Pick Omni Flash to edit, not invent

Pick Omni Flash to edit, not to invent. Gemini Omni Flash Edit is conversational video editing: video in, instruction in, revised clip out. Category edit-video, 13 credits on the card, audio off. It is not a text-to-video model. If this week you do not have a plate yet, you are not on Flash Edit. You are on a generate row — Gemini Omni Video (T2V / I2V / R2V, 4–10 seconds, 4K listed, 9 credits) or Veo, Seedance, Kling.

Google’s Omni 1.1 Flash post (27 August 2026) is a different surface: gemini-omni-1.1-flash, 10-second prior context, extend in 10-second hops to 40 seconds, 360p drafts Google claims are faster and cheaper than Omni 1.1 720p, upscale to 1080p or 4K. Versely does not publish an Omni 1.1 slug. Do not route a brief to a slug we do not have. This radio is Flash Edit versus invent, on the rows that actually exist here.

Edit is a note on a plate. Invent is a new shot.

If the sentence starts with “change,” “replace,” “crop,” “warm,” “remove,” or “keep this and…,” you are on Edit — if the plate is worth keeping. The camera move you already liked should not go back into a lottery.

If the sentence starts with “a shot of,” “new,” “another take,” or “from a different angle,” you are on generate. Edit cannot recompose a shot from a camera that was never in the file. It cannot invent the other two setups in a three-shot scene.

Flash Edit on this catalog is a visual pass. Audio is flagged off. Do not use it to rewrite a spoken line. A new line is a voice pass, a dub, or a generate that was always a talking shot.

Omni Video is the invent door in the same family: prompt plus optional references, 4–10 seconds, 16:9 or 9:16, 4K listed. Use it when there is no plate. Do not use Flash Edit with a random clip attached “so we can be on Omni” when you meant a new shot. That is how you pay to revise a file you do not want.

Google’s 27 August 2026 note also describes first-and-last-frame interpolation, video reference up to 3 seconds, and access via AI Studio, Gemini Enterprise, Flow, and the Gemini app. Those are Google’s doors. If you are extending on Google’s 10-second hop, that is extend, not Flash Edit. Extend continues time. Edit revises pixels. Invent samples a new take. Three radios. This page is only: Flash Edit when the plate exists and the change is local.

This week’s jobs that belong on Flash Edit

One object, one attribute. Replace the mug. Change the jacket. Swap the box for the branded pack. Name the thing. Say everything else unchanged.

A restyle that should keep the blocking. “Golden hour, same framing.” The motion is why it looks like a grade instead of a different video.

A crop, a punch-in, a cleanup on a performance you will not re-roll. Generating a tighter take is how you lose the performance.

Chain those. Edit two sees the output of edit one. “Make the jacket red” then “now make it evening” is one plate with two notes. Two generates with those prompts is two coin flips.

Stop editing and cut on a timeline when you need a hard cut, titles, a second clip, or precise audio sync across a sequence. Conversational edit is for narrowing one plate. Past a handful of turns on a long clip, the conventional timeline is cheaper than another edit call on the whole duration — Flash Edit meters the source.

Do not Flash-Edit a SKU that never held. If the label is already wrong, you are editing a miss. Go back to a locked still and a generate (or I2V). Edit is not a SKU locker.

Do not Flash-Edit to add a sentence of dialogue. Audio is off. Overlay the line, or generate a talking take on Veo.

This week’s jobs that are invent

No plate. You have a sentence. Omni Video, Veo, Seedance, Kling — pick the generate that matches dialogue, SKU, or silent motion. Not Edit.

The blocking is wrong. New angle, new room, different performance. Generate. If the last frame is good and you want more of the same move, that is extend, which is also not Edit.

Three camera positions. Three generates (or three stills animated), then a cut. Three edit turns will not invent the other two cameras.

A hero you meant to failover because Edit felt cheaper than a new Veo. If you need a new Veo, pay for a new Veo. A Flash Edit of a Kling draft will not become a Veo talking take.

Sora-shaped leftovers: do not “Omni Flash” a Sora clip into a new idea and call it a migration. Migration is a new generate on a living row, from a still if identity matters. Edit keeps Sora’s last mistakes too.

The Monday radio

For each request in the chat, ask: do we already have a plate we would be sad to lose?
Yes, and the change is local and visual → Gemini Omni Flash Edit. One note per turn. Cap the turns.
No plate, or the idea changed → Omni Video or another T2V/I2V row. Set aspect first.
Need more time on a keeper → extend row, not Flash Edit, not a new invent.
Need a new line of speech → not Flash Edit.

The AI video generator sits next to edit in the product. Adjacent is not interchangeable. Inventing on Edit is how you attach a dummy clip. Editing on a generate row is how you throw away a signed plate.

When the plate is signed and the note is a crop or a jacket, sign up and pick Flash Edit. When there is no plate, pick a generate. Do not ask Omni Flash to invent a shot it was not given.

FAQ

Is Omni Flash Edit the same as Omni 1.1 Flash scene extension?

No. Google’s 27 August 2026 Omni 1.1 Flash post is scene extension and draft/upscale on Google’s surfaces, model id gemini-omni-1.1-flash. Versely’s Flash Edit row is gemini-omni-flash-edit, video-to-video conversational editing, audio off, 13 credits. Do not treat extend hops to 40 seconds as this slug.

Both Omni Video and Flash Edit are “Omni.” Can I use either?

Omni Video invents (and can take references) at 4–10 seconds. Flash Edit revises a source clip. If you only remember “Omni,” you will pick the wrong one. Name Edit versus Video on the slate.

Will Flash Edit keep the native audio on a Veo plate?

Treat the Versely Edit endpoint as visual. The card flags audio off. Do not plan on a rewritten line or a repaired bed. If the VO has to change, do that as a separate voice or dub step after the picture is right.