Found Footage turns one pick - what the footage catches - into a vertical clip built to read as a real recording rather than a film: an imperfect frame, a camera that reacts late, focus that hunts, sound the camera itself picked up, and a hard cut before the question is answered. The rig and the place come with the sighting, so there is nothing else to choose. Nothing is uploaded and nothing is edited afterwards.
It is a short clip built to look like an accidental recording of something that should not exist: a shot of a creature or an anomaly, half-obscured, badly framed, and cut off before anyone gets a clear look.
The format works on the gap between what is in the frame and how badly it is filmed. A clean, well-composed shot of a lake monster reads as a visual effect. A smeared, over-exposed, half-blocked one reads as evidence, which is why every part of this template is asking the model for a worse-looking image on purpose.
The ones spreading now are generated. The TikTok account .ai.enthusiast has passed 13 million views on a single Amazon river sighting in this exact format, and the whole genre is built to leave the question open rather than to answer it.
That is also the honest reason the template refuses to give you a clear shot: a clip that resolves the creature stops being the format. Aicut generates the clip; what you caption it is yours.
One pick: what the footage catches. Four presets - a giant, a pale figure, a lake monster or a tentacle - or your own sentence. Everything else is already written into the template.
You describe the creature and nothing else. The distance, the occlusion, the beat timing, the camera's reaction, the lens artefacts, the sound and the hard cut are all the template's job, so a plain description works as well as a carefully worded one.
No, and that is deliberate. Every clip is shot either from the deck of a small boat or from the open door of a low-flying helicopter, and which one you get follows the sighting you picked: a lake monster is filmed from the water, a giant can be filmed from either.
Those are the two rigs the real viral clips in this format actually use. A drone was never one of them - drone footage is too smooth and too high to read as an accident, which is the opposite of what this format needs. Tying the rig to the sighting instead of asking you also removes the one way the old form could contradict itself: a lake monster filmed on a desert highway.
The rig is not cosmetic. It decides what sits in the bottom of the frame and what the microphone picks up - the boat's engine idling under everything, or the rotor loud over it. A helicopter clip with no rotor on the audio track reads as a composite immediately, which is one reason the sound is generated with the picture instead of being added afterwards.
No. Found Footage is text-only, which is what separates it from most AI video tools in this genre - those animate a still you supply, so you need a subject before you can start.
Because nothing is uploaded, the sighting can be anything you can describe. The sighting picker takes a custom answer as well as its presets, so a creature that is not on the list is one sentence away.
Yes, and only the kind a real camera would have caught: wind on the microphone, water, the engine or the rotor of whatever the operator is on, and the reaction of whoever is filming. No music and no narration.
This is the single loudest tell of a generated clip, so the template states it twice and the model renders the audio in the same pass as the picture rather than as a separate step.
A 9:16 clip at 720p, ten seconds by default, in about two and a half minutes - already framed for TikTok, Instagram Reels and YouTube Shorts.
The length is a real choice rather than a slider over one script. The beats are written as timecodes - the ordinary approach, the reveal, the panic, the grab, and the aftermath on the last frames - and every offered length gets its own timings rather than being squeezed into a script written for a different one. Six, eight, ten or twelve seconds under Advanced, and each one is a script written for that length instead of a long clip cut off before its payoff. The video lands in your Creations, where you can publish it to TikTok, YouTube and Instagram.
4 tokens for the default 10-second clip on Seedance 2.0 Mini at 720p. That is about 12 videos a month on Creator ($19.99, 50 tokens) or 25 on Automate ($39.99, 100 tokens).
All three models are priced per second, so the number moves with the length you pick and the generate button always carries it before you spend anything. On Seedance 2.0 Mini at 720p: 2.4 tokens at 6 seconds, 3.2 at 8, 4 at 10 and 4.8 at 12.
Seedance 2.5 and Seedance 2.0 are the better-looking renders and both cost 1.2 tokens a second at 720p - three times Mini - so the same four lengths are 7.2, 9.6, 12 and 14.4 tokens on either. Dropping to 480p under Advanced roughly halves every one of those.
