VEO First Last Frame: the still is the contract (4s, 6s, 8s, 4K, 40cr)
VEO First Last Frame is image-to-video. If the label is wrong, every second is waste.
VEO First Last Frame is image-to-video. If the label is wrong, every second is waste. The catalog row VEO First Last Frame interpolates between first and last frames — that is the whole description. It does not invent a product from a paragraph. It does not extend a keeper. It spends 40 credits to travel from a still you already approved to a still you already approved.
Google files this next to the rest of the VEO line. Native audio is on. Durations are only 4s, 6s, and 8s. Output quality includes 720p, 1080p, and 4K. Aspect ratios are 16:9, 9:16, and auto. A 15-second talking scene from text is the wrong seconds.
Both ends have to exist before you press generate
First-last frame is a two-still contract. Frame one is the opening photograph. Frame two is the landing photograph. The model solves the motion between them. The image-to-video tool is the right door when you have those stills. It is the wrong door when you have a sentence and a hope.
That is why the still is the spec. A bottle that must stay the bottle needs a front pack and a three-quarter pack, same lighting family, same lens height. A room that must go from empty table to plated dish needs both photographs. If one end is a sketch in the prompt, you do not have a first-last job yet. Make the stills. Then buy the interpolate.
Veo 3.1 is the sibling when the job is a scene from text with native audio. Same family, different input shape. Do not “just try first-last” on a prompt that never named a last frame. The 40-credit sticker is not a cheaper Veo 3.1. It is a different machine.
4s, 6s, 8s — there is no ninth second on this row
The duration menu is not a suggestion. 4s, 6s, 8s. That is the entire list. A product orbit that needs 12 seconds is not this model with the slider imagined further to the right. An eight-second interpolate of a 180° turn is a different film from a four-second lid-seat. Pick the length that matches the distance between the two stills.
Styles on the row are Cinematic, Documentary, and Abstract. Motion levels are Medium and High. Frame rates are 24fps and 30fps. None of those replace a missing last frame. They flavour a travel you already bounded. If the two stills are not connectable in eight seconds — macro of a cap to a wide of an aisle — the model will warp, not turn. Shorten the geographic ask, or pick a different job.
4K is the ceiling, not a default you owe every take. 720p and 1080p are on the quality list. Spend 4K when the plate is a hero. The sticker is 40 credits either way. The waste is the wrong job, not the badge.
Native audio rides along. Write it or refuse it.
Audio is listed as on for this row. A silent prompt still gets a soundtrack you did not choose. If the interpolate is a pack turn you will score later, say so in the prompt — no music, no voice, room tone only — or plan to replace the bed in the editor. Do not discover a documentary swell under a SKU you meant to keep mute.
The complementary mistake is treating this as a talking-head row. First-last will interpolate a face from pose A to pose B. It will not lip-sync a script. If the mouth is the shot, you are in a lipsync or talking generate, not in a two-still travel.
For a product that has to remain the SKU, the longer argument is first-last as the product-turn model. This page is only the VEO endpoint: image required, 4 / 6 / 8 seconds, 4K available, 40 credits, native audio, Google. The best image-to-video list is the rest of the category. Use this row when both photographs exist and eight seconds is enough travel.
FAQ
Can I run VEO First Last Frame from a single still?
The catalog requires an image and describes the job as generating between first and last frames. One still is half a contract. Use a plain image-to-video row if you only own the start. Use this row when you own both ends.
Is 4K automatic at 40 credits?
No. The quality list is 720p, 1080p, and 4K. The catalog sticker is 40 credits. Pick the tier you will actually ship. 4K does not rescue two stills that disagree about the label.
Why not always take 8s?
Because 8s is the ceiling, not a free upgrade. A lid that seats in two seconds looks padded at eight. A 4s take of a short mechanical move is the honest length. The waste is extra seconds of a morph you already finished.
How is this different from Veo 3.1 text-to-video?
Veo 3.1 is a text-to-video scene clip. This row is image-to-video between two frames. Same provider, same short-clip family, different input. If you do not have two stills, you are not on this row yet.