Gangsta Bird
Find an accomplice
Prompt 2
SEEDANCE 2.5 · IMAGE TO VIDEO · @IMAGE 1 FORMAT: 15s / 9:16 vertical / one continuous handheld shot, no cuts / natural sync sound, no music STYLE: raw candid smartphone video shot from across the street, looks like a bystander filmed it and posted it. Photorealistic, low contrast, muted desaturated colors, flat overcast daylight, real concrete and asphalt texture, phone sensor noise, slight rolling-shutter wobble on fast moves, focus hunting, digital zoom softness when zoomed in. Handheld level 2 to 3: the camera is a phone held in someone's hands, it bounces with the operator's breathing and small steps, reframing lags behind the action, no smooth gimbal, no stabilization. Everything is real animals with accurate anatomy, size and feather/fur texture. No CGI feel, no cinematic grading. REFERENCES: @Image 1 is the exact FIRST frame, start with no change. Lock everything in it: the grey feral pigeon standing upright on the cracked sidewalk, the small black Wayfarer sunglasses on its face, the small black pistol held in its right wing, the grey concrete wall, the curb, the litter, the cigarette butt, the white lane line, the passerby's legs in dark shorts and black clogs at the right edge. The pigeon is the only hero. The passerby simply finishes walking out of frame to the right. VIDEO PROMPT: 0.0s to 2.0s: Open on the exact frame of @Image 1. The passerby's legs step out of frame right. The pigeon turns its head, then starts walking, upright on two feet with a slow, unbothered swagger, pistol still in its right wing hanging at its side, barrel down. It steps off the curb onto the asphalt. The phone camera drifts slightly and reframes late, a small bounce from the operator shifting weight. 2.0s to 4.0s: The camera performs a shaky digital zoom in on the pigeon as it walks along the road toward the camera side of the street, slight focus hunt, the image gets a bit softer and noisier as the zoom tightens. Sunglasses stay on. 4.0s to 6.0s: The camera pulls the zoom back a little and pans left, revealing the middle of the street ahead of the pigeon: a scruffy tabby street cat crouched on the asphalt with a real small grey mouse held in its mouth, and two more street cats, one black and one orange, sitting close on either side of it, all three looking down at the mouse. The pigeon keeps walking straight at them. Distant traffic passes in the background. 6.0s to 8.5s: The pigeon stops about two meters from the cats, plants its feet, raises the pistol straight up over its head with its wing, pointed only at the empty sky, and fires two short blank shots upward, then a third. Small sharp muzzle flash, a puff of smoke, the wing kicks slightly with each recoil. The phone camera jolts hard with each crack, the operator flinches, the frame jumps and over-rotates, then corrects back onto the pigeon. Nothing and no one is hit. 8.5s to 11.0s: All three cats startle at the first crack. The tabby drops the mouse. The cats bolt with real cat physics, low bodies, scrambling paws, tails puffed: the tabby sprints far down the road and shrinks into the distance, the black cat darts left under a parked car, the orange cat leaps sideways to the right and vanishes behind the curb. The freed mouse scurries fast across the asphalt and squeezes into a storm drain opening in the curb near the camera. The camera whips toward the fleeing cats, lags, then finds the pigeon again. 11.0s to 15.0s: The pigeon lowers the pistol back to its side, barrel down, and continues its walk straight toward the camera across the road with the same slow gangster swagger, head bobbing, feet clicking on the asphalt. The camera performs a second shaky digital zoom in, slowly tightening on the pigeon's head and black sunglasses as it walks, focus breathing, sensor noise rising. The pigeon fills the lower frame, sunglasses on, pistol hanging at its side, and walks past the lens at the last moment. End on the pigeon close and slightly out of focus, camera still handheld and unsteady. AUDIO: Natural street ambience only: distant traffic hum, a car passing, wind on the phone microphone, the passerby's clogs slapping away, the pigeon's claws ticking on concrete then asphalt, <three short sharp pistol cracks echoing off the concrete wall, a fraction apart>, a muffled phone-mic clip and the operator's small startled breath on the first crack, cats hissing and yowling once, the scramble of paws on asphalt, a tiny mouse squeak and a scrape at the drain, then the ambience settles back to traffic and the pigeon's steady footsteps growing closer to the microphone. No music, no narration, no spoken words. CONSISTENCY: The pigeon is the same bird from @Image 1 for the entire shot, identical feather markings and size. The black Wayfarer sunglasses stay on its face for the whole video and are never removed, never slip. The pistol never leaves its right wing. The pigeon stays upright on two feet, real pigeon anatomy, walks like a small person. The cats are real street cats, the mouse is a real small mouse. The location is the same street, same grey wall and curb, from the same across-the-street viewpoint. One continuous shot from one phone. NEGATIVE PROMPT: no cuts, no scene change, no smooth gimbal movement, no stabilization, no cinematic color grading, no cartoon, no CGI look, no plastic or toy look, no extra pigeons, no extra people in the foreground, no camera or phone visible, no drone, no sunglasses removed, no pistol dropped, no pistol pointed at any animal or person, no injuries, no blood, no cats hit, no mouse hurt, no floating objects, no teleporting cats, no morphing, no warping, no text, no captions, no watermark, no music, no narration.
Other prompts in this tutorial
Prompt 1
NANO BANANA PRO · TEXT TO IMAGE · NO REFERENCE Low-quality candid smartphone photo taken from across the street, handheld, slightly tilted, mild JPEG compression, soft focus, phone sensor noise, muted desaturated colors, flat overcast daylight with no visible sun. Real feral animal with accurate anatomy, size and feather/fur texture, nothing cartoonish, nothing CGI. The animal looks like it was genuinely photographed on the street, not composited. Not posed, not staged. No text, no watermark, no visible faces of people. A grey feral pigeon standing upright on both feet on a cracked concrete sidewalk like a small person, wearing tiny black Ray-Ban Wayfarer sunglasses, holding a small black pistol loosely in one wing at its side, barrel pointed at the ground, head turned slightly as if watching someone leave. Setting: a plain grey rendered wall behind, litter and a cigarette butt in the gutter, faded white lane line on the asphalt in the foreground, the legs of a passerby in dark shorts and black clogs cut off at the right edge of the frame mid-stride. Eye level, shot from about 8 meters away with a phone camera, 4:3 framing, pigeon small in frame.
Tutorial
DENIZ AKKABAK · SEEDANCE 2.5 TUTORIAL · @DENIZ.AKKABAK
MADE IN INVIDEO · invideo.io/i/DENIZ-ig
Gangsta Bird
A feral pigeon in Wayfarers walks up to three street cats, fires three blanks at the sky, frees a mouse — and the phone filming it flinches like a real person.
A TUTORIAL BY DENIZ AKKABAK · @DENIZ.AKKABAK
PIPELINE NANO BANANA PRO → SEEDANCE 2.5 · FORMAT 15S / 9:16 / ONE SHOT · GENRE STREET COMEDY
Nobody believes cinema anymore, but everybody believes a stranger's phone. That is the whole trick behind this video. The pigeon, the sunglasses and the pistol are the joke — the credibility comes from everything the prompt says about the camera: sensor noise, rolling-shutter wobble, focus hunting, a digital zoom that gets softer the tighter it goes, an operator who breathes, shifts weight and reframes late. The prompt spends more words on the flaws of the phone than on the bird, and that is exactly why the bird reads as real.
The workflow is two moves inside invideo. Nano Banana Pro casts the impossible character in a deliberately low-quality "witness photo" — a pigeon small in frame, seen from across the street, with a passerby's legs as scale. Seedance 2.5 then locks that photo as the exact first frame and plays out fifteen seconds of choreography by the second: walk, reveal, three blanks at the sky, cats bolt, mouse escapes, swagger into the lens. The still sets the quality ceiling; the video never has to fake more polish than a phone would have.
THE PIPELINE AT A GLANCE
- 01
The witness photo — Nano Banana Pro
One text prompt, no reference. A candid, tilted, JPEG-crushed smartphone photo from eight meters away: grey pigeon, Wayfarers, pistol at its side, cracked sidewalk, litter, a passerby's legs cut off at the right edge. Low quality is the specification, not a defect.
- 02
The standoff — Seedance 2.5
The photo becomes @Image 1, the frame-zero lock. A timecoded prompt runs six beats across fifteen seconds while the AUDIO block writes the three pistol cracks, the phone-mic clip and the mouse squeak. One continuous handheld shot, natural sync sound, no music.
STEP 01 — NANO BANANA PRO
The Witness Photo
The still has one job: look like it was taken by someone who did not mean to take it. Everything in the first paragraph is a quality downgrade — "slightly tilted, mild JPEG compression, soft focus, phone sensor noise" — and everything in the second paragraph is staging that pretends not to be staging. The pigeon stands "like a small person" but the prompt never says pose; it says "head turned slightly as if watching someone leave", which gives the bird a thought instead of a stance. The passerby's legs "cut off at the right edge of the frame mid-stride" do three things at once: they prove scale, they prove the photo was cropped by chance, and they hand Seedance a character who can simply walk out of frame in the first two seconds.
Two guardrails matter here. "Real feral animal with accurate anatomy … nothing cartoonish, nothing CGI" stops the model from smoothing the pigeon into a mascot, and "No text, no watermark, no visible faces of people" keeps the frame free of anything that would break the bystander illusion. Eight meters, eye level, pigeon small in frame — the distance is what makes the pistol plausible; up close it would be a prop, from across the street it is a rumour.
Prompt 1
NANO BANANA PRO · TEXT TO IMAGE · NO REFERENCE Low-quality candid smartphone photo taken from across the street, handheld, slightly tilted, mild JPEG compression, soft focus, phone sensor noise, muted desaturated colors, flat overcast daylight with no visible sun. Real feral animal with accurate anatomy, size and feather/fur texture, nothing cartoonish, nothing CGI. The animal looks like it was genuinely photographed on the street, not composited. Not posed, not staged. No text, no watermark, no visible faces of people. A grey feral pigeon standing upright on both feet on a cracked concrete sidewalk like a small person, wearing tiny black Ray-Ban Wayfarer sunglasses, holding a small black pistol loosely in one wing at its side, barrel pointed at the ground, head turned slightly as if watching someone leave. Setting: a plain grey rendered wall behind, litter and a cigarette butt in the gutter, faded white lane line on the asphalt in the foreground, the legs of a passerby in dark shorts and black clogs cut off at the right edge of the frame mid-stride. Eye level, shot from about 8 meters away with a phone camera, 4:3 framing, pigeon small in frame.
OUTPUT
STEP 02 — SEEDANCE 2.5 IN INVIDEO
The Standoff
Attach the still as @Image 1 and the prompt treats it like a crime-scene inventory: "@Image 1 is the exact FIRST frame, start with no change", followed by a list of every object it must keep — the sunglasses, the pistol in the right wing, the wall, the curb, the litter, the cigarette butt, the lane line, the legs. Then it retires the only element that could compete with the pigeon: "The passerby simply finishes walking out of frame to the right." From that point the street belongs to the bird.
The VIDEO PROMPT is written in six timecoded blocks, and each block pairs one camera behaviour with one action. The reveal is the smartest beat: at 4.0s the phone "pulls the zoom back a little and pans left" to find the three cats and the mouse — the audience discovers them at the same moment the operator does, which is how a real clip would feel. At 6.0s the pistol goes up "pointed only at the empty sky" and the camera "jolts hard with each crack"; the recoil is written into the operator, not just the gun. At 11.0s a second digital zoom tightens on the sunglasses as the pigeon walks past the lens, and the prompt ends the shot "slightly out of focus, camera still handheld and unsteady" — refusing a clean final frame is what keeps it candid.
REFERENCE INPUT — ATTACH IN THIS ORDER
Prompt 2
SEEDANCE 2.5 · IMAGE TO VIDEO · @IMAGE 1
15s / 9:16 vertical / one continuous handheld shot, no cuts / natural sync sound, no music
raw candid smartphone video shot from across the street, looks like a bystander filmed it and posted it. Photorealistic, low contrast, muted desaturated colors, flat overcast daylight, real concrete and asphalt texture, phone sensor noise, slight rolling-shutter wobble on fast moves, focus hunting, digital zoom softness when zoomed in. Handheld level 2 to 3: the camera is a phone held in someone's hands, it bounces with the operator's breathing and small steps, reframing lags behind the action, no smooth gimbal, no stabilization. Everything is real animals with accurate anatomy, size and feather/fur texture. No CGI feel, no cinematic grading.
@Image 1 is the exact FIRST frame, start with no change. Lock everything in it: the grey feral pigeon standing upright on the cracked sidewalk, the small black Wayfarer sunglasses on its face, the small black pistol held in its right wing, the grey concrete wall, the curb, the litter, the cigarette butt, the white lane line, the passerby's legs in dark shorts and black clogs at the right edge. The pigeon is the only hero. The passerby simply finishes walking out of frame to the right.
0.0s to 2.0s: Open on the exact frame of @Image 1. The passerby's legs step out of frame right. The pigeon turns its head, then starts walking, upright on two feet with a slow, unbothered swagger, pistol still in its right wing hanging at its side, barrel down. It steps off the curb onto the asphalt. The phone camera drifts slightly and reframes late, a small bounce from the operator shifting weight. 2.0s to 4.0s: The camera performs a shaky digital zoom in on the pigeon as it walks along the road toward the camera side of the street, slight focus hunt, the image gets a bit softer and noisier as the zoom tightens. Sunglasses stay on. 4.0s to 6.0s: The camera pulls the zoom back a little and pans left, revealing the middle of the street ahead of the pigeon: a scruffy tabby street cat crouched on the asphalt with a real small grey mouse held in its mouth, and two more street cats, one black and one orange, sitting close on either side of it, all three looking down at the mouse. The pigeon keeps walking straight at them. Distant traffic passes in the background. 6.0s to 8.5s: The pigeon stops about two meters from the cats, plants its feet, raises the pistol straight up over its head with its wing, pointed only at the empty sky, and fires two short blank shots upward, then a third. Small sharp muzzle flash, a puff of smoke, the wing kicks slightly with each recoil. The phone camera jolts hard with each crack, the operator flinches, the frame jumps and over-rotates, then corrects back onto the pigeon. Nothing and no one is hit. 8.5s to 11.0s: All three cats startle at the first crack. The tabby drops the mouse. The cats bolt with real cat physics, low bodies, scrambling paws, tails puffed: the tabby sprints far down the road and shrinks into the distance, the black cat darts left under a parked car, the orange cat leaps sideways to the right and vanishes behind the curb. The freed mouse scurries fast across the asphalt and squeezes into a storm drain opening in the curb near the camera. The camera whips toward the fleeing cats, lags, then finds the pigeon again. 11.0s to 15.0s: The pigeon lowers the pistol back to its side, barrel down, and continues its walk straight toward the camera across the road with the same slow gangster swagger, head bobbing, feet clicking on the asphalt. The camera performs a second shaky digital zoom in, slowly tightening on the pigeon's head and black sunglasses as it walks, focus breathing, sensor noise rising. The pigeon fills the lower frame, sunglasses on, pistol hanging at its side, and walks past the lens at the last moment. End on the pigeon close and slightly out of focus, camera still handheld and unsteady.
Natural street ambience only: distant traffic hum, a car passing, wind on the phone microphone, the passerby's clogs slapping away, the pigeon's claws ticking on concrete then asphalt, <three short sharp pistol cracks echoing off the concrete wall, a fraction apart>, a muffled phone-mic clip and the operator's small startled breath on the first crack, cats hissing and yowling once, the scramble of paws on asphalt, a tiny mouse squeak and a scrape at the drain, then the ambience settles back to traffic and the pigeon's steady footsteps growing closer to the microphone. No music, no narration, no spoken words.
The pigeon is the same bird from @Image 1 for the entire shot, identical feather markings and size. The black Wayfarer sunglasses stay on its face for the whole video and are never removed, never slip. The pistol never leaves its right wing. The pigeon stays upright on two feet, real pigeon anatomy, walks like a small person. The cats are real street cats, the mouse is a real small mouse. The location is the same street, same grey wall and curb, from the same across-the-street viewpoint. One continuous shot from one phone.
no cuts, no scene change, no smooth gimbal movement, no stabilization, no cinematic color grading, no cartoon, no CGI look, no plastic or toy look, no extra pigeons, no extra people in the foreground, no camera or phone visible, no drone, no sunglasses removed, no pistol dropped, no pistol pointed at any animal or person, no injuries, no blood, no cats hit, no mouse hurt, no floating objects, no teleporting cats, no morphing, no warping, no text, no captions, no watermark, no music, no narration.
OUTPUT — CHECKPOINTS
STEP 03
Anatomy — Why the Bystander's Phone Sells a Pigeon With a Pistol
Seven decisions carry this prompt. Each one is tied to a line you can point at.
- 1
The contract
"15s / 9:16 vertical / one continuous handheld shot, no cuts / natural sync sound, no music." One line tells the model the whole piece lives inside a single clip someone posted. No cuts means no editing grammar to hide behind, and no music means the pistol cracks have to carry the drama alone.
- 2
Ugly on purpose
"Low contrast, muted desaturated colors … phone sensor noise, slight rolling-shutter wobble on fast moves, focus hunting, digital zoom softness when zoomed in." Every flaw is a realism token, and every one of them matches the still — the video never becomes sharper than the photo it started from.
- 3
The operator is a character
"Handheld level 2 to 3 … it bounces with the operator's breathing and small steps, reframing lags behind the action." Later, at 6.0s, "the operator flinches, the frame jumps and over-rotates." The person holding the phone reacts to the story, and their reaction is what makes the gunshots feel loud.
- 4
Frame-zero lock
"@Image 1 is the exact FIRST frame, start with no change" plus an itemised inventory down to "the cigarette butt" and "the white lane line". The still is not a mood board, it is the opening frame — and the CONSISTENCY block re-states the three things that must survive: same bird, sunglasses never slip, pistol never leaves the right wing.
- 5
Choreography by the second
Six blocks, six beats: walk (0–2), zoom (2–4), pan-left reveal (4–6), three blanks (6–8.5), the scatter (8.5–11), the swagger into the lens (11–15). Each block gives the camera one move and the animals one action, so nothing has to happen simultaneously and nothing has room to morph.
- 6
Sound as structure
The AUDIO block scripts the clip like a foley sheet: "the passerby's clogs slapping away", "three short sharp pistol cracks echoing off the concrete wall, a fraction apart", "a muffled phone-mic clip and the operator's small startled breath", "a tiny mouse squeak and a scrape at the drain". The shots are written in sound before they are seen, which is how the model learns to time the recoil.
- 7
The insurance
"No pistol pointed at any animal or person, no injuries, no blood, no cats hit, no mouse hurt" keeps the comedy harmless, while "no smooth gimbal movement, no stabilization, no cinematic color grading" keeps the camera bad. The NEGATIVE PROMPT protects both ends of the joke: nobody gets hurt, and nobody gets a nice shot.
THE RESULT
Output — A Pigeon Clears the Street
Check it against the prompt: the passerby's clogs leave the frame in the first two seconds, the pigeon steps off the curb with the pistol barrel down, the pan left at four seconds finds three cats over one mouse, the pistol goes straight up with a visible muzzle flash and the frame jolts, the tabby shrinks into the distance while the black cat goes under the parked car, the mouse reaches the storm drain, and the final zoom tightens on Wayfarers that never slipped. Where the video slightly outruns the script — this cut lands at roughly thirteen and a half seconds instead of fifteen — the beats are all there, in order, in one shot.
THE WORST CAMERA WINS
Nobody argues with a bystander's phone.
Swap the pigeon for a raccoon with a briefcase, swap the cats for a row of seagulls, swap the pistol for a whistle — the skeleton carries any candid street comedy. A deliberately cheap still sets the quality ceiling, a frame-zero lock turns it into the opening shot, six timecoded beats give the camera and the animals one job each, the AUDIO block times the punchline, and the NEGATIVE PROMPT makes sure nobody gets hurt and nobody gets a nice shot.
Nano Banana Pro and Seedance 2.5 both run inside invideo — the witness photo and the standoff were made in the same workspace, one after the other, with the still passed straight into the video model as @Image 1. Start with the link below and rebuild this shot in an afternoon.
WRITTEN, DIRECTED & PROMPTED BY DENIZ AKKABAK · @DENIZ.AKKABAK