SOURCERecording & transcript
No timingTranscript view
Platform administrators can create independent projects. Institutional project and team access are managed in research administration.
Archiving hides the project and stops new research actions. Restore keeps records available without reactivating research assignments. No data is permanently deleted.
Loading workspace…
A PCM is a saved configuration of pedagogical style and strategy preferences, derived from material or designed by you. It is not a student profile or the course knowledge graph. Selecting a model here does not activate it in GRITT.
Retrieve audio → transcribe → analyse teaching patterns and strategies → save a draft model.
All supplied target-speaker segments are processed in successive windows. A prominent voice is selected provisionally; confirm the teacher before using the model. The budget covers speech recognition and analysis estimates. Infrastructure is unpriced. Processing pauses before a next estimate exceeds the budget; increase the total budget to resume. No automatic paid retries. Drafts are saved without activating them in GRITT.
YouTube may refuse server access, even when you can watch the video. In that case, attach an audio/video file or transcript and resume. No screen recording is required. Two-hour / storage limits still apply. Whole-explanation synthesis proposes source-grounded links across windows, within a bounded context. These are reviewable interpretations; emotions and learning effects are not established by them.
This is the manual source editor. Use Analyse YouTube link above for automatic processing. A playlist is never imported as a whole.
Up to 128 MiB and two hours per file; the audio derivative must fit 64 MiB. Project audio storage: 256 MiB. The original upload is removed after conversion; retain your original recording. Attaching audio does not certify that an existing transcript matches it.
Transcript view
Speech recognition can mishear words and confuse speakers. Proposed voices are not identified people. Phrase timestamps are not validated waiting-time measurements.
The cue scan is a provisional signal search, not a validated interpretation. Confirm speaker identity from the recording; prominence alone does not identify the teacher.
Observed frequencies and deliberately chosen preferences remain separate. The candidate can be inspected and downloaded before being registered in the research library.
FROM CONFIGURATION TO CONVERSATION
Inspect one of your own narrative sessions. For shared comparisons, use the authorised parallel conversations in the research workspace.
Estimates, recorded provider usage and unpriced infrastructure are separate. Failed or cancelled calls can still cost money. Central usage entries are linked here, never charged twice.
Each revision records its author, reason and parent. Loading an older revision does not change the current model.