Skip to main content
The interview is in Hirify, but the text stops too early, skips passages, or doesn’t exist. Whatever is missing from the transcript will also be missing from the interview notes and the analysis. So that’s where to start.

​
An interview marked Too short

The Too short label shows when the recording lasts less than a minute. It also covers a recording of normal length in which no speech was picked up, for example when the microphone stayed muted or the wrong audio input was selected. There’s nothing to recover in that case. Check the equipment, then start a new recording.

​
The text stops partway through the interview

There are three common causes. Rule them out in this order.
A video meeting restarted after a drop, or a call that was redialed, creates two separate interviews. Each one only holds its half. You can combine them. Open Interviews, select both, then click Merge.
If it was removed from the video meeting, or if the meeting was closed and reopened on another link, the recording stops there. The rest of the conversation doesn’t exist anywhere.
In person, recording relies on the phone’s browser. A locked screen, an incoming call, or switching apps can interrupt it. Keep the page in the foreground for the whole interview.

​
The text is there, but wrong

The transcription hears correctly but writes what it knows: mangled client names, unrecognizable internal job titles, and industry acronyms turned into something else. Your organization’s vocabulary fixes this.
1

Open the vocabulary

An administrator opens Settings, then Custom vocabulary.
2

Add your terms

In Terms specific to your company, click Add, then enter the proper names, products, and job titles that come up in your interviews.
3

Check the industry vocabularies

The Enabled industry vocabularies section shows the fields already covered for your organization.
The vocabulary applies to interviews recorded afterward. It doesn’t rewrite a transcript that already exists.

​
What a poor transcript brings down with it

When the transcript is poor, the screens that depend on it say so plainly instead of making things up.
  • Key takeaways can show “This call doesn’t have enough content to generate a summary.” Another message says that no speaker was marked as the candidate: “Key takeaways are only generated for interviews with a candidate present.”
  • The analysis can show Analysis unavailable. A Run the analysis again button appears when another attempt makes sense.
  • The interview notes can show “Interview notes generation failed”, with “The transcript and the analysis aren’t affected.” and a Try again button.
Another attempt starts from the same transcript. If that transcript is too poor, the result won’t change.

​
Good to know

  • A transcript can’t be transcribed again. The audio is processed once. What wasn’t picked up can’t be recovered.
  • In person, the number chosen in Participants when you generate the QR code helps separate the voices. It’s a help, not a guarantee.
  • A partial transcript is still usable. The interview notes and the analysis rely on what exists and point out what they didn’t find rather than filling the gap.
  • If passages are attributed to the wrong person, that’s not a transcription problem but a speaker identification problem.
  • A noisy room, a loudspeaker placed far away, or several people talking at once degrade the recording before processing even starts.

Improve transcripts

The vocabulary, in detail.

Merge two interviews

Join an interview that was split in two.