Seedance 2.5 prompts: 3 found-footage clips that look filmed on a phone
Seedance 2.5 prompt on Twistavo, an AI video studio for character swap: replace a person in a real video from 1 to 4 photos; plans from $3.99/month.
Updated
These three clips come from our own Seedance 2.5 runs (August and September 2026), made from text only, with no reference image. Each one tries the same trick: one impossible event filmed as if a bystander held a phone. The prompts are long on purpose. They fix the camera position, the light, the scale references and the sound, so the model has nothing left to invent except the event. Below are the exact prompts as we sent them, the settings, what Twistavo charges today for the same settings, and what went wrong.
- Seedance 2.5: specs, inputs and credits
- Seedance 2.0 vs Seedance 2.5: what changed
- Kling 4.0 Flash: another text-to-video model to run these prompts on
Setup
- Model
- Seedance 2.5
- Settings
- Seedance 2.5 text-to-video, 9:16, one continuous take, native audio on. The whale and cinema clips ran at 480p with the standard bitrate; the kayak clip ran at 720p with the high bitrate. The clips on this page are the original results, muted and compressed for the web: the originals carry the model's own sound (no music was asked for or added).
- Models used
- Seedance 2.5
1. A sperm whale glides over a seaside cafe terrace (15 s)
- Text to video, no reference
- Aspect ratio 9:16
- Duration 15 s
- Resolution 480p
- Audio on
- Bitrate standard
386 credits in Twistavo
Prompt
An impossibly enormous sperm whale glides through open air just above an ordinary Mediterranean seaside cafe terrace in daylight. Its square head and long charcoal-grey body are already visibly suspended above the umbrellas at the opening. Adult customers, tables and the stone buildings behind them make its overwhelming scale unmistakable: the whale is as long as two city buses. It is one solid, anatomically convincing animal, not a shape made of clouds. Render the entire scene as extremely realistic, ungraded live-action: wrinkled whale skin with subtle natural imperfections, ordinary weathered facades, woven cafe chairs, canvas umbrellas and ceramic cups. Low afternoon sunlight illuminates the animal and the terrace from the same direction. Its broad shadow travels across the paving, tables and people as its body passes overhead. Flying is the only impossible thing; the surroundings retain their everyday material behavior. Stage one continuous, readable passage from left to right across the open plaza. Keep the whale clearly separated from rooftops, umbrellas and people. Observe from terrace level with enough width to retain customers and architecture as scale references; a restrained upward reframe follows the animal without hiding its body or turning the scene into a montage. The customers interrupt their ordinary actions at different moments: a cup stops halfway to a mouth, two heads turn upward, one person slowly rises from a chair. Subtle reactions, not synchronized theatrical screaming. Umbrella fabric stirs gently. The heavy torso passes, followed by a slow natural movement of the tail flukes. End after the tail clears the terrace and sunlight returns to the tables, with the customers still staring after it. Preserve the same whale, people and location throughout. Native sound: cafe ambience, a ceramic cup touching its saucer, wind through umbrella fabric, quiet surprised breaths and a deep organic rumble during the passage. No music, narration or scripted dialogue. No written text, logo, watermark, fantasy glow, cartoon styling or glossy video-game surfaces.
Tip: What worked: naming two scale references (customers and stone buildings, plus "as long as two city buses") kept the whale the same size for all 15 seconds, and "flying is the only impossible thing" kept the cups, chairs and umbrellas behaving normally. What did not: from second 6 to 12 the whale's belly skims the umbrella tops, although the prompt asked for clear separation, and the body fills the top half of the frame. If you want air between the animal and the umbrellas, give a distance ("two storeys above the umbrellas") instead of an adjective.
2. A 4D cinema screen spills a wave into the aisle (12 s)
- Text to video, no reference
- Aspect ratio 9:16
- Duration 12 s
- Resolution 480p
- Audio on
- Bitrate standard
309 credits in Twistavo
Prompt
Original fictional 4D cinema video, photorealistic and shot like a surprised adult audience member recording vertically from the middle row. Viewer IS the recording camera: no phone, screen-in-a-phone, filming hand, selfie stick or rig inside the image. Normal red cinema seats with a handful of adult patrons, aisle and a huge wall-to-wall projection screen visible. Natural imperfect handheld movement and sensible exposure. One continuous readable event, never a trailer montage. Keep theatre architecture and audience positions coherent. No real film footage, celebrities, licensed characters, writing, logos, subtitles or watermark. Native room tone, mechanical seat sounds, startled gasps and precisely synchronized environmental effects. No music and no spoken narration. All people remain safe; this is a playful impossible cinema illusion, not an actual disaster. The projection shows a sunlit empty tropical shoreline, with a modest turquoise wave approaching the camera inside the projected scene. 0-3s the sea looks almost convincingly connected to the aisle. 3-7s a shallow sheet of turquoise water spills smoothly out from the BOTTOM EDGE of the projection and flows gently along the theatre aisle; keep its height below the soles of the seats, never flooding or submerging anyone. Two nearby patrons lift their feet, laughing with surprise. 7-10s a little rubber duck floats down the aisle and bumps gently against the front seat. 10-12s one patron lifts their 3D glasses, looks at the duck, then toward the recording camera in disbelief. The theatre never becomes an outdoor beach. The water-screen boundary and the duck are both readable. The reveal is delightful and absurd, no destruction, panic, screaming or injury.
Tip: What worked: the second-by-second beats (0-3 s, 3-7 s, 7-10 s, 10-12 s) landed in order, and the closing beat, a patron lifting the 3D glasses and looking into the lens, is the moment that sells it as phone footage. "Viewer IS the recording camera" kept phones and hands out of the frame. What did not: the spill itself happens off camera. The first 4 seconds hold on the screen with no water, then the camera turns to the row and the water is already on the floor. If the spill is your hook, ask for the bottom edge of the screen to stay in frame until second 7.
3. A kayaker rolls under water holding a full cup (10 s)
- Text to video, no reference
- Aspect ratio 9:16
- Duration 10 s
- Resolution 720p
- Audio on
- Bitrate high
601 credits in Twistavo
Prompt
Raw vertical smartphone witness video filmed from knee-deep crystal-clear turquoise water beside a quiet rocky cove in daylight. An ordinary adult kayaker sits in a narrow recreational kayak only a few metres from the phone, already holding a completely full transparent plastic cup high above one shoulder so the challenge is understood immediately. Friends nearby laugh and gasp. The kayaker slowly leans sideways, the kayak rolls fully upside down and the person disappears under the water while the raised arm fights to keep the cup upright above the surface. Hold the suspense naturally for a moment. Then the kayaker completes the roll and resurfaces upright, soaked and laughing, still holding the visibly full cup without spilling. One believable continuous handheld phone take at water level, imperfect framing, natural autofocus and exposure, real water resistance and splashes, no cuts, no slow motion, no polished commercial look. Native synchronized water splashes, paddle and kayak creaks, surprised laughter and cheering only; no narration, no music. No text, no subtitles, no logo, no brand, no extra limbs, no duplicated people, no impossible morphing. End directly on the successful reveal and genuine reaction.
Tip: What worked: the cup is raised in the first frame ("so the challenge is understood immediately"), which gives the clip its hook before anything happens, and the ending lands on the full cup and the cheering. What did not: "at water level" let the camera dip under the surface for the middle of the clip, so you watch the hull from below instead of the arm holding the cup up. Write "the phone stays above the surface" to keep the stunt in view. This was also the most expensive run here: 720p with the high bitrate costs about twice the 480p clips per second.
Make it in Twistavo
- Open Create in Twistavo and choose Seedance 2.5 in the video models.
- Leave the image slots empty: all three clips are text to video.
- Paste one prompt as written, then change only the event sentence (the whale, the wave, the cup) to your own idea.
- Set Aspect ratio 9:16, the duration and resolution listed under the example, and keep audio on if you want the model's own sound.
- Check the credit quote shown before you run it, and start at 480p: the cost goes up with resolution, bitrate and seconds.
- Run two or three takes and keep the one where the event stays readable from the first to the last second.
Variations to try
- Swap the event and keep the camera block: a hot-air balloon landing in a supermarket car park, filmed from a parked car.
- Keep the cinema prompt and change what leaves the screen: snow, autumn leaves or a paper plane instead of water.
- Turn the kayak stunt into a bicycle wheelie over a puddle with a full coffee cup, same bystander framing and same no-music rule.
- Ask for a locked camera instead of a handheld one to compare how much of the realism comes from the shake.
Notes
All three clips are AI-generated: the people, places and animals do not exist. Footage built to look like it was filmed on a phone should be labelled as AI when you publish it, and most platforms now ask for that label. Credit costs are Twistavo's quotes for the same model and settings on November 18, 2026; your quote can change with the settings you pick. Results change from run to run.