1. Separate a recorded transcript from live dictation
A Loom transcript is generated from the audio inside a stored video. It is timestamped against playback and can drive closed captions. Live dictation begins with your microphone now and inserts editable text into a selected field.
Use Loom when the video is the source of truth. Use Dictation when the intended result is a title, share message, comment, document, or other draft rather than a recording.
2. Record the correct sources in Loom Desktop
Loom Desktop lets the creator choose Full Screen, Window, Custom size on an eligible plan, or Camera only. The creator separately selects camera state and a microphone source before starting.
That microphone belongs to Loom’s recording workflow. It can capture your narration into a stored video and later transcript. It is not the same permission or artifact as macOS Dictation in an editable field.
- Choose the screen and camera capture intentionally.
- Verify the microphone input and audio indicator before recording.
- Confirm participant consent before capturing a call or other people.
3. Use the transcript and captions Loom already creates
Loom says transcription is available on all plans and supports more than 50 distinct languages. After recording, the Transcript panel aligns timestamped text with the video; creators and viewers can jump to moments, search phrases, or copy the transcript.
Closed captions are based on the video transcript, so correcting transcript text corrects captions. Eligible creators and admins can download captions as an SRT file; plan and role requirements apply.
4. Repair language detection before exporting text
Loom detects a video language from the first 30 seconds of audio and uses it for transcripts, titles, summaries, and captions. Background noise, accents, or multiple languages can produce the wrong result.
Open the Transcript menu, choose Language, and regenerate when that control is available. Loom notes that the selection guides detection but does not guarantee the resulting language. Review names, product terms, numbers, and links before reusing the text.
5. Treat recording privacy and transcript visibility as one decision
A transcript does not remove the recording. Loom workspaces have link and audience settings such as anyone with the link, workspace or company access, and only people added. Workspace administrators can restrict public links and set defaults.
Creators can also control whether viewers see the transcript. Check the current video access, transcript visibility, comments, downloads, workspace role, and sharing destination before distributing sensitive speech.
6. Dictate ordinary text around a Loom without making another recording
When the actual job is to write a title, share message, or comment, focus that editable field and trigger the Dictation shortcut configured under System Settings → Keyboard. Speak, stop, review the visible draft and destination, then publish or send it yourself.
Website or Loom microphone permission is not required merely for system Dictation to target an ordinary supported text field. Loom recording permission and system Dictation remain separate inputs with different processors and outputs.
7. Troubleshoot the microphone path that actually failed
If a Loom recording has no audio, check the selected source, Loom’s audio indicator, and macOS Microphone, Accessibility, and Screen Recording permissions for the app or browser. If a video has audio but no transcript, allow processing time and follow Loom’s current transcript-repair guidance.
If ordinary Dictation produces no text, test the same shortcut in TextEdit. If TextEdit works, refocus the Loom text field and try a short unsent draft. Do not grant broader Loom recording access to solve a system Dictation issue.
8. Use IraVoice only when the intended artifact is editable text
IraVoice 0.7.1 is a separate option for macOS 26+ Apple-silicon Macs when a held-key workflow should create editable text across supported Loom, email, document, and developer fields. Speech recognition and supported formatting run on the Mac after model setup; paid continuation needs one Gumroad key check.
IraVoice does not record a Loom, transcribe uploaded or saved video, create captions, edit footage, select viewers, change link privacy, or publish a comment. It targets a supported focused editable field and falls back to the clipboard when insertion is uncertain.