Image B floats in mid-air in the vertical gap between the two hands, positioned in the exact center-foreground of the frame, closer to the camera than the person, tilted at a slight dynamic three-quarter angle that shows its most recognizable side. CRITICAL FOCUS LOGIC: the product is the sharpest element in the image, rendered in crisp macro-level detail — material, texture, edges, and branding exactly as in Image B with no redesign; the person's face and body sit slightly behind in soft focus, recognizable but gently blurred, eyes looking straight into the camera through the gap between the product and the hands. Scale and physics: render the product at a believable real-world size relative to the hands — a sneaker spans roughly two hand-lengths, sunglasses roughly one palm-width, a watch or ring smaller than the palm — floating with a subtle sense of suspended weight, no strings, no stand, no support. Both hands show natural skin texture, knuckle creases, and realistic finger spacing. Wardrobe: an elevated minimal outfit in soft neutral tones — an oversized cream-ivory wool overshirt or coat with visible fabric texture, worn over a simple tonal base layer — unless Image A clearly establishes a different signature style, in which case echo that style in neutral tones so the product remains the only strong color statement in the frame. Background: a seamless luxury white studio backdrop with a very subtle vertical gradient — soft warm white falling into a faint cool grey-white, premium and minimal, with gentle studio depth behind the person. Editorial photographic treatment: shot like a premium eyewear campaign, 85mm f/2 look — shallow depth of field with the focal plane locked on the floating product, soft even studio light with one broad diffused key from the front-left, delicate highlight roll-off on the product's surfaces, natural skin texture in the blurred zones, no digital smoothing. Constraints: identity locked to Image A with no drift; the product locked to Image B with no redesign, no color shift, no invented branding, no duplicated or morphed parts; exactly two hands, anatomically correct with no extra fingers; the product floats free with no visible support; background stays clean luxury white gradient with no props, no text, no watermark.
OUTPUT — THE EDITORIAL STILL
STEP 02
Setting Up Seedance 2.5 Inside invideo
Open invideo's generative models catalog and pick Seedance 2.5 — the premium video model with reference images, video and audio support — from the video lineup. It sits at the top of the catalog next to Seedance 2.0, Kling, Veo and the rest of the frontier models; this guide's spell runs specifically on 2.5 because a five-stage transformation needs its 20-second single-generation window.
Then build the request exactly as in the screenshot below: attach the four images in order — the Step 01 editorial still first, then the green, blue and brown sheets — set the duration to 20s, the aspect to Portrait (9:16), outputs to x1, and confirm the mode reads Seedance 2.5 · Reference to Video before pasting the prompt.
REFERENCE INPUTS — ATTACH IN THIS ORDER
STEP 03
Prompt 2
Five stages inside one continuous shot, each closing with an explicit End state. Paste it as written.
SEEDANCE 2.5 · 9:16 · 20S · SINGLE CONTINUOUS SHOT · REFERENCE TO VIDEO
FORMAT: 20s / 9:16 / cinematic sneaker transformation / single continuous shot
STYLE: Photorealistic smartphone capture with cinematic polish: sharp phone sensor detail, natural micro-shake, realistic autofocus, soft studio daylight, seamless backdrop, subtle grain, real-world physics, no fantasy glow.
REFERENCES:
@Image 1 defines the person: preserve their exact face, hair, wardrobe and accessories, no changes. Also defines the opening pose: green sneaker floating between their hands.
@Image 2: multi-panel sheet of the SAME single green suede sneaker, beige sole. Design detail only; ignore the panel grid, never render split-screen.
@Image 3: multi-panel sheet of the SAME single light-blue suede sneaker, pale-yellow and green sole. Design detail only; ignore panel layout.
@Image 4: multi-panel sheet of the SAME single brown-and-tan sneaker, dark grey sole. Design detail only; ignore panel layout.
VIDEO PROMPT:
[Stage 1 - approx 0-4s] The person centered, right hand raised above, left palm open below, the green sneaker suspended in the gap between their hands, rotating on its vertical axis. Camera: one continuous push-in from wide to a chest-height medium shot. Their fingers glide around the floating shoe without touching it; hands stay normal, no light or particles. End state: medium framing, green shoe rotating.
[Stage 2 - approx 4-9s] Mid-rotation the green sneaker separates into its real materials: suede panels lift apart, laces unravel, fine threads drift in a tight controlled cloud between their hands, then weave back together into the blue sneaker from @Image 3. Clean minimal particulation, no sparks, no glow, no smoke. End state: one blue sneaker rotating in place.
[Stage 3 - approx 9-13s] The blue sneaker disassembles the same way, reassembling into the brown sneaker from @Image 4. End state: one brown sneaker rotating between their hands.
[Stage 4 - approx 13-16s] The brown sneaker duplicates with a brief subtle spatial blur: a mirrored second sneaker slides out of the first; the pair hovers side by side as correct left and right shoes. End state: exactly two brown sneakers floating.
[Stage 5 - approx 16-20s] The person presses both palms gently downward. The pair glides down, aligns with their feet and wraps onto them, laces settling. They turn and walk out of frame with relaxed confident strides; camera holds, slight tilt-down to the shoes. Hard stop.
AUDIO: <soft room tone, fabric and thread whispers at each disassembly, low whoosh as the pair descends, two subtle sole impacts, fading footsteps> No dialogue, no music.
CONSISTENCY: Keep the person's face, wardrobe, accessories, lighting unchanged. Exactly one person. One shoe in Stages 1-3, exactly two in Stages 4-5.
NEGATIVE PROMPT: no glowing hands, no magic energy, no sparks, no smoke, no split-screen, no panel borders, no duplicate person, no face distortion, no wardrobe change, no extra shoes, no cartoon, no cheap CGI, no leftover debris, no text, no watermark.
STEP 04
Anatomy — Why the Spell Holds
- 1
One still, two jobs
@Image 1 carries both the identity and the opening pose — because Step 01 staged the exact first frame, the video begins already composed, and Seedance spends zero seconds finding its shot.
- 2
The sheet waiver, three times
Every sneaker reference is a multi-panel sheet, so each one repeats the same clause: design detail only, ignore the panel grid, never render split-screen. One waiver per sheet keeps four references from leaking layouts.
- 3
Materials, not magic
The transformation is engineered from real components — suede panels lift, laces unravel, threads reweave — while the negative prompt bans glow, sparks and smoke. Physics-based transitions are what let a supernatural event pass the eye's realism check.
- 4
End states as checkpoints
Every stage closes with an explicit End state line — green rotating, one blue, one brown, exactly two, worn and walking. Checkpoints stop a 20-second morph chain from drifting mid-sequence.
- 5
Counted objects
CONSISTENCY pins the arithmetic: one person, one shoe in Stages 1-3, exactly two in Stages 4-5 — and the pair is specified as correct left and right shoes. Object counts are where transformation videos usually break; this prompt locks them by name.
- 6
Sound of craft, not sorcery
The audio is thread whispers, a low whoosh, two sole impacts, fading footsteps — no music, no dialogue. Quiet, material sound keeps the spell inside the same realism contract as the visuals.
- 7
The insurance
The negative prompt closes every classic failure at once: glowing hands and magic energy (tone drift), split-screen and panel borders (sheet leakage), duplicate person and extra shoes (count drift), leftover debris (incomplete reassembly).
THE RESULT
Output — Twenty Seconds, Three Shoes, One Pair
Audit it stage by stage: the push-in lands on a chest-height medium exactly as Stage 1 orders, each disassembly stays a tight thread cloud with nothing glowing, the pair appears as true mirrored left-and-right shoes, and the film ends on the tilt-down and a hard stop. Five end states, five delivered.
SEEDANCE 2.5 LIVES INSIDE INVIDEO
One tab. Every frontier model. Your spell.
Everything in this guide — the editorial still and the 20-second transformation — was generated inside invideo, where Seedance 2.5 sits one click away in the model picker alongside the rest of the frontier lineup. Bring your own face and your own product, and run the same alchemy.
WRITTEN, DIRECTED & PROMPTED BY DENIZ AKKABAK · @DENIZ.AKKABAK