Skip to content

Creation guide

AI Face Swap vs Video Recast: Which Fits Your Clip?

Face swap usually targets the face region in an existing clip. Video Recast uses reference people to regenerate a source-driven performance. They are related creative tasks, but they make different promises about bodies, clothing, occlusion and sound. Choose the method by what you want to change and what must remain recognizable.

Updated 2026-10-27

Open the creator
Existing ballet recast example; not a guarantee of frame-by-frame identity.

Start with what should change in the clip

If you only want to replace a face while keeping the original performer’s body, a face-region editing tool may be closer to the task. If you want a character from a reference image to take part in an existing performance, character recasting is the more relevant workflow. Videopogo should be evaluated as a reference-driven generation service, not as a promise to lock every original pixel, garment and body movement while changing only the face.

Understand the inputs and identity limits

Video Recast starts from an existing clip and separate character reference images. General Recast has its own source-video duration and file limits. A fixed template with a longer source belongs on its dedicated page; a rounded duration in a gallery is not permission to submit a file whose actual duration exceeds a model limit. Explicitly choose each reference before creating. Receiving or adding a photo is not the same action as submitting it to a model.

Turns, occlusion, overlapping people, fast gestures and small faces make a result harder to assess. A model can change facial detail, clothing, body shape or background while interpreting the performance. No sample establishes perfect identity in every frame. Begin with readable people and moderate movement, inspect the whole output, then decide whether this type of generation fits the intended use.

Separate original audio from newly generated sound

The current Normal Recast workflow is intended to preserve the source performance and soundtrack. Other Hotel music styles use a different reference-driven generation workflow and create new audio. They are not a promise to sing an existing commercial track word for word. Template lengths, frame shapes and sound contracts are specific to each page, so do not infer one template’s behavior from another template’s preview.

Permission to use a portrait is different from permission to use the source performance, music and final distribution. Check each separately, particularly before commercial use. Do not use a recast clip to mislead people into believing that the depicted person genuinely made a statement or performed an action. An output’s availability for download does not by itself establish every right required for your project.

Use a fixed template when you already like the performance

A ready-made scene reduces the number of prompt and settings decisions. Hotel Lobby, Zombie Love Story and Raindance use two character images, while Bulin 47 and Burning Bridges each use one character image. Read the selected page before preparing inputs; a workflow that makes sense for a duet is not automatically the right upload pattern for a single-character scene.

Watch the original example, check person positions, duration, shape and price, then prepare your photos. Template packs and account credits have different scopes. Verify what the selected purchase covers rather than assuming every balance can fund every template. A payment return or restored draft does not begin generation; review the inputs and use the explicit create action.

Choose photo animation when you do not need a source performance

If there is no existing choreography or camera move to preserve, ordinary image to video may be simpler. It starts from the still image’s composition and a description of new motion. It does not promise to reconstruct the timing of an unrelated clip. Choosing the task first helps you avoid changing models repeatedly when the real issue is a mismatch between the desired outcome and the mode you selected.

Check the delivery before making another attempt

Inspect the complete face, body, occlusion, motion and soundtrack. A technical failure is different from a successful result you would like to change. Another creative version after success is a new paid generation. If a submission or archive remains uncertain, inspect the existing task first rather than paying again to diagnose the situation. Download the result you want to keep, and retain your own original inputs and permissions.

Related template tutorials