Once a second pedestrian walked close to the camera around 0:16–0:17, the matting broke down and large chunks of the original street scene bled through around her instead of being erased.
What was measured
Motion handling
Does quality degrade when the subject moves quickly, gestures, or leans?
decisive for this rankingtransformation
A video-background tool has to keep working when the subject moves; failure here means the core effect breaks down. (3 of 3 judges)
What was given, what came back
Test input: Busy Urban Street · video · group: remove-or-replace-video-backgrounds-using-ai
Input — what we sent
Busy Urban Street
A video of a person walking through a busy city street with multiple people, moving vehicles, and a visually complex urban background. It was used to stress segmentation and tracking under heavy background distraction and continuous motion.
Why this input is hard
- · Complex background segmentation
- · Handling background distractions
- · Subject tracking
- · Temporal consistency across frames
- · Edge accuracy during continuous movement
Output — unretouched
Also checked on this input — same tool, 1 other criterion
Provenance
- Observation
- a42137aa-e7d2-47d6-88b4-8d6475beb11c
- Evidence run
- 10155228-7ce8-438a-a210-547331ef080b
- Study
- Remove or Replace Video Backgrounds Using AI
- Research task
- 86ba42c2d
- Tested at
- not recorded
- Source
- first-party
- Evidence state
- verified
- Proof shown
- input + output shown
- Cost / latency
- not captured
- Repeat run
- not captured
- Tester
- not captured
The last three rows are honest blanks, not placeholders — our capture has no field for them yet.
Query this
get_evidence({
tool: "kapwing",
scenario: "remove-or-replace-video-backgrounds-using-ai"
})MCP · mcp.aidemos.com/api/mcp
Free with attribution.
Same input, same check — 7 other tools
measured on Motion handling
Bria.ai✓ WorkedBoth pedestrians remained separately isolated while walking, so multi-subject motion did not cause subject loss.Cutout.Pro✓ WorkedKeeps a moving foreground pedestrian isolated while the background stays removed, and even a second person can appear as a separate cutout in one frame.Descript⚠ StruggledThe requested in-motion subjects are rendered more static and posed than the source’s forward-stride feel, so motion energy is flattened.Fotor✓ WorkedThe tool tracked 3–4 background pedestrians at once through a full 20-second continuous-motion clip without merging or dropping anyone.InVideo AI✓ WorkedIt kept the subject and secondary pedestrians consistent through a full ~20-second continuous walking shot, with no ghosting or motion breakdown.Media.io✓ WorkedThe cutout held up as several people shifted position through the frame, including one person moving from center toward the right and others entering at the edges.Picsart✗ FailedIn a two-person moving street clip, the primary walker stayed isolated but the second pedestrian was erased entirely, indicating a single-subject lock-on in dynamic scenes.
This evidence is published in
Real inputs and real outputs, no retouching · every cell queryable via API & MCP · aidemos.com