Qualitative research transcription

Focus group transcription: what automation handles, and what still needs your ears

Focus groups are the hardest test of any transcription tool, so here is the honest version.

With one voice speaking at a time and a decent microphone, Sonix separates participants reliably and keeps each turn attributed as the conversation moves around the table. You can rename six speakers to participant codes in one pass and search every group in the project from one box.

What needs review

Crosstalk is where every automated system, Sonix included, loses precision. When two participants talk over each other, expect merged or misattributed fragments in those moments. The timestamped editor makes the cleanup fast: play the disputed seconds, fix the label, move on. Plan a verification pass on overlapping segments before you code, the same pass most methodologies already require.

Stack the odds in your favor

Use an external microphone placed centrally, open with a ground rule about speaking one at a time, and have the moderator use names so voices get anchored early. For how automated speaker separation works, see the speaker diarization guide.

NVivo, ATLAS.ti, and MAXQDA: export, then import

Sonix is not a CAQDAS tool and does not write native project files; it produces the clean, structured transcript those tools are built to import.

The workflow is export, then import, and it preserves the two things coding depends on: speaker turns and timestamps.

Export formats, as documented

Sonix exports to 18+ formats, including DOCX, TXT, PDF, SRT, and VTT, with options to include speaker labels and timestamps at the interval you choose. The export menu includes a dedicated NVivo export that keeps speakers and timestamps intact, and the API adds TXT, SRT, VTT, and JSON for teams automating a pipeline. The interview transcription guide walks through the generic interview workflow.

NVivo

Use the NVivo export, or a Word export with speaker labels on. NVivo reads the speaker-formatted transcript, so you can autocode by speaker and filter a focus group down to a single participant's turns.

ATLAS.ti and MAXQDA

Both import Word and plain text documents. Export DOCX with speaker labels and timestamps, import, and code as usual. Keep timestamps on if you want to pair the transcript with its media file inside your analysis tool.

One honest caveat: the live export menu decides the final format list, and it grows with the product. Check it on your first file before you standardize a protocol for your lab.

When your methodology requires true verbatim

Thematic analysis usually wants a clean read; discourse analysis, conversation analysis, and some IPA work need every um, false start, and repetition, because the hesitation is data.

Sonix captures full verbatim transcription automatically: filler words, false starts, and repetitions are preserved rather than smoothed away, with word-level timestamps throughout. If your protocol specifies verbatim, you do not need a separate vendor or a premium tier for it.

Document your choice of verbatim or clean transcription in your methods section, and apply it consistently across the dataset. If you need the clean read instead, edit the fillers out in the browser editor and the timestamps still hold.

IRB, consent, and data security for human subjects recordings

Recorded interviews are human subjects data, and your IRB will ask where that data goes; here is what Sonix gives you to answer with, stated as the security documentation states it.

• SOC 2 Type II certified, with AES-256 encryption at rest and TLS in transit. Data is stored with AWS in the USA.
• No AI training on your data. Nothing you process through Sonix is used to train Sonix systems. Your recordings and transcripts stay yours.
• Deletion on your schedule. You can view, export, download, and delete your data at any time. When you delete, both audio and transcripts are wiped from the platform, which supports the retention limits many protocols specify.
• Access controls. Role-based permissions for team accounts, password-protected sharing, and no employee access to your content without your express written consent.
• GDPR, DPA, and SCC available for research involving EU participants. For clinical work, Sonix offers HIPAA-compliant transcription with a BAA, detailed on the clinical researchers page.

Three practical notes for your protocol

Name the transcription tool in your IRB application rather than discovering the question at review. Make sure your consent language covers recording and automated transcription, not just recording. And map the Sonix deletion controls to your required retention period, since "we can delete it" only helps if someone is assigned to do it.

None of this is legal advice, and your IRB's determination governs. The full detail lives on the security page.

Cross-language research, one tool and one workflow

Multilingual studies usually die in the handoff between transcription vendors; Sonix runs a cross-national study through one tool and one security review for the whole team.

Sonix transcribes in 54+ languages including Spanish, French, German, Mandarin, Japanese, and Arabic. The workflow that holds up methodologically: transcribe each interview in its source language, verify it against the audio with a native-speaking team member, then translate the transcript for the rest of the team.

Translation covers the same languages with a side-by-side view, so your bilingual reviewer can check the English rendering against the source line by line. Code from the source language where your methodology requires it, and from the translation where it does not. Both versions carry the same timestamps back to the same audio, and machine translation is a review tool, not a certified translation.

Pricing that fits a grant budget

Human research transcription services typically run $1 to $2 per audio minute, which is $60 to $120 per recorded hour with days of turnaround; Sonix prices by the hour of audio, in minutes of turnaround.

Pay as you go

$10 per hour of audio, no subscription. Right for a dissertation with 15 interviews or a one-off study: twenty hours of dissertation interviews costs $200 instead of $1,200 or more with a human service.

Subscriptions for steady use

• Core, $25 a month with 5 hours of transcription and translation included: steady single-researcher use.
• Advanced, $50 a month with 20 hours included: a lab mid-study.
• Pro, $80 a month with 40 hours included: high-volume teams and research firms.

Additional hours on any subscription are $10 per hour, so a heavy fieldwork month does not force a plan change. Line-item either model in a grant as hours of audio times the rate.

The academic discount is real and self-serve

Anyone with a verifiable .edu or international equivalent school email gets 20% off every plan, every seat, every renewal, applied at checkout, and it covers the whole lab account. Registered nonprofits qualify through a one-time document check. Every new account starts with 30 free minutes to test your own audio, no card required. The full cost math, including the self-transcription time trade, is in the transcription cost guide.