12-second example · sel_area
Remove a title and headline overlay from city street footage
A 12-second downtown street clip with a large white headline and a yellow episode line over a building facade. Watch the input and the real sel_area API output with the exact region and charge.
sel_area- Clip length
- 12 s
- Source size
- 1280 × 720
- Recorded API charge
- $0.05625
- Processing
- 133.7 s
Measured 2026-10-20. USD API charge for 15 billable seconds; web points are separate. 174.8 seconds from submission to completion, including validation and queue time. This is a recorded run, not a latency guarantee.
When to use this workflow
Reuse authorized B-roll without an old episode title or headline card.
Why this removal mode?
One 758 × 164 pixel box covers both headline lines with about 16 px of margin, about 13.5% of the frame. The text is static, so a single selected region is enough.
Result review
Both headline lines are no longer readable in the reviewed frames, and the facade, street lamp and windows continue through the region. Lettering of the hotel sign that was partly covered looks slightly distorted compared with an unobstructed sign.
What this example cannot prove
The hidden area is a building facade with a hotel sign behind the headline. Reconstructed sign lettering and window edges can look slightly warped; this clip does not establish results on faces or fast camera motion.
No model version was returned by the public job response. One short clip cannot predict results for other footage. Check the full result at normal speed and around each word change.
Reproduce this API request
Upload your own authorized video first, then replace the upload ID. Coordinates refer to this example’s 1280 × 720 source pixels; recalculate them for a different video.
{
"type": "remove_text",
"input": {
"upload_id": "upl_your_uploaded_video"
},
"mode": "sel_area",
"rect": {
"x1": 262,
"y1": 104,
"x2": 1020,
"y2": 268
}
}Send Authorization: Bearer $UNMARKAI_API_KEY from your server and a stable Idempotency-Key for this request. Never expose the key in browser code.
Source and demonstration method
Text overlay added for this test with ffmpeg drawtext (DejaVu Sans Bold, white 76 px headline fading in at 0.4 s, yellow 34 px episode line fading in at 0.9 s) on licensed footage trimmed to 12 s and scaled to 1280 × 720 at 25 fps. The after video is the actual API result, without manual cleanup.
Footage by Get Lost Mike on Pexels, used under the Pexels License. Prepared at 1280 × 720, H.264, 25 fps, with no audio. The comparison places the processed input and downloaded API output on the same timeline; only comparison labels and encoding are added.
Use owned or licensed footage for caption replacement. Keep attribution and required disclosures in content you publish.
Explore the next workflow
All video text removal examples · Video text remover · AI caption remover