Many AI-video exports arrive at 1080×1920, the standard portrait resolution for a 9:16 vertical video. That frame often contains carefully composed subjects near the edges, so even a small crop can cut off hands, text, faces or background detail. A better workflow keeps the original canvas and limits processing to the small region that needs cleanup.
Start by confirming the real source resolution
Do not rely only on what a social app says. Check the actual file before processing. A video may look vertical while being encoded at 720×1280, 1080×1920 or another size. A watermark profile calibrated for one resolution can miss the target area on another.
- 1080×1920 = full-HD portrait and the strongest current fit for the supported recent Gemini story profile.
- 720×1280 = smaller portrait and may require a different profile.
- 1920×1080 = landscape; do not force it into a portrait workflow.
Why cropping is a weak first choice
Cropping works only when the visible mark sits outside the important composition. On a portrait video, a corner crop can change the aspect ratio, make the subject feel off-center and force a second resize before posting. It may also reduce image quality when the edited frame is stretched back to 1080×1920.
A local region-based cleanup keeps the frame boundaries intact. That means the final video remains suitable for Instagram Reels, YouTube Shorts or other 9:16 placements without adding black bars or zooming the image.
Use the correct visible-watermark profile
Recent AI exports and older Veo exports can use different visible marks. A cleanup tool should not assume that every file has the same logo or position. On geminiwatermark.space, recent portrait videos use the new diamond profile, while older supported Veo clips can use the legacy profile.
- Open the remover and choose Video.
- Select the profile that matches the source.
- Upload the original file, not a version already compressed by a messaging app.
- Preview the cleaned result before downloading.
Preserve detail around the cleaned area
The hardest frames are not plain skies or flat walls; they are frames where the mark overlaps clothing, hair, leaves, shadows or a moving object. In those cases, a hard rectangular blur is easy to notice. Better cleanup uses nearby visual information and a soft transition around the edited region so the result does not look like a pasted block.
Even with a good local process, review several moments in the clip rather than checking only the first frame. A background can move underneath a fixed mark, which means one timestamp may look perfect while another reveals a texture mismatch.
Check the final export before publishing
After processing, confirm four things: resolution, duration, audio and the cleaned corner. If the browser exports WebM, keep the file when your publishing platform supports it; otherwise convert a copy to MP4 with a quality-focused encoder rather than repeatedly re-exporting through social apps.
Use the browser remover for supported visible Gemini and Veo profiles. The original 1080×1920 frame is kept during the targeted story workflow.
Open video removerWhat this workflow does not do
This process is for supported visible watermark overlays. It does not remove SynthID or other invisible provenance signals. It is also not a general-purpose object remover for arbitrary logos across the entire frame.
Only edit media you own or have permission to modify. Watermark designs and placement can change over time, so unsupported exports may need a newer profile rather than more aggressive processing.