Hebrew interview transcription
Built for journalists, qualitative researchers, doctoral students, and documentary makers who record long-form Hebrew interviews with several speakers. Upload the raw file - up to 500 MB, which fits an 8-hour MP3 - and get a right-to-left transcript with speakers separated, plus an AI summary with key points. Rename speakers to pseudonyms, export to Word, and import into NVivo, ATLAS.ti, or MAXQDA.
What this tool does
A long Hebrew interview is a different beast from a short meeting clip. Participants talk over each other, drop an English or Arabic word mid-sentence, trail off, correct themselves. RecordX transcribes the whole file, separates up to 10 speakers, and writes an AI summary with key points, action items, and decisions - in Hebrew, right-to-left.
Each file is transcribed in one language, so English or Arabic words inside a Hebrew interview won't always come out right. For participant names and field terms, add them to your custom vocabulary in Settings and pick Hebrew explicitly.
A free account gives you 60 transcription minutes every 30 days - enough to try it on a real interview up to an hour long. Longer or more frequent interviews need a paid plan; see pricing.
How it works
- Upload the interview. Sign up at recordx.io, open the web app, and drag the file in. Files up to 500 MB go in as they are, so a 2-hour interview saved as MP3 or M4A fits without splitting.
- Wait for the transcript. Speaker separation, the transcript, and the AI summary with key points all run automatically.
- Rename, pseudonymize, export. Rename Speaker 1 to a real name or to P1 for publication, then export to Word with "Include transcript" turned on. Each turn comes out as "Name: text", ready to import into ATLAS.ti, NVivo, or MAXQDA.
Formats supported
The usual field recorder and phone formats work with no conversion step: WAV from a ZOOM H4n or H5, a Tascam DR-series, or a Sony PCM recorder; M4A from iPhone Voice Memos or an Android recorder; plus MP3, FLAC, OGG, and WebM. Video works too - drop in an MP4 from Zoom. The limit is 500 MB per file. A 2-hour interview as MP3 or M4A fits easily; a long uncompressed WAV can go over, so export it to MP3 or M4A first.
What kinds of Hebrew interviews work
- Academic research interviews - semi-structured interview guides, multi-session designs, two to three speakers per session, long transcripts you plan to code.
- Journalist source interviews - long-form sit-downs where every quote has to be tied to the right speaker.
- Doctoral dissertation participant interviews - IRB-approved recordings with pseudonymized export for the appendix and the data archive.
- Oral history recordings - elderly participants and long sessions, turned into a full written record of what was said.
- Documentary subject interviews - on-camera recordings from a shotgun mic or lavalier into a field recorder; a full transcript to pull lines from for the paper edit.
What you get back
Speaker-labelled transcript
Up to 10 speakers separated, each turn tagged. Rename them to real names, or to P1 / P2 / P3 pseudonyms for an IRB or Helsinki committee, before you share or publish.
AI summary and key points
A short summary, a longer structured one, key points, action items, and decisions - written in Hebrew. Ask follow-up questions about the interview in the AI chat.
Word export for coding
Export to Word with "Include transcript" turned on and you get the full transcript with speaker names, ready to import into NVivo, ATLAS.ti, or MAXQDA. PDF export is there too.
Beyond file uploads
AI notes for every interview - bot or no bot, your call. The same RecordX account captures interviews across every surface a conversation can happen on:
- Record from the web browser - open RecordX on the web and hit record. Captures mic audio for a sit-down interview on your desk, or system audio for a call playing in another tab. The transcript starts building as soon as you stop.
- Chrome extension for remote interviews - captures Zoom, Google Meet, Microsoft Teams, or Webex running in the browser. No bot joins the call.
- Android native app - captures field interviews, in-person sit-downs, and phone interviews on speaker, with the phone in your pocket. Runs with the screen locked. iOS coming.
- Auto-join from your calendar - connect your Microsoft Outlook calendar and a notetaker auto-joins the interviews you accept.
- Softphone - team plans - phone the source from inside RecordX on a shared organization DID. Phone interviews land in the same library as everything else.
- File upload - the baseline, when you already have the audio - a voice note from the source, a recorded call, a transferred field tape.
FAQ
How well does it handle several speakers in one Hebrew recording?
Speaker separation is on by default and handles up to 10 voices, with each turn tagged by speaker. Rename Speaker 1 / Speaker 2 to real names in a couple of taps, or to P1 / P2 / P3 if you need to pseudonymize for publication.
Can I pick Hebrew explicitly and add names and field terms?
Yes. The language is auto-detected by default, or you can pick Hebrew yourself. Add participant names, place names, and jargon to your custom vocabulary in Settings - it applies when you pick the language explicitly. Each file is transcribed in one language.
Can I export to Word for ATLAS.ti, NVivo, or MAXQDA?
Yes. Export to Word with "Include transcript" turned on - it's off by default, which gives you the summary only. Each turn comes out as "Name: text", and the .docx imports straight into ATLAS.ti, NVivo, or MAXQDA for coding.
Is my recording confidential? Who sees the audio?
Audio files are deleted automatically after processing, so only the text stays in your account. Transcripts are encrypted at rest, kept for up to 12 months, never used to train AI models, and not shared with other customers or third parties. For an IRB or Helsinki committee, rename speakers to pseudonyms before you export, so the speaker labels carry no real names. Names said aloud stay in the text, so check for them before you share.