Guida · Video con l'AI
Il semaforo,
un mini film con l'AI
Sono Tommaso Siveri, art director freelance in Italia: costruisco automazioni e sistemi creativi con l'AI.
Questo mini film l'ho fatto in cinque passaggi: disegno, fotogrammi con Nano Banana Pro, clip con Kling 3.0, la mia voce con Seedance 2.5, montaggio. Qui sotto c'è ogni scena, e ogni prompt lo copi con un tocco.
Tommaso Siveri · scritta il 2 novembre 2026.
Gli strumenti cambiano in fretta.
AI generatedPrima le immagini,
poi il video
Il film è fatto di cinque scene. Per ognuna ho seguito sempre gli stessi passaggi, e una clip l'ho generata solo quando tutti i fotogrammi mi tornavano.
- 01Ho scritto la storia e ho disegnato ogni inquadratura, a mano: cosa si vede, da dove guarda la camera, chi parla.
- 02Per ogni inquadratura ho generato il fotogramma con Nano Banana Pro, caricando come riferimento le foto delle facce e le scene già fatte.
- 03Quando c'è un movimento, faccio anche il fotogramma finale. La clip parte dal primo e arriva al secondo.
- 04Le clip le ho generate con Kling 3.0, da 5 secondi, a 1080p. Quella in cui parlo l'ho fatta con Seedance 2.5, caricando la mia voce.
- 05Alla fine ho montato le clip, con la musica e il cartello.
Un fotogramma sbagliato lo rifai in un minuto. Una clip sbagliata costa di più, e se la faccia cambia da una scena all'altra il film non regge. Per questo si sistema tutto sulle immagini, prima.
Scena 1 di 5
Lei al semaforo si gira

AI generated
AI generatedI fotogrammi · Nano Banana Pro


AI generatedRiferimenti: il suo profilo per l'iniziale, il primo piano e la scheda del personaggio per il finale.
Prima ho fatto la scena con lei che guarda avanti, poi una passata solo per rendere vera la faccia, con una sua foto di profilo. Per il fotogramma finale le ho girato la testa, con le sue foto di fronte.
Image 1 is the scene, image 2 is a close-up of the same woman in the same left profile. Keep image 1 exactly as it is: same composition, framing, pose, head angle, gaze, light, car, seat, seatbelt and sweater. Change only the realism of her face and hair to match image 2: the same nose, jaw and chin, the same natural skin texture, visible freckles, matte skin without glossy shine, loose flyaway hairs. Same face and identity. No new light sources, no new objects.
Image 1 is the scene, images 2 and 3 show the same woman. Keep image 1 exactly as it is: same composition, framing, light, car, seat, seatbelt, sweater, skin texture and the same woman. Change only her pose: she has turned her head toward the car beside her, almost facing the camera, and her eyes look just slightly past the lens, at the passenger of the other car, not straight into the camera. Her face seen from the front matches images 2 and 3 exactly: same nose, jaw, chin, eyes, lips and freckles. Curious, slightly guarded expression, lips closed. No new light sources, no new objects.
La clip · Kling 3.0
Static camera on a long telephoto lens, shallow depth of field. The woman sits at the wheel of her stopped car, looking ahead through the windscreen, calm and absent-minded. After a moment she senses someone looking at her and slowly turns her head toward the car beside her, her eyes settling just past the camera, curious and slightly guarded. Subtle natural movement, a slight breath, a few strands of hair moving. No camera movement, no other people, no cuts. Sound: quiet city ambience, idling engine.
Scena 2 di 5
Il finestrino si abbassa

AI generated
AI generatedI fotogrammi · Nano Banana Pro


AI generatedImmagine 1: un fotogramma di un reel che mi piaceva, solo per la composizione (non lo metto qui, non è mio). Immagine 2: la scena 1, per capelli e luce. Immagini 3 e 4: la mia faccia.
Use image 1 only for the composition and framing: an over-the-shoulder shot from inside one car, looking out through its open driver's window at the car stopped right next to it. Do not copy the people, the clothes or the car colours from image 1. Foreground, left third of the frame: the back of the head of the woman from image 2, her long straight jet black hair with a centre parting, very out of focus, filling the left edge of the frame, her face not visible. Middle ground, in focus: a dark navy blue car stopped side by side, its front pointing to the right of the frame, its near side window fully rolled down. In the near seat, closest to us, sits the man from images 3 and 4, the passenger: same face, same eyes, same nose, same moustache and short beard, same dark tousled hair with a short fringe, same skin texture, wearing a black crew-neck t-shirt. His head is turned toward us and he looks at the woman in the foreground, his eyes just to the left of the lens, calm neutral expression, lips closed. In the far seat, behind him, a young woman with dark hair tied back is driving, seen in profile, looking straight ahead toward the right of the frame, wearing a plain beige top, softly out of focus. Warm late afternoon sunlight, the same light and colour as image 2, blurred city street in the background, telephoto lens, shallow depth of field, natural documentary photography style, 35mm film grain, no retouching. No dark circles under his eyes, neatly groomed moustache and short beard.
Ritocco della mia espressione · GPT Image 2.5
Nel primo fotogramma guardavo lei con la coda dell'occhio e le sopracciglia aggrottate: sembravo inquietante. Ho cambiato solo l'espressione, tenendo tutto il resto.
Edit image 1, a close-up crop of a man in a car window. Make a clearly visible change to his head and expression. Turn his head a little more toward the viewer, so that his eyes look straight ahead from the centre of the eye sockets, not from the corners of his eyes: no side-eye. Give him an open, friendly, pleasant expression: eyebrows slightly raised and relaxed, no frown at all, eyes bright and gently open, the cheeks lifted a little by a small warm smile with the lips closed, like a man who has just noticed someone nice in the next car. Keep his identity exactly as in image 1: same face, eyes, nose, moustache, beard, hair, skin texture and pores; images 2 and 3 are the same man, use them only to stay true to his face, ignore their background. Keep the window frame, the car interior, the driver behind him, the light, the colours and the film grain of image 1 unchanged. Do not beautify, do not smooth the skin, no dark circles.
Edit image 1. Change ONLY one thing: the side window of the dark blue car, in front of the man in the passenger seat, is now completely closed: the glass goes all the way up to the top of the window frame, with no gap at the top. The glass is lightly tinted dark grey, with a soft, faint reflection of the street on it, so the man behind it is only partly visible, a darker silhouette with his face barely readable. Keep everything else in image 1 exactly identical: the man's head position and pose, the driver, the woman's hair in the foreground, both cars, the light, the colours, the framing and the background. Natural photography, 35mm film grain, no retouching.
La clip · Kling 3.0
Static camera on a long telephoto lens, shallow depth of field, from inside her car past the back of her head. Both cars are stopped at the light and do not move: the dark blue car stays in the same place in the frame. In the car beside her, the dark tinted passenger window slowly rolls down, revealing the man in the passenger seat, ending exactly as in the end frame. The man is alive and natural, never frozen: he shifts a little in his seat, his head moves slightly, he tilts it a touch and gives her a small friendly nod as the window goes down, he breathes and blinks naturally. He keeps the same small, warm smile from start to finish, his eyes soft and relaxed. The young woman driving his car keeps looking straight ahead at the road the whole time, in profile, with calm, steady, relaxed eyes; she blinks normally and only moves her hands a little on the wheel; she never squints, never rolls or darts her eyes, never looks down. The woman in the foreground moves her head slightly and a few strands of her hair move. No camera movement, no cuts. Sound: the soft hum of an electric window motor, idling engines.
Scena 3 di 5
Lei: sorpresa, sorriso, dubbio

AI generatedIl fotogramma · Nano Banana Pro

AI generatedRiferimenti: la scena 1 e una sua foto in primo piano.
Un fotogramma solo per tre clip: cambia solo l'espressione, così la faccia resta la stessa.
Close-up photograph of the woman from images 1 and 2, same face, same nose, same jaw and chin, same freckles and hair, seen through the open driver's window of the same white car with cream leather seats as image 1, her head and shoulders large in the frame. She sits at the wheel wearing the same black crew-neck knit sweater and seatbelt as image 1. Her head is turned toward us and she looks just to the side of the lens, calm neutral expression, lips gently closed. Warm late afternoon sunlight, the same light and colour as image 1, the car interior and the street behind her softly out of focus. Telephoto lens, shallow depth of field, natural documentary photography style, 35mm film grain, no retouching.
Le clip · Kling 3.0
Her car is stopped and does not move. The street seen through the window behind her stays completely still. She looks straight into the lens. As she recognizes the man looking at her, her eyes widen slightly and her eyebrows lift in quiet surprise, her lips part a little, then she holds the look. Subtle and natural, not exaggerated. She does not speak. She blinks once. Nothing else in the frame moves. Keep her face exactly as in the start image. Static camera, telephoto lens, shallow depth of field. Sound: quiet city ambience, idling engine.
Her car is stopped and does not move. The street seen through the window behind her stays completely still. She looks straight into the lens at the man smiling at her, and slowly breaks into a small, warm, slightly shy smile, then holds it. She does not speak. She blinks once. Nothing else in the frame moves. Keep her face exactly as in the start image. Static camera, telephoto lens, shallow depth of field. Sound: quiet city ambience, idling engine.
Her car is stopped and does not move. The street seen through the window behind her stays completely still. She keeps looking straight into the lens the whole time. Right away her expression turns clearly skeptical: she frowns, raises one eyebrow higher than the other, narrows her eyes and tilts her head to the side, as if to say "what are you talking about?" She keeps her lips pressed together the whole time, her mouth stays closed, she does not speak. She holds this skeptical look until the end. Nothing else in the frame moves. Keep her face exactly as in the start image. Static camera, telephoto lens, shallow depth of field. Sound: quiet city ambience, idling engine.
Scena 4 di 5
Io sorrido, poi parlo

AI generatedIl fotogramma · Nano Banana Pro


AI generatedRiferimenti: la mia faccia e la scena 2 con la faccia sfocata, per tenere auto e luce.
Close-up photograph of the man in images 1 and 2, exactly this real man, sitting in the passenger seat of the dark navy blue car from image 3, with its side window fully rolled down, seen from the car stopped next to it. In image 3 his face is blurred on purpose: take his face only from images 1 and 2. His head and shoulders are large in the frame, his face in the same three-quarter angle as image 1, and he looks just to the left of the lens, calm neutral expression, lips closed. Black crew-neck t-shirt. Behind him, softly out of focus, the same young woman from image 3 is driving, seen in profile, in her plain beige top. The same warm late afternoon sunlight and blurred street as image 3. Telephoto lens, shallow depth of field, natural documentary photography style, 35mm film grain, no retouching. No dark circles under his eyes. Do not beautify or redesign his face.
Il sorriso · Kling 3.0
His car is stopped and does not move at all. The window frame around him stays exactly in the same place in the frame. The street seen through the windshield behind them stays completely still. The man in the passenger seat looks straight into the lens and breaks into a calm, warm smile with his lips closed, then holds it. He does not speak. The driver behind him keeps both hands still on the wheel and keeps looking straight ahead. Nothing else in the frame moves. Keep his face exactly as in the start image. Static camera, telephoto lens, shallow depth of field. Sound: quiet city ambience, idling engine.
La battuta · Seedance 2.5, con la mia voce caricata
The man in the passenger seat looks into the lens and says in Italian, with the voice from the audio reference and his lips in sync with it: "Sai che tutto questo non è reale?" Calm and slightly amused. Then he holds the look with a sly, knowing smile. Only small natural head movements. Keep his face exactly as in the start image. The driver behind him keeps looking straight ahead. The camera stays still. Telephoto lens, shallow depth of field. Only his voice from the audio reference, no music.
Scena 5 di 5
L'auto se ne va
AI generated
AI generatedIl fotogramma finale · Nano Banana Pro, in due passate


AI generatedPrima tolgo l'auto, poi giro la sua testa. Una modifica per volta.
Edit image 1. Remove the dark blue car completely, with the man and the driver inside it: the car has driven away. Show only what was behind it: the empty lane and the city street beyond, with the buildings, the trees and the traffic far away softly out of focus, in the same warm late afternoon light. Keep everything else in image 1 exactly identical: the woman's out-of-focus hair in the foreground on the left, the inside of her car, the window frame, the camera position, the framing, the telephoto lens and the colours.
Edit image 1. Change ONLY the woman in the foreground: she has turned her head toward the right of the frame, toward the front of her car, following with her eyes the car that has just driven away, so we now see a little more of her face in soft profile from slightly behind. She is the woman in image 2: same profile, same nose, same lips, same long straight jet black hair, same black sweater. She stays very out of focus, as in image 1. Keep everything else in image 1 exactly identical: the empty street, the buildings, the light, the inside of her car, the window frame, the framing and the colours.
La clip · Kling 3.0
Static camera on a long telephoto lens, shallow depth of field, from inside her car past her head. At first the dark blue car beside her is still stopped and the man in the passenger seat keeps the same easy, friendly look and small warm smile as in the start image, lips closed, his eyes relaxed. Then he calmly turns his head to face straight ahead, toward the windshield, and looks at the road in front of the car, his face now seen in profile. From that moment he keeps looking straight ahead through the windshield at the road until the car is gone: he does not look at the driver next to him and he does not look back at the woman. The car pulls away forward and drives out of frame to the right, revealing the empty street behind it. As the car leaves, she slowly turns her head to the right to follow it, ending exactly as in the end frame. The car moves forward only, it does not fade or vanish. No camera movement, no cuts. Sound: an engine pulling away, quiet city ambience.
Le domande che mi fanno
Quali strumenti servono per fare un mini film con l'AI?
Io ho usato Nano Banana Pro per i fotogrammi, Kling 3.0 per le clip e Seedance 2.5 per la clip in cui parlo, con la mia voce caricata. Il ritocco di un'espressione l'ho fatto con GPT Image 2.5. Poi un programma di montaggio per mettere insieme clip, musica e cartello.
Perché prima le immagini e poi il video?
Un fotogramma sbagliato si rifà in un minuto. Una clip sbagliata costa di più. Se la faccia cambia da una scena all'altra il film non regge, quindi la clip la genero solo quando tutti i fotogrammi tornano.
Come si tiene la stessa faccia in tutte le scene?
Ogni fotogramma parte dai riferimenti: le foto delle facce, la scheda del personaggio e le scene già fatte. Cambio una cosa per volta. Nella scena 3 un fotogramma solo fa partire tre clip, e cambia solo l'espressione.
Quanto dura una clip fatta con Kling?
Le mie sono da 5 secondi, a 1080p. Quando c'è un movimento faccio anche il fotogramma finale: la clip parte dal primo e arriva al secondo.
Chi può aiutarmi a fare un video con l'AI per il mio brand?
Io. Faccio immagini di campagna, spot e film brevi con l'AI, con una regia e un'art direction dietro. Si comincia con una call di 30 minuti. Se invece vuoi imparare a lavorare così, c'è l'affiancamento.
Altre guide
- Claude Design è gratis: un volantino nel piano Free
- Comfy Agent: l'agente che monta i workflow di ComfyUI
- Come fare immagini di campagna coerenti con Claude e Gemini
- Impaginazione automatica in InDesign, senza AI
- Come rinominare le immagini di reference con Claude
- Stitch e Pomelli, gli strumenti gratis di Google per la grafica
Se ti è servita, seguimi
Su Instagram faccio vedere come lavoro con l'AI, da art director. Se la vuoi girare a qualcuno, mandagli questa pagina.