TaughtHuman decisions for better models

Audio and video labeling with time kept intact.

Review a whole recording or clip, mark the precise interval where an event occurs, and keep every human decision linked to the source media and task contract.

Request early access How contributor points work
Evidence registerTemporal evidence needs more than a play button.

Taught records the exact time range, source identity, task revision, and reviewer action as one reviewable object.

  1. Listen / watch
    Source in context

    Reviewers can inspect the media before and after the event.

  2. Locate
    Exact interval

    Boundaries use integer microseconds rather than drifting floating-point timestamps.

  3. Decide
    Human disposition

    Whole-asset and time-segment tasks keep distinct review contracts.

  4. Preserve
    Temporal provenance

    The accepted interval remains tied to source, task, reviewer, and revision.

Native review contracts for sound and motion.

Audio and video projects fail when time ranges, units, and task meaning change between import, review, and export. Taught makes those boundaries explicit so a segment can be audited and reproduced.

Whole-asset decisions
Classify a complete recording or clip when the task concerns the entire source.
Time-segment geometry
Mark start and end positions for events, speech, actions, or other task-defined intervals.
Integer time units
Microsecond boundaries avoid rounding ambiguity across browser, API, and export representations.
Source-aware review
The timeline, media identity, and task revision travel with the human disposition.

A temporal review path that can be replayed

  1. 01
    Choose the temporal contract

    Decide whether evidence applies to the whole asset or a bounded interval.

  2. 02
    Import with provenance

    Keep source identity and duration available before a reviewer begins.

  3. 03
    Review in context

    Inspect neighboring sound or motion before fixing the interval and disposition.

  4. 04
    Export only supported evidence

    Proposal, training, export, and inference remain separate capabilities for each configured task.

Current temporal scope

  • Audio and video have native whole-asset and time-segment review contracts.
  • Temporal labeling support does not imply that every task has proposal, training, export, or inference support.
  • Text, document, and PDF workflows remain platform targets until dedicated task contracts exist.

Questions teams ask before they begin

Direct answers based on the current Taught product and deployment boundaries.

Can a task label an entire recording or clip?

Yes. Whole-asset tasks are distinct from tasks that require a bounded time segment.

How are segment times represented?

Temporal geometry uses integer microseconds so the same boundaries can survive review, API transport, and export without floating-point drift.

Does native labeling mean native model training?

No. Labeling, proposals, training, export, and inference are separate task capabilities and must be reported independently.

Bring the decisions, data, and model version into one record.

Request access for a labeling project, a custom-model workflow, or an integration with a promoted model.

Request early access