PersatePersate documentation

Opening a proceeding

The detail view for a single session — the video player, transcripts, speaker identification, and the choice between the Record Analysis and Enhanced Record pipelines.

Selecting a card on the Live proceedings list opens the Detail view for that proceeding. The screen is split into two regions: the video player at the top, and the analysis panel below. The two regions share state — seeking the video repositions the analysis cursor, and selecting a transcript line jumps the video to the corresponding moment.

A breadcrumb at the top of the screen reflects the originating sub-view (Ongoing, Previous, or Scheduled), followed by the session title.

The video player

The player is positioned in the upper portion of the screen.

  • For ongoing sessions the live stream is loaded from the official Sejm broadcast channel. Playback begins automatically once the stream is available.
  • For previous sessions the recording is served from the Sejm archive. Playback begins automatically when the page is opened.
  • For scheduled sessions no stream is available until the session begins. The player is replaced with a placeholder reading "Scheduled — analysis will be available once the session begins". For sessions starting within the next thirty minutes, the player attempts to connect in advance so playback may begin promptly when the broadcast goes live.

A toggle in the player controls displays or hides the face overlays — coloured rectangles indicating speakers detected in the current frame. The setting is preserved across sessions for the user's account.

If the stream cannot be prepared for playback, the player is replaced with a Playback unavailable notice describing the failure mode. A retry control re-attempts the stream resolution.

The two analysis tabs

For previous sessions the analysis panel is preceded by a tab strip with two entries:

  • Record Analysis — the realtime pipeline. Available for both live and recorded broadcasts.
  • Enhanced Record — the comprehensive pipeline. Available only for recorded broadcasts.

For ongoing sessions only the Record Analysis view is shown; the Enhanced Record pipeline operates on completed recordings. For scheduled sessions both views are hidden until the broadcast begins.

A status dot beside each tab label reflects the current state:

IndicatorMeaning
Hollow ringNot started. The pipeline has not yet been requested.
Spinning circleThe pipeline is starting, buffering, or in progress.
Pulsing cyan dotA real-time session is currently running (Record Analysis only).
Filled dotResults are available.
Alert iconThe pipeline failed. The full error is shown in the panel body.

Record Analysis

The Record Analysis tab presents the output of the realtime pipeline. The pipeline runs in two modes depending on the proceeding's state:

  • Live Session — the realtime mode used during an ongoing broadcast. Transcription, speaker detection, and face recognition are produced as the session unfolds. The most recent segments appear at the foot of the panel; the cursor follows the live broadcast position.
  • Priority Review — the realtime mode applied retroactively to an archived recording. The same realtime pipeline is run against the historical broadcast, producing transcription with speaker labels and face identification within minutes rather than the longer turnaround of the comprehensive pipeline.

Where no Record Analysis has yet been produced for the proceeding, the panel offers a Start Record Analysis control. Selecting it begins a Live Session for an ongoing broadcast or a Priority Review for an archived recording, depending on the proceeding's state.

Once analysis is in progress, the panel displays a queue indicator (when applicable), followed by transcript segments as they become available.

Transcript segments

Each segment shows the speaker (where identified), the spoken text, and a timestamp. Selecting a segment seeks the video to the corresponding moment. During live playback the active segment is highlighted; selecting an earlier segment temporarily detaches the cursor from the live position.

Speaker profiles

Where the system has matched a face to a known parliamentary figure, the speaker name is shown alongside the segment. Unidentified speakers are labelled by a generated identifier (e.g. Speaker 3) and may be reconciled with known profiles in a later release.

Enhanced Record

The Enhanced Record tab presents the output of the comprehensive pipeline. This pipeline is slower than Record Analysis but produces a higher-fidelity result — a polished transcript, topic discovery, a structured timeline of the session, and complete facial recognition across the recording.

For recordings on which the pipeline has not yet been run, the panel offers a Start Enhanced Record control. Once initiated, progress is reported as a percentage and a stage label (e.g. Detecting faces, Recognizing faces, Transcription & diarization, Active speaker analysis). The operation may be cancelled in-flight; cancellation discards the partial result.

When complete, the panel displays the full set of derived artefacts. Selecting any item with an associated timestamp seeks the video to the corresponding moment.

Choosing between the two pipelines

The two pipelines are complementary rather than alternatives.

  • Record Analysis is appropriate when timeliness matters — during a live broadcast, or when a recently concluded session must be reviewed within minutes.
  • Enhanced Record is appropriate when accuracy and structure matter — when the session is to be quoted, summarised, or used as a reference document.

Initiating one pipeline does not preclude the other. Both may be run for the same proceeding; the corresponding tabs surface their results independently.

For most users, the typical sequence is as follows:

  1. Open Live Proceedings in the morning, switch to Ongoing to see what is currently broadcasting and to Scheduled to plan the rest of the day.
  2. For sessions in progress, open the proceeding and start a Live Session — the transcript builds in real time and may be referenced before the session concludes.
  3. For sessions of higher importance, request Enhanced Record afterwards to obtain the structured artefacts; the result is typically available within an hour for a session of average length.
  4. Use the transcript-to-video seek to revisit specific moments without scrubbing through the full broadcast.

On this page