METHODOLOGY AND DATA HANDLING

A useful first pass, with its limits visible.

AdScanVideo turns a video into structured observations. It analyzes video and available audio to produce timestamped observations. The report is designed for creative research, production planning, notes, and review. It is not a forensic conclusion or a substitute for expert verification.

What happens during an analysis

  1. The input is validatedUploads must use a supported video format and stay under 200 MB. Free and single-credit videos are limited to 10 minutes; Monthly Premium videos can be up to 45 minutes.
  2. A temporary workspace is createdThe server processes the video in an isolated temporary directory. The original filename is not used as a server path.
  3. Video and audio are preparedThe native analyzer receives video and audio together. Speech and visual descriptions can be inaccurate; the model is instructed to avoid inventing details and to mark uncertainty.
  4. The video is analyzedThe analysis provider processes the video at its configured media resolution. This can miss brief actions and fine details. An explicitly configured fallback can use selected frames and available transcripts.
  5. The chosen mode shapes the reportFind moments, Dashcam timeline, Shot breakdown, Ad intelligence, Podcast notes, Marketing copy, Tutorial steps, and Summary use instructions tailored to their outcome.
  6. Temporary media is deletedThe uploaded video, extracted audio, captions, and frames are removed when the job finishes, fails, or times out. Deletion of the provider upload is requested after processing; provider retention is governed by its applicable terms. Media is not kept as a customer video library.

What is retained

The service retains the generated analysis text and limited operational metadata needed to deliver results, enforce the daily allowance, diagnose reliability, and understand processing cost. Operational records can include job status, timestamps, mode, duration, selected-frame count, transcript availability, and cost. Analytics events exclude filenames, video URLs, report text, contact details, and job identifiers.

What the report can get wrong

  • Speaker names are unknown unless the video supplies them.
  • Speech recognition can miss words, accents, names, or technical terms.
  • Camera movement and intent can be misinterpreted from visible evidence.
  • Fast cuts can compress several visual changes into one time range.
  • Legal, medical, financial, and factual claims require independent verification.

How to review a result

Use timestamps to return to the original footage. Verify quotations against the audio, confirm names and numbers, and treat creative recommendations as a starting point. For production work, add the lens, lighting, safety, ownership, and scheduling decisions that pixels alone cannot reveal.