Turn long recordings into clean text.
Turn an authorized lecture, interview or voice memo into text you can check and reuse. Longer recordings can show a live, progressive transcript draft; finish recording and wait for processing, then give the relevant reviewed passages to an Agent for a summary, translation or follow-up draft.
Before you record
Confirm participants' consent and your organization's rules for recording, transcription and sharing. Check microphone and system-audio access for the voices you need. Make a short authorized sample, then replay it before a long session. For an existing audio file, keep the original and use only material you are allowed to process.
Free download · macOS & Windows · No credit card
Illustrative example—not a recorded Cue run.
…so the deadline moves to Thursday, and design review stays on Monday.
Two open questions: pricing page copy, and who owns the launch email.
Let's get the draft out today so legal has time to look at it.
— check against the audio before reuse
44% lower median polish latency in the published benchmark
Featured in Google DeepMind's Gemmaverse — see the Gemma 4 dictation case study
What people transcribe with Cue
Lectures & talks
With permission, capture a talk. Check key terms and references, then ask for the main arguments from the reviewed passages.
Interviews & research
Keep questions separate from the participant's answers. Replay a quotation before using it, and mark overlapping or unclear speech as uncertain.
Voice memos & drafts
Record a thought on your desktop or import an authorized audio file accepted by Cue's picker. Review the text before turning it into a plan.
Recording meetings? See meeting notes without a bot →
From audio to a checked next step
- Record or import. In Cue, open Notes → New note → Start recording, sign in if prompted, and check the active recording state. Already have an authorized recording? Choose Notes → Import audio and a file accepted by the picker. Keep the original file.
- Check before summarizing. Choose Finish after recording and wait for processing. Open the note's Transcript and replay consequential passages. Check names, numbers and negations; leave unknown owners or dates unresolved. Correct a working copy without erasing the source.
- Give an Agent a bounded request. Explicitly provide the reviewed passages, source references, open questions and one desired output. Ask for a draft—not permission to send, publish or change files. Do not assume another Agent already has the note or earlier conversation.
Use the recording-to-Agent context tutorial to practice with a fictional number correction, an unresolved speaker and a checkable context packet. Its copyable instruction separates evidence, uncertainty and allowed actions.
Keep the modes separate. For a short message in another app, use Dictation. Notes recording and audio import are different entry points. A desktop interface does not mean every processing step stays local; read the privacy policy before sharing sensitive material.
A faster measured text-polish step
Google DeepMind's Gemmaverse profile reports that running Gemma 4 locally for Cue's text-polish step cut median polish latency by 44%, lifted feature usage by 30%, and dropped marginal inference cost for that step to zero in the published May 2026 benchmark. Read how the benchmark path maps to the desktop dictation workflow →
If recording or transcription fails
Only my voice was captured
A working microphone does not prove system audio was captured. When Cue offers it, use Open Settings to check audio access, then Retry system audio. Continue microphone only does not guarantee remote voices will be included. Check another short authorized sample. If the needed voices are missing, stop and use an authorized source; an Agent cannot reconstruct speech that was not recorded.
Transcription failed or no note appeared
If the recording was saved, open its note and use Retry transcription when offered. Check the displayed sign-in, connection or processing error first. A short or silent capture can show No recording saved; do not assume its audio can be recovered. Keep the original imported file. For an import failure, use the file types accepted by the picker and follow the displayed error rather than assuming every audio format is supported.
Email Cue support with your Cue version, operating system and a redacted error description—not private recording audio or transcripts. See the full recording recovery guide for the next check.
Voice transcription FAQ
Is Cue voice transcription free?
Cue is free to start. Cue Plus is $19.99/month.
How long can a recording be?
Cue supports long-form recording workflows, but this page does not promise a maximum duration or unlimited recording. Check the limits and errors shown in your installed version and current plan. Test an authorized sample before an important long session, keep original imported files, and wait for processing after you finish.
What is the difference between voice typing and transcription?
Voice typing inserts short dictated text into the app you are using. Use Notes for long recordings or audio import, then review the transcript. Dictation and Agent use separate configured modes; check the shortcuts and controls in your installed version.
Which languages does Cue transcribe?
Cue supports voice input in the languages currently available in the app. Check Cue before relying on a language for your workflow.
Turn the next recording into useful text.
Start with Cue on Mac or Windows, then dictate, transcribe, and continue the work in the app already in front of you.
They hear what you said.
Cue sees what you're doing.
And does the thing, in supported apps.
Use a transcript as evidence, not a final decision
Correct a relevant excerpt before turning interview notes into requirements. If sources disagree, resolve the conflict before handing context to an Agent. Choose the next task in the meeting and interview guides.