Save a question once, run it against authorized private video or audio, and give your product timestamped results with stable scanner-version and source metadata. Video find_moments can feed explicit rendering without publishing media automatically.
These patterns use video or audio you are authorized to process. ChatClipThat does not install a recorder, discover third-party sessions automatically, or replace consent and retention controls.
Find repeated actions, visible errors, stalled steps, and abandoned flows across screen recordings. Preserve grounded timestamps for the product team instead of reducing each session to an unsupported conclusion.
find_momentsclassifierUse all six scanner methods for timestamped objections, needs, workarounds, and decisions. Use video find_moments when the result may become a reviewed clip.
extractorfind_momentsLocate confusing objectives, repeated failures, reactions, bugs, and high-signal moments in uploaded playtests or authorized gameplay VODs.
monitorfind_momentsTurn walkthrough video or support-call audio into timestamped findings that can be routed into review and follow-up workflows.
classifierextractorApply the same versioned question to an authorized video or audio catalog, attach metadata, and query compatible observations across completed runs.
summarizerscorerRank timestamped observations by relevance and let a human approve useful findings. Render approved video find_moments results by their Moment IDs.
observationsrendersA scanner is a saved, versioned question. A run snapshots that version against one ready private video or audio upload. An observation carries a typed value, relevance, and grounded timestamps.
Each recipe names the API resources, retry boundary, provenance fields, and current availability assumptions it depends on.
Submit up to 100 durable items, preserve provenance, and respect shared capacity.
ReviewBuild an evidence review queueNormalize typed observations without flattening relevance into confidence.
MediaRender only approved momentsKeep structured analysis separate from optional MP4 generation.
EventsHandle completion reliablyPoll today and adopt signed, idempotent webhook handling where enabled.
API usage is metered from your ChatClipThat account wallet. A paid account is required to issue API credentials, and the console shows only the checkout choices currently available to your account.
Usage rounds up to the next minute. A video find_moments window bills only its selected duration; without a window, the full upload is measured and billed. Each scanner run is billed independently.
Run three saved scanners against that same upload and the total is 30 credits. Reusing an upload avoids another transfer, not another analysis charge.
Polling and structured JSON do not add charges. Video find_moments results can be rendered explicitly; scanner completion itself never publishes media, and audio has no render path.
Current access and limits. Paid accounts receive one proof analysis for one window up to 2 hours and one proof render totaling up to 60 seconds. Paid analyses have no fixed duration ceiling: each request must name an explicit window, fit the available credit balance, and stay within the account source-size limit. Longer windows use the same bounded chunk pipeline automatically. Video is capped at 8 GiB for the proof tier and 10 GiB for paid tiers; audio is capped at 2 GiB. Beginner, Hobby, and Creator allow 1, 2, and 3 concurrent runs with source retention of 7, 30, and 90 days. Enterprise limits are contracted. ChatClipThat system failures are refunded automatically.
See how saved scanners, typed observations, provenance, and optional rendering form a reusable application layer—and where a dedicated session replay product still fits.