Choose the useful moment from a longer recording.

A product demonstration or interview may contain one moment that belongs in a short video. Store the original privately, choose its start and length, then prepare that range as editable footage and separate audio. The original stays unchanged.

Upload once, then select a range.

  1. In the video studio, open Reuse workspace media → Prepare clips from a larger recording.
  2. Choose an H.264 MP4 up to 64 MiB and 20 minutes, at most 1080p and 60 fps, with optional AAC audio. Confirm your permission to store it and prepare clips.
  3. Upload the original. Transfers use 5 MiB parts. Pause transfer stops after the current part; already confirmed parts remain private.
  4. Select a start time and a length from 1 to 30 seconds, in tenths of a second. Confirm permission for the selected range and choose Prepare this range.
  5. Return to its status when preparation finishes. This step continues after you leave the page.

For example, a 150-second recording can supply the moment from 125 to 128 seconds. Enter a start of 125 and length of 3. Preparing that range does not insert it into your video.

Find a passage in the original.

For a stored recording, open Transcribe and search this recording, select English or Spanish, confirm your permission, and choose Transcribe full original. Processing continues after you leave; return to its retained status. Two attempts per workspace per UTC day and one active attempt apply, including failures and cancellations.

A complete transcript uses the original recording’s clock. Search its passages, compare the wording with the audio, then select a passage and adjust its start and length. Choose transcript range for preparation fills the range form; confirm its permission and prepare it separately. It does not insert media or captions into your current video.

Download the full candidate as VTT or SRT. VTT includes a source note with the original’s hash, language and model. Keep an SRT with its separate source receipt. If no usable speech is found, check the original audio and selected language before using another attempt. Names, wording, pauses and audio-window boundaries need human review.

Review suggested highlights.

When enabled, open Find highlight candidates inside a complete transcript. Describe your goal, choose an interval on the original clock, confirm permission, and choose Suggest highlights. You can leave and return to the retained result. Two attempts per workspace per UTC day and one active attempt apply, including failures and cancellations.

The model can select up to three distinct, contiguous passages of 1–30 seconds. Their text and times come from the transcript. Only complete passages inside your interval are considered; dense transcripts may require a shorter interval. If no passage fits, choose a range manually. Suggestions are editorial candidates, not verified claims or predictions of views.

Read each candidate and its surrounding context, then choose Select this range. Adjust the transcript range if needed and choose it for preparation. Prepare and listen to the result before confirming captions or saving a separate video. Selecting a suggestion alone does not change your current video. Retained review remains available when new suggestions are disabled.

Prepare several reviewed ranges together.

Open Prepare several reviewed ranges and enter two or three distinct, nonoverlapping moments on the original clock. Each range must last 1–30 seconds. You can add the selected transcript passage as range 3. Check the context and permission for every range, then choose Prepare reviewed batch.

The complete batch is reserved together; every range counts toward the same five daily preparation attempts. It must fit the remaining daily and library capacity. Processing runs one range at a time and continues after you close the page. Return to Clip batches for each child’s status and use Review batch range to inspect a ready clip before starting its separate editable video.

Cancel unfinished ranges hides late results and retains ready clips. Admitted failures and cancellations still count. If request confirmation is lost, retain and retry the same batch request; it recovers its receipt without another admission. Refresh history before clearing a retained request. Failed or uncertain native preparation never retries automatically. A batch does not approve or publish videos.

Resume an interrupted upload.

Keep the page open while bytes transfer. After reopening, select the interrupted recording and choose Reselect original file. Use the exact original file: its length and SHA256 must match. Confirmed parts are reused. A different file is refused, so it cannot silently change the stored recording.

Unfinished transfers expire after 24 hours. Cancel a transfer you no longer need. Five stored originals and two active transfers apply per workspace; pending storage cleanup can continue to occupy a slot.

Review footage and audio separately.

A ready range contains a silent H.264 clip at 30 fps with a maximum 720-pixel short edge and 1280-pixel long edge. Smaller footage is not enlarged. When the original contains AAC audio, its range also produces a separate mono WAV at 16 kHz.

Play both previews, check the chosen moment and permissions, then explicitly use the footage in the selected scene or use its audio as the primary track. Footage replaces that scene’s background. Primary audio replaces the current narration and clears captions. Each action has Undo; nothing changes merely because preparation finishes.

To make an independent clip, open Start a separate video from this range. Choose its format, whole-source or center-crop framing, and whether to keep the original-range audio. After previewing and confirming permission, choose Start separate editable video. Its full range opens as an unsaved draft. If you selected the range from a complete transcript, explicitly confirm that you checked its wording against this range’s audio to include captions. Their times are rebased to the new video and remain editable. Edit its title, framing and captions, then Save video. The video previously open remains unchanged; unsaved edits receive the normal leave-page warning.

Adjust framing and source trim, scene timing, and captions, then save a new version. For a short prepared video, review pauses and filler words before saving. Watch the completed export with sound before sharing.

Keep the original and its permissions.

Download original recording checks the stored length and hash before handing the unchanged MP4 to your browser. Keep its usage permissions with your copy. Original files, range history and prepared media remain private to the workspace.

Deleting an original is permanent and prevents future ranges. Active preparation, original transcription or highlight selection blocks deletion; existing prepared clips, completed transcripts and their original range identity remain. Five preparation attempts per UTC day apply, including failed and cancelled attempts. A failed or uncertain encode is not automatically repeated.

The editable-file limit remains 6 MiB and a render can use up to 12 MiB of media. Compositions remain limited to two minutes at 720p or one minute at 1080p. This workflow supports manual ranges and, where enabled, reviewed highlight candidates from a full-original English or Spanish transcript. Reviewed batches retain explicit ranges and separate output review. Speaker tracking, autonomous editorial selection/publication and 4K export are still unavailable.

Open the video studio →

Upload through your agent workflow.

Use the private media CLI guide to inspect a request, upload an original or resume numbered recording parts with scoped workspace credentials. Human review remains a separate step.