Product Swap, character swap or lip sync: which workflow fits?

The short answer

Choose Product Swap when the visible item is what needs to change. Choose Recreate when you want a different main character to follow a reference performance. Choose Video Lip Sync when an existing human-speaker video needs new speech. Choose AI Talking Photo when the starting point is a still portrait. These tasks use different inputs and should not be described as interchangeable. Product Swap currently has limited approved-account access; the other tools have their own published input rules and quotes.

Replacing the item: Product Swap

A product swap starts with a video plus a replacement product photo. The target is one visible item in one continuous shot. You identify the original product in a frame rather than replacing the presenter. Similar shape, viewing angle and handling make the shot easier to assess.

UseLipSync's current beta accepts 4–15 second source videos and prepares a 4–8 second selection, up to 50MB. The frame opens automatically; draw a box to select the item. Supported whole-second output follows the selection, so you do not have to choose a second duration. Fine packaging text and unchanged surrounding pixels are not guaranteed.

Replacing the main character: Recreate

Recreate uses a character image and reference performance to generate a new version of the action, camera and timing. It is the relevant starting point when the person or character should change. It is not a precise product-editing mask, nor a tool for independently replacing everyone in a crowded scene.

A request to put a new presenter into a product demonstration is a character-replacement task. A request to keep the presenter while showing a different bottle is a product-replacement task. Review the relevant tool's own upload limits and quote rather than assuming the same setting applies to both.

Changing the words: Video Lip Sync

Video Lip Sync starts with an existing supported human-speaker video and new speech from text or an uploaded audio track. The visual performance remains the basis for the new spoken delivery. It is the relevant starting point when the script, language or voiceover is the main change.

Swapping a product visually does not update its name in the soundtrack. If a presenter names the old item, resolve that mismatch before publishing. Do not assume product replacement, new dialogue and subtitle rewriting happen in one operation.

Starting from a still portrait: AI Talking Photo

AI Talking Photo turns a portrait into a speaking video. It answers a different need from editing an existing product shot: there is no original filmed performance to preserve. Use a clear, authorized portrait and follow that tool's speech and image requirements.

A product photograph alone is not a source video for Product Swap. If you need a wholly new advertisement from a packshot, this limited replacement workflow is not a promise of a complete ad-generation pipeline.

Treat sound, claims and permissions as separate checks

Product Swap offers Keep original sound and No sound. Keeping sound is a generative request, so speech, music and timing can vary. Muting removes sound; it does not rewrite an old claim or make the visual result an endorsement. Check the final content against the real product and intended message.

You need permission for the source, images and sound you use. A successful render does not establish advertising rights or product performance. This comparison makes no claim that a model result is exact, that all accounts have beta access, or that existing Recreate examples prove product replacement.

Open the workflow that fits

• Product Swap availability and approved-account workspace
• Recreate: replace a character
• Video Lip Sync: change spoken words
• AI Talking Photo: start from a portrait