Screen Recorder — part 3 (share: links, captions, titles and analytics)
Share links can be copied while a video is still uploading. Captions are made for recordings that contain audio; AI titles and summaries have separate cloud-AI privacy controls. A shared link reports basic viewing activity in the Review Window.
On this page
What it is
This page covers sharing a recording: links, captions, AI titles and summaries, the privacy of saved edits, viewer analytics, and how an agent reads frames from a recording. Starting a recording is on Screen Recorder; reviewing, editing and storage are on part 4.
Where to find it
Share from the recording bar, review window, library, or editor. Captions and AI summary settings are in Screen Recording settings.
How it behaves
Sharing
Uploading checks that each portion contains all the bytes expected from the saved video. If the file ends unexpectedly, that upload fails instead of silently publishing a shortened copy. A failed background piece upload releases its pending work so the recorder can use its fallback upload path.
Sharing uploads the recording and its transcript captions to a link. Anyone with the link can open it unless that share also requires a password or workspace sign-in. Before your first share, Omniscio asks you to acknowledge that the recording and captions may contain other people's voices or information. Declining the notice does not stop recording, but sharing remains unavailable until you acknowledge it.
Choose an expiry before the first share in the review window: 1 day, 7 days, 30 days, or Never. Quick-share links default to Never; change Quick share link expiry in Screen Recording settings → Sharing. The regular Default link expiry setting applies to new shares and defaults to 7 days. Choosing Never requires an extra confirmation. Once a recording is shared, its review window shows the current link and expiry; it cannot change that link's expiry there. You can stop sharing later; links associated with the recording are removed when the service can reach them.
Video links support files up to 50 GB when the multipart upload route is available. If only the single-file upload route is available, its limit is 500 MB; Omniscio tells you if the recording is too large for that route. If link sharing refuses a large recording, trim it or use Publish to Vimeo to upload through your own Vimeo account. Snip shares have a 10 MB limit.
With Finish and share or automatic sharing enabled, Omniscio can copy a reserved link as soon as you finish. The link may first show that the video is still uploading, then become playable when it is ready. While it uploads, the link already looks like the video page: the Omniscio bar, the recording's thumbnail in the video frame, an upload animation, and the real upload progress underneath. Shortly after, a handful of frames from across the recording are added and play in a loop, even on a page that is already open. When the camera is on from the start, a small quick copy uploads while you record and plays first; the full-quality copy replaces it when it is ready. If your computer is too busy to keep the full-quality copy up during the take, the quick copy still uploads and the link plays it within seconds of Stop, while the full-quality file is built and then swapped in. If that full-quality build fails, the link keeps playing the quick copy and Omniscio tells you once that full quality isn't coming; the recording then has no local video file. In Screen Recording settings → Sharing, Copy share links to my clipboard controls automatic sharing and whether a successful share is copied. Turn it off if you prefer to open the share receipt and use its Open button. If the link cannot be reserved, the ordinary share flow is available after the recording is saved. Captions are not required before sharing; a transcript that arrives later can update the same share page. An early link may play muted until the viewer chooses to turn sound on, depending on the browser. A custom URL name can be set when a share is created through the CLI or an agent; the review window's custom link name field is not available yet (it shows Coming soon), so links created there use an automatically generated address.
Sharing automatically can be disabled in Screen Recording settings. A configured sensitive-window rule can prevent automatic sharing if a matching window appears during recording; the take is still saved for you to review and share manually. Uploads that are already in progress are stopped when a take is cancelled, discarded or deleted.
Saved edits and sharing privacy
Edits are saved alongside the original recording, so reopening it restores the edit. Sharing or exporting after an edit waits for that edit to save; if saving fails, publication stops and you can retry. Imported videos can also be played and exported from the library.
When you share an edited video, the poster is taken from the edited video and the transcript is matched to the parts you kept, so speech you cut is not included. The summary and chapters are left out, because their text may describe removed content. Unedited recordings can include the original poster and transcript. A share reports success only when its link is available. If you stop sharing, the app removes links it can reach for the recording's current original and edited videos.
Captions for new recordings
Captions for new recordings is on by default. Omniscio transcribes recordings that are at least 3 seconds long and contain audio; the transcript appears as captions and makes the recording searchable. A take without microphone or system audio has no transcript to generate. Turn this option off to wait until you ask for captions from a saved recording.
By default captions are made by Omniscio cloud speech: the recording's audio — including any system audio of other people on a call — is sent through Omniscio's cloud to our speech-to-text provider, currently Soniox. This happens whether or not you use Voice Control dictation. If you are signed out, your plan does not include cloud speech, or the cloud service fails, Whisper transcribes the recording on this computer instead and no audio leaves it. To keep every recording on this computer, choose any speech engine other than cloud speech in Settings → Voice Control, or turn Captions for new recordings off. Captions do not hold up saving or sharing. If a transcript is not ready when a share is created, the share page explains that captions were not ready; they can appear there later if transcription succeeds.
With cloud speech selected — the default — transcription starts while you record: the audio streams to Omniscio cloud speech as the take runs, so captions are ready the moment you stop. A take you cancel or discard has already been streamed. Turn Captions for new recordings off, or choose an on-device speech engine, to stop this.
Cloud captions count against your plan's monthly speech allowance — the same minutes dictation spends, and the plan's figures are on FlowVoice. When the allowance is used up, captions are made on this computer instead (English only) until it resets. Choose an on-device engine in Voice Control settings to keep recordings from using the allowance.
Auto-titling
The Auto-title recordings option is off by default. When enabled, Omniscio can send a transcript to Claude to suggest a short title. Automatic summaries can also create a title, but only after you have seen the separate summary notice and the automatic summary setting remains on. That notice explains what is sent, and the first recording that triggers it is not sent. A title you type yourself always takes priority. If a title cannot be generated, the recording remains available under its capture-time name.
Automatic titles require a transcript with speech, a configured cloud credential, and remaining daily AI allowance. A title is not a substitute for a manual name, and it does not change the recording itself.
Auto summary and chapters
Auto-summarize recordings is on by default. Before the first eligible recording is sent for a summary, Omniscio shows a one-time notice explaining that a cloud AI will summarize recordings; that first take is not sent. The setting can be turned off from the notice or Screen Recording settings.
For recordings with speech, the summary adds a short overview and chapter list to the review window and public share page. Chapters jump to their point in the recording. Silent recordings do not need a model call. A summary that fails can be retried from the recording's review window with Generate AI summary; once a summary exists, the same button reads Regenerate summary. You can also use that button to request a summary yourself. Summary requests are paid AI calls and stop at the shared daily allowance. Summaries and chapters are omitted from shares of edited recordings because their text may describe content removed by the edit.
How fast each take was (the timing trail)
The app keeps a small, privacy-limited timing record for recent takes to help diagnose delays in saving, sharing or transcription. It does not contain recording titles, transcripts, file paths or share links, and it does not make a take wait.
Viewer analytics
For a shared recording, the Review Window can show how viewers used its public link: view count, unique viewers, average watched percentage and where viewers stopped. Detailed charts appear after 5 views; below that, the panel shows only counts. You can opt in to being told when the link is watched: your phone or desktop gets an operating-system notification while Omniscio is running, and the server emails you when it is not, so a view reaches you either way. Both are coalesced to at most one per share per hour. The email names the recording and the time; it names the viewer too when they were signed in to a workspace share, and says "someone" otherwise. On an ordinary link share nobody can be signed in, so every viewer there is anonymous — including you previewing your own link.
Analytics use viewing sessions to record progress, seconds watched, a coarse country and a pseudonymous identifier. Raw IP addresses and browser identifiers are not stored in the analytics record, and sessions expire after 180 days. A signed-in viewer's account id is kept only long enough to name them in that one email and is then removed from the session record. The viewer page sends this information in the background; it does not block playback.
Agent frame access (CLI)
An agent can request a still frame or a marked crop from a recording through the CLI. This lets it inspect a moment without asking you to describe the image.
Related
- Screen Recorder — choose sources and make a recording.
- Screen Recorder, part 2 — the controls while you record and recovery.
- Screen Recorder, part 4 — review, edit, projects, storage and settings.
Change the description, chapters and transcript
Open a saved video's review and choose Edit description and chapters. Change the description, rename chapters or adjust their start times, then choose Save changes. Chapter times refer to the original recording; shared edited videos map them to the retained footage. Use Edit captions to change transcript text and timings. You can remove every caption after confirming, then open the editor again to add captions back.
Saved empty descriptions, chapters and transcripts stay empty when automatic generation finishes later. Unsaved drafts stay in your editor while background updates arrive. Explicitly generating content again can replace the version that was present when generation started; a newer manual save still wins.
An open share page updates as its thumbnail, playable preview, full-quality video and text become available. It keeps the viewer's playback position and play/pause state. If your text saves on this computer but the shared page cannot update, choose Retry shared page. A protected share left open beyond its access lifetime may ask you to reopen the original link.
Last verified 2026-11-12