Ice-rescue snow-saw bars write the cut into Veo 3.1 at 40cr, 4K, native audio
An ice-rescue snow-saw bar cutting a photographed block writes the kerf and the diegetic cut into Veo 3.1 at 40cr, 4K, 4s / 6s / 8s, native audio; do not generate a victim on the ice.
An ice-rescue snow-saw bar cutting a photographed block writes the kerf and the diegetic cut into Veo 3.1 at 40cr, 4K, 4s / 6s / 8s, native audio; do not generate a victim on the ice.
A snow-saw listing is this bar: tooth pitch, rail length, the drive you actually ship, the ice block on a training pad you own. The leftover is the kerf and the sound of the cut. A body on the ice is a person you did not consent. A constructed instructor walking “how to cut a victim hole” is a presenter you did not hire. Photograph the bar. Write the cut. Leave the casualty off the frame.
Photograph the bar and the ice block
Procurement needs the serial they quoted: bar length, tooth family, the scabbard, the ice you actually train on. Text-to-video of “dramatic ice rescue, victim in the hole, cinematic lake” invents a cousin bar and a private figure.
Shoot the approved bar as a line card: square-on of the rail, teeth readable. A demo ice block on a pad you own is the second plate — empty of people, empty of a hole you would call a victim opening. Crop a name badge. Crop a real casualty drill you do not have a release for. Hardware is the spec. A body is not.
The AI product video generator is the hardware contract: reference photos are the ground truth. Zoom to 100% and read tooth pitch before you keep a take. Prompt only camera and cut on the unit. Re-describing “ice rescue in progress” is how the model draws a second chassis and a person in the water. If the only photo includes a gloved hand or a face, crop the person or shoot again. Empty bar, empty block.
Veo 3.1 writes the kerf and the diegetic cut
Veo 3.1 is Google's enhanced VEO 3 row: text-to-video, no still required, official 4K tier, clips of 4s, 6s, or 8s, listed 40 credits. Native audio is on. A silent prompt still gets a soundtrack you did not choose. This leftover spends that mix on the cut — the hiss of the bar, ice chips, a kerf opening in a block, not in a torso.
Write the sound the way you write the lens. Name the bar. Name the block. Quote nothing a mouth would say. “Locked close on an ice-rescue snow-saw bar cutting a photographed training block, kerf opening, ice chips, diegetic cut, no people.” Stop. “A certified ice-rescue expert explaining what you should do if someone falls through” is a health-shaped persona.
Duration is a closed set: 4s, 6s, 8s. There is no 5-second step and no 10-second hold. Start at 4s. Pad to 8s and the back half invents a second hole. Forty credits is the catalog price, including the audio pass. 4K is the ceiling, not a licence to mint a casualty.
Write the sound into the prompt is the row's own audio rule. Diegetic means the sound is in the picture: kerf, chips, motor. It is not a host.
Product video versus an empty-handed text start
AI video generator is the launcher. Generate a video from text is the named agent job when you start empty-handed: no still required, Veo 3.1 as the row. That is the right door for a mechanism demo that does not have to match a locked JPEG.
It is the wrong door if the bar has to be the serial on the cut sheet. Veo does not take your photograph as frame one. Description is not a GA. You will get a cousin rail. When the SKU must survive, product video from the bar still is the identity job. Veo remains the native-audio cut when the leftover is the hiss, not the badge. Do not glue those jobs into one prompt and hope Veo “uses the photo.” This catalog row is text-to-video.
Add a text overlay burns one authored line after picture lock — bar length, “training block only.” Do not ask Veo to typeset tooth pitch into the rail.
Do not generate a victim on the ice
A photoreal person in the hole is a casualty. A photoreal instructor walking a protocol is a health-shaped expert. You did not cast either. Synthetic presenters on health and finance are a YPP kill. YouTube's third inauthentic-content bucket is AI personas that present as human experts on health. Ice rescue sits next to that domain. Labelling the face AI does not leave the bucket.
Do not prompt a body in the water, even “anonymised.” Do not prompt a dummy “for scale.” Do not prompt a child or a hand reaching through the hole. If someone must speak a protocol, that someone is a real named credentialed person on a different take. Generation can still do the bar. It cannot be the consult, and it cannot be the victim.
FAQ
Can Veo 3.1 ingest our bar photo as frame one so the serial holds?
No. This row is text-to-video. An image is not required and is not the contract. When the bar must match a JPEG, use the product video generator on that still. Veo writes the diegetic cut from words: 4 / 6 / 8 seconds, 4K, 40 credits, native audio.
If I omit sound from the prompt, is the clip silent?
No. Native audio is on. A prompt that never mentions sound still returns a soundtrack the model chose. Write the kerf, the chips, or silence. Do not write a lecture and call it a stem.
Can I generate a victim in the hole if I blur the face or label the clip as training?
No. A body on the ice is a person you did not consent. Blur and a label do not create a grant. Photograph an empty training block. Prompt the bar. Leave the casualty off the frame.
Is a disclosed AI rescue instructor acceptable on the product channel?
Disclosure does not move a constructed health expert out of YouTube's third bucket. Keep the bar plate. Keep the overlay. If someone must speak a protocol, put a real named instructor on camera. Do not generate a victim so the mouth has a subject.