Push-in

Move the viewpoint closer to a subject to concentrate attention on a detail or reaction. Unlike a lens zoom, a physical approach changes the spatial relationship between foreground and background. Russian catalog label: наезд камеры.

Prepare the direction

Choose the starting composition, the action and the intended end state. Put the direction in the scene brief or the request for Prompt for the model (Промпт под модель). Select the video model separately and review the final instructions before rendering.

Original example:

Begin with the watchmaker and the whole workbench visible. Approach slowly until the watchmaker’s hands and the open watch fill the image.

Model applicability

Availability: supported as a natural-language authoring instruction for the current video families. It is not a separate hardware camera control or a promise that a model will comply. Choose a mode that can accept the scene’s required inputs; first-frame constraints can conflict with an incompatible starting view.

Outcome: no dedicated controlled success claim for this exact technique/model pair is established by this guide. Model pages report their specific observations. Start with a short test and keep the model, quality and mode in your notes. Seedance 2.0 is one documented starting point; the model guide explains the other families.

Inspect the result

Check that the approach is continuous, the subject remains identifiable, and the background changes naturally. A cut to a close-up or a simple enlargement is a different result.

If the result misses the technique, revise one conflicting instruction or input at a time. Missing evidence means untested, not unsupported. See also pull back.

See the model–technique applicability matrix for connected operations, dated outcomes and configurations that remain untested.

Reviewed 14 October 2026 against the public technique catalog and Studio authoring contract. This example has not been rendered for this documentation.