Sam 3 Video Segment: pick the row, not the brand (1cr)

Sam 3 Video Segment is one catalog generate. Use it because this job matches this row.

Sam 3 Video Segment is intelligent video segmentation for object detection and precise video editing. Content type: video. Category: edit-video. Credits: 1. Audio empty. No duration list. No max output resolution. Features: video_editing, segmentation, object_detection. That is a mask. It is not a movie. If you opened Sam because you needed a clip, the brand did not fail you. The row pick did.

A mask is not a generate

Segmentation finds objects across frames so an editor can isolate, replace, or protect them. The glossary term is video segmentation. Precise video editing in the description is the after: you segmented, now you cut, restyle, or composite. Sam 3 is the first half of that sentence.

requires_image is false and the duration list is empty because Sam 3 is not selling a 5s or 10s invent. It labels objects on footage you already paid to have. The catalog field is 1 credit; confirm the live quote against the clip's length. That is still a credits figure, not a dollar figure, and it assumes there is an object in frame worth detecting.

There is no Sam provider hub on this site. The model page is the roster. Prompting: Sam 3 Video Segment prompting. Downstream edits live under video editing — captions, crop, background removal — not under a generate row.

One credit for object detection

Quote the row and stop decorating:

• 1 credit (catalog field)
• edit-video
• segmentation / object detection
• no audio
• no listed resolution
• no listed durations

If the brief is "make me a 10-second 1080p street scene," this is the wrong aisle. If the brief is "cut the product out of the 10-second street scene we already shot," this is the aisle.

People click cheap video rows when the budget is tight. Segmentation is cheap because it is not inventing pixels of a world. It is labeling pixels of a world you already paid for. Using it as a generator is how you get an empty mask of a scene that does not exist.

Precise editing starts after the generate

Order of operations:

1. Have a clip (shot or generated on a generate row).
2. Run Sam 3 Video Segment if you need the object isolated.
3. Edit: restyle, replace, protect, caption.

Skipping step 1 is the brand mistake. Skipping step 3 is the other one — a segment file sitting in a folder is not a shippable. Remove video background is a related task when the object is the subject and the background is the problem. Sam is the detection/segmentation model in the catalog for the precise version of that idea. Read the slug: sam-3-video-segment.

Pick the row because you need a segment

"We use Sam" is not a pipeline. "We need object-accurate masks on this clip, 1 credit, edit-video" is a pipeline. If you needed motion from a still, that is image-to-video. If you needed a silent 1080p invent, that is a text-to-video row with a duration list. If you needed lipsync, that is a lipsync row. Sam will not stand in for those. It will segment.

FAQ

Is Sam 3 Video Segment a text-to-video model?

No. Category is edit-video. Description is segmentation for object detection and precise video editing. 1 credit, no durations, no resolution listed, no audio.

Why is there no 5s / 10s / 1080p on the spec?

Because this row does not generate a clip at a chosen length or ceiling. It segments a source. Length and resolution come in with the file.

Can I use it to invent a character?

No. Object detection labels what is already in the frames. Inventing a character is a stills or video generate. Segment after the character exists.

What do I do with the output?

Edit. Isolate, protect, replace, composite. The catalog's "precise video editing" is the point of the mask, not a second generate hiding inside the slug.