Police reports are written after the fact, sometimes hours or days after an incident. They rely on memory, which is fallible. They are shaped by training, institutional pressure, and the human tendency to construct coherent narratives from chaotic events. None of this is necessarily malicious—but the result is that police reports frequently contain inaccuracies when compared against body camera footage.
For defense attorneys, finding those inaccuracies is the difference between winning and losing a case. But manually cross-referencing a multi-page narrative report against 45 minutes of body camera footage is painstaking, time-consuming work. Most defenders simply do not have the hours.
FrameCounsel's contradiction detection engine automates this process.
The contradiction detection pipeline operates in three stages:
FrameCounsel accepts police reports in PDF, DOCX, and plain text formats. The system parses the document into structured claims—discrete factual assertions extracted from the narrative. For example, a sentence like "The subject was non-compliant and refused to exit the vehicle" becomes two claims: (1) the subject was non-compliant, and (2) the subject refused to exit the vehicle.
Each claim is matched against the video transcript and visual timeline. FrameCounsel identifies the time window where the described event should have occurred, then examines both the audio transcript and visual context for that period.
The system uses semantic similarity matching rather than simple keyword search. It understands that "refused to exit the vehicle" should correspond to a segment where the officer made a request and the subject's response is captured on audio.
When a discrepancy is found, FrameCounsel classifies it into one of several categories:
Each contradiction receives a confidence score and is linked to the specific timestamp in the video where the discrepancy occurs.
In Brady v. Maryland (1963), the Supreme Court held that prosecutors must disclose evidence favorable to the defense. Contradictions between police reports and body camera footage are exactly the kind of material evidence that Brady requires.
FrameCounsel does not just find contradictions—it generates court-ready documentation that maps each discrepancy to specific video timestamps, report page numbers, and applicable legal standards. Defense attorneys can export a contradiction report that is ready to attach to a Brady motion or use in cross-examination.
In internal testing with partner defense attorneys, FrameCounsel's contradiction detection engine identified an average of 4.7 material discrepancies per case that the attorneys had not previously caught through manual review. In one case involving a contested traffic stop, the engine found 11 contradictions between the arrest report and body camera footage—three of which were classified as factual contradictions serious enough to warrant a suppression motion.
The tool does not replace attorney judgment. It amplifies it. By surfacing every potential discrepancy and providing the video evidence to support it, FrameCounsel lets defenders focus their limited time on the contradictions that matter most.
Segment Anything Model 3 lets defense teams isolate objects, people, and details in footage with surgical precision — all on-device.
An analysis of the growing disparity between prosecution and defense access to forensic video analysis technology, and what the legal community must do about it.
A comprehensive guide to the best AI tools available for public defenders in 2026. From body camera analysis to case management, these tools help public defender offices handle crushing caseloads more effectively.
On-device body camera analysis, contradiction detection, and court-ready reports. No credit card required.