It reads what they said. It never looks at their face.
Two hundred video answers is nine hours of watching. Hireall transcribes each answer and scores the text against the scorecard you already wrote for that job — with the sentence that earned every score sitting right next to it. The picture is thrown away before anything is sent.
Your rubric, applied the same way to everyone.
A first pass on the words — so your team spends its hours on the shortlist.
Your scorecard is the rubric
Criteria come from the job's own active scorecard, so the machine is judging what your interviewers already agreed to judge. No scorecard on the job, and it falls back to four plain criteria.
Every score quotes its source
Each criterion carries a short quote lifted from the transcript and the answer it came from. You can check the claim in one click instead of taking it on faith.
Silence is not a score
When an answer contains no real evidence for a criterion, it stays unscored and says so. Our servers enforce that: a criterion with no quote behind it cannot carry a number.
Only the audio leaves
The video track is stripped on our side and the file deleted before the audio is sent for transcription. There is no frame of the recording for anything to analyse.
Optional language estimate
For roles where the language is the job, switch on a CEFR estimate read from the transcript. It is off by default, set per job, and it says plainly that it cannot judge pronunciation.
Available to your rules
Route on the video score like any other condition. Rejections and offers still stop for a person, and a rule waits for the evaluation rather than treating a missing score as a low one.
Audio out, text back, picture never
Most claims about video AI are vague about what is actually sent. Here it is: we strip the video track locally and delete the file, send the audio for speech-to-text in the EU, then score the resulting text against your rubric. The transcript is stored with the answer and deleted on the same schedule as the recording.
What it refuses to do
A video evaluation tool is only as good as the things it declines to measure. This one is told to ignore accent, pronunciation and grammar, not to reward length or fluency for their own sake, and never to guess at a protected characteristic. The overall figure is not written by the model either: our servers average the criteria that actually had evidence.
When there are more answers than hours
Volume funnels, multilingual pipelines and distributed panels.
Two hundred answers on Monday
A first pass arrives before anyone presses play, so the team starts with the answers worth their attention.
- Runs as answers complete
- Criteria scored per answer
- Feeds the combined evaluation report
Candidates in several languages
The spoken language is detected per answer, so a Turkish and an English answer to the same question are read on the same terms.
- Language detected, not declared
- Optional CEFR estimate per job
- Accent explicitly excluded from scoring
Reviewers in four time zones
Everyone arrives at the same structured read of the same answers instead of forming an impression at 23:00 on their own.
- Same rubric applied to every candidate
- Quotes make a second opinion cheap
- Kept apart from human scorecard averages
What it analyses, and what it does not
The questions legal teams ask before this gets switched on.
No, and not as a policy choice you have to take our word for: the video track is discarded on our side before anything is sent, so no picture of the candidate reaches a model at all. What is scored is the text of what they said. Emotion recognition in the workplace is prohibited in the EU, and this product is built so the question cannot arise.
The audio is transcribed in the EU. The resulting text is then sent to the language model that scores it, and those providers may process the request outside the EU under the transfer mechanisms in our Data Processing Agreement. We would rather say that plainly than claim the whole pipeline is European.
From the job's own active scorecard, so the machine is applying the rubric your interviewers already use. If the job has no scorecard, four general criteria are used instead, and the report records which of the two produced the scores.
No. The video score counts as an AI-assisted field, which means any rule that would reject, make an offer or hire on it presents that step to a person for one-click confirmation. A missing score makes a rule wait, not fall through to the other branch.
It is deliberately kept separate. The number never enters your team's scorecard average and is stored as its own record, because a rubric applied to a transcript is not the same evidence as a person who watched the answer.
A first pass on every answer
See a video interview scored end to end in a 30-minute demo.
Talk to our team
Tell us about your roles and team size — we'll map Hireall to your hiring process and the plan that fits.
Contact sales Priced per company, not per seat