Harmonic shears jaws bind Gemini Omni Video at 9cr; quota is seven

Gemini Omni Video is 9cr, 4/6/8/10s at 4K, quota images + videos*2 + character_ids ≤ 7: reference the photographed Harmonic shears; do not claim native audio on this row, and do not generate a surgical field or talking surgeon.

A Harmonic-shears listing a buyer can match is this pair of jaws: blade length, clamp face, handpiece colour, the cable boot. Those pixels live in a photograph. Omni restages. A paragraph about an energy device in a bright theatre will mint a cousin blade, a drape, and a person you did not hire. Photograph the shears. Attach that still. Spend the seven-slot budget on the instrument, not on a field.

Photograph the jaws; the field is not the SKU

The honest plate is the instrument you ship. Square-on of the closed jaws. A three-quarter that still reads the clamp. Empty of people, empty of tissue, empty of a tray you do not sell with this SKU. Crop a handwritten QC tag you would not put in a listing. Do not prompt the model to “clean” a bench. Erasing clutter redraws the plastic behind it.

Text-to-video of “ultrasonic shears sealing a vessel in a modern OR” invents a cousin jaw and a body. That is a different catalog row, and it is a case you did not consent. Keep the still as a line card: even light, logo readable, jaws in focus. If the only photo includes a gloved hand, crop the hand or shoot again. A generated glove is still a generated body.

The AI product video generator is the hardware contract: reference photos are the ground truth. Zoom to 100% and read the jaw profile before you keep a take. Prompt only camera on the unit. Re-describing “Harmonic shears ready for surgery” is how the model draws a second chassis.

If frame one has to be the JPEG you approved, AI image-to-video treats the upload as frame one. Omni does not. Mixing those jobs is how a catalog still becomes a theatre you did not light.

Gemini Omni Video is 9 credits and a quota of seven

Gemini Omni Video is Google's multimodal video row: a prompt plus optional reference images, source video clips, character IDs, and audio IDs. Catalog output is 720p, 1080p, or 4K, at 4 / 6 / 8 / 10 seconds, 16:9 or 9:16. The listed cost is 9 credits. Categories are text-to-video, image-to-video, and reference-to-video. An image is not required to press generate. That is the tell. A theatre Omni film with no shears photo in the stack is a mood. Procurement cannot bid from a mood.

Gemini Omni Video: the stack is the brief is the row's own contract. This leftover only spends it on the photographed jaws. The quota is arithmetic: images + videos×2 + character_ids ≤ 7. One approved instrument still is a stack. Seven moods of “the vibe of surgery” is how you blow the quota and still miss the clamp.

A source clip counts double. Character IDs invite a constructed clinician. Do not spend a slot on a surgeon ID. One face of the jaws, maybe a second angle of the same serial, is a kit. Prove the stack at 4s. 4K is the ceiling, not the personality: a bad stack at 4K is a sharper wrong blade.

Do not claim native audio; do not letter the blade length

The catalog does not mark native audio on this row. audio is empty. Treat soundtrack as an input you may pass (audio IDs), not as a guaranteed mix. Quoted dialogue does not become a stem. If the clip needs a bed, score after picture lock, off a plate with no mouth in frame.

Blade length, generator compatibility, “for demonstration only” — those strings are claims. Glyphs in a generate have to be right every frame. They will not be. Add a text overlay burns one authored line for the whole clip — top, center, or bottom. It does not transcribe. Write the string as it sits on the cut sheet. Do not prompt a surgeon demonstrating on a dummy, and do not run auto-captions hoping a transcriber invents a part number nobody said.

No surgical field, no talking surgeon

A photoreal person next to energy shears is a clinician, and a photoreal field is a case. You did not cast either. Synthetic presenters on health and finance are a YPP kill: AI personas as human experts on health. A constructed surgeon walking a sealing protocol is that shape. Labelling the face AI does not leave the bucket.

Do not prompt a draped field. Do not feed Omni a theatre tour and hope the jaws in the corner are yours. Do not spend character_id slots on a talking applications person. If someone must speak a protocol, that someone is a real named credentialed person on a different take. Generation can still do the jaws. It cannot be the consult, and it cannot be the field. Attach the photograph. Count to seven. Leave the theatre off the stack.

FAQ

Can I skip the shears photo and prompt Omni to “match our Harmonic jaws”?

No. A sentence is not a plate. An image is not required to press generate on this row, which is how teams skip the still. If the jaws have to be yours, attach the photograph. Quota and 4K do not pin a clamp you never uploaded.

Does Gemini Omni Video add an OR soundtrack or a surgeon VO?

Do not claim native audio. Catalog audio is empty. An audio ID is an optional input, not a mix you get for free. Score after lock. Do not prompt a lecture and call it a stem.

Why not fill all seven reference slots with theatre stills?

Because videos count double and character IDs invite a constructed clinician. Images + videos×2 + character_ids must stay ≤ 7. One approved instrument still is the stack. Seven moods of surgery still miss the SKU if the jaws photo is not in it.

Is a disclosed AI surgeon acceptable on the product channel?

Disclosure does not move a constructed health expert out of YouTube's third bucket. Keep the instrument plate. Keep the overlay. Do not generate a surgical field or a talking surgeon so the mouth has a subject.