The best recording to upload to Speech Analyzer is the rough one you made on Tuesday

A few habits make the recording easier to work with.

1. Say the talk out loud at real pace. Whispering through it or skipping slides gives the report a talk you will never give.
2. Put the microphone close. A phone on a table across the room produces a recording the transcript will struggle with, and a struggling transcript means coaching attached to words you never said.
3. Do not stop and restart. A recording with pauses for you to check notes reads as dead air.
4. Name the file something you will recognize in a month. You will want to find it again.

None of this needs equipment. A phone propped on a stack of books is enough.

Check the file before you upload it

Two limits matter. The server accepts files up to 500 MB, and supported audio and video formats are validated before processing. If a long video recorded at high quality goes over, export a smaller copy. Audio formats are supported too, so uploading just the audio is an option when the video is the problem.

What happens after you press upload

The page lays out three steps. Your media goes directly to private storage and is checked before processing. The report then finishes in the background. And a refresh does not create a second analysis, because the job state is durable.

In practice that means you can close the tab. Make coffee. Come back. You do not need to babysit a progress bar, and you do not need to worry that reloading will queue the same talk twice and leave you with two reports to sort out.

Privacy is stated just as directly. Recordings are not public. They are stored privately and served through short lived signed links, only to the processing job and your account. If the talk includes client material you cannot share, that is worth knowing before you hit upload.

Open the report with one question

When the report finishes, resist reading everything. The page's own advice is to start with the largest useful finding, review that one moment, and rehearse it. Pick one timestamp, listen to the 30 seconds around it, and decide what you would change.

The page is also upfront about the limits. Response scores are model estimates, not measured audience brain activity and not medical information. Transcripts and generated coaching can be wrong, so listen to the source recording before you rely on any claim. Treat the report as a way to decide where to spend your next hour, not as a ruling.

Then do it again

One upload tells you very little. A second upload of the revised version tells you whether your edit helped. Keep the rough recording, fix one thing, record again, and run both.

Two posts cover what comes next. Rehearse the three lines the Moments tab flags turns the report into a short practice plan, and why Compare matters more than any single score shows how to read the two recordings side by side.

• speech analysis
• rehearsal
• recording

Back to all posts