Transcription for Thesis Research: A Masters Student's Workflow

The Realistic Plan

For a masters thesis with 8 to 15 interviews, AI transcription plus careful manual review is the right choice. Human transcription at current rates runs around $1.99 per minute (verified against Rev's 2026 pricing), which means 12 one-hour interviews would cost roughly $1,400. Most masters students do not have that budget, and they should not need to spend it. AI transcription gets you from raw audio to workable text in under an hour, and the review process that follows is where real analytic engagement begins anyway.

The doctoral-scale version of this workflow, with committee approvals and larger sample sizes, is covered in the dissertation interview transcription guide. This post focuses on the masters scope: tighter timeline, supervisor rather than committee, and a research budget that is often zero.

IRB Compliance Before You Upload Anything

Your thesis IRB approval almost certainly says something about how recordings can be handled. Verify three things before using any cloud transcription service.

Does the service train AI models on your data? For most IRBs, the answer must be no. Read the terms of service, not just the marketing copy.

How long does the service retain your audio? The retention period must align with what your IRB approval specifies. Some services delete files immediately after processing; others keep them for weeks or months by default.

Where is data stored? Some IRB protocols specify geographic restrictions, especially for research involving participants from the EU or certain sensitive populations.

For research involving vulnerable populations, controversial topics, or anonymous sources, the conservative option is local transcription using open-source Whisper. The trade-off is real: local processing requires more technical setup and is slower. But it removes cloud data handling from the equation entirely.

ConvertAudioToText does not train models on user audio and has a defined retention schedule. Verify any service's policies against your specific IRB document before uploading participant recordings.

A Workflow Built for Masters Scale

This seven-step workflow covers an 8 to 15-interview thesis project from data collection through analysis-ready transcripts.

Step 1: Write Your Methods Document Before the First Interview

Before you collect any data, write down:

• Recording device and settings
• File naming convention (e.g., P01_2026-03-15.mp3)
• Backup procedure (two locations within 24 hours of each interview)
• Which transcription tool you will use and why
• Your accuracy review process
• Anonymization procedure
• Retention and deletion schedule

This becomes a section of your methods chapter. Writing it in advance means you actually follow it and can describe it precisely. Reconstructing this document retroactively always reads weaker to supervisors because it reveals that the practice was inconsistent.

Step 2: Record With Transcription in Mind

Audio quality is the largest single predictor of transcription accuracy. A decent recording in a quiet room yields AI transcripts in the 90-95% accuracy range. A phone recording in a noisy cafe can drop to 70% and require twice the review time.

For in-person interviews, a dedicated handheld recorder placed between you and the participant outperforms phone recording significantly. For remote interviews, record locally in the meeting platform at the highest available quality and download the file immediately after.

After each interview, back up the original audio file to two separate locations within 24 hours. Losing audio in a masters thesis is a project-ending problem with no recovery path.

Step 3: Transcribe in Batches, Not All at Once

Set aside a half-day after every 3 to 4 interviews. Upload them as a batch. AI transcription processes files in parallel, so 4 interviews finish in roughly the same time as 1. Batching also forces you to stay current with your data rather than deferring everything to the final weeks.

If you are conducting interviews in a language other than English, make sure your transcription tool supports that language explicitly. Accuracy varies by language and by tool, and for less-resourced languages you should plan extra review time.

My take: the single most damaging habit I see in thesis projects is saving all transcription for the end. By then, you have forgotten the texture of early interviews, your timeline has collapsed, and you are trying to do mechanical work at the same time as deep analytic work. Transcribing in batches as you go is not more effort; it is less.

For a qualitative research transcription overview, including guidance on format choices and workflow options, that post covers the broader context.

Step 4: Review Each Transcript Carefully

Plan 15 to 30 minutes of review per hour of audio. Open the transcript next to the audio. Scrub through and correct:

• Proper nouns: names of people, places, organizations, and institutions
• Field-specific terminology your tool has not encountered before
• Numbers and dates, where AI errors are frequent
• Speaker labels where automatic speaker diarization misidentified a turn

For a typical masters project with 10 interviews averaging 45 minutes each, the review stage is 4 to 8 hours of focused work. This is not wasted time: reading through transcripts while checking against audio is often where your first real analytic observations form.

Step 5: Anonymize Before You Code

Strip identifying information from every transcript before you begin analysis. This is both an ethical obligation and an IRB requirement, and doing it systematically before coding means you never accidentally include identifying details in your coded excerpts.

A defensible protocol:

• Replace participant names with consistent pseudonyms (use a separate encrypted key file that maps pseudonyms to real identities and never share it)
• Replace specific employer or organization names with generic descriptors ("a regional nonprofit" instead of the actual name)
• Replace specific dates with relative references ("about two years before the interview")
• Generalize any details that could identify a participant within their professional or social network

The ethics of interview transcription covers the IRB and consent dimensions in more depth, including what to do when a participant says something identifiable mid-interview that they may not have intended to share.

Step 6: Choose an Analysis Tool That Fits Your Scale

For a masters thesis, the right QDA tool depends primarily on what your institution already provides.

Check your university library first. NVivo and MAXQDA both offer institutional site licenses, and many universities have them. If your library has a license, you get full access for free. NVivo's student annual license runs around $130 if you are purchasing independently; MAXQDA and ATLAS.ti offer comparable academic pricing.

For 6 to 10 interviews: Taguette (free, open-source) handles basic text tagging and code-and-retrieve work well at this scale. Microsoft Word with comments is also workable, even if it feels low-tech.

For 10 to 15 interviews: Dedoose runs about $14.95 per month and adds mixed-methods capability if your thesis combines qualitative and quantitative data.

If your institution has institutional access: Use it. The learning curve on full QDA software is real, but the structure it enforces on large transcript sets pays off during the theme development stage.

The NVivo vs AI transcription comparison covers how to combine AI-generated transcripts with QDA software, including the import formats each tool accepts.

Step 7: Code and Theme Systematically

Plan 3 to 5 hours of coding per one-hour interview. For a 10-interview thesis project with 45-minute average interviews, that is 22 to 37 hours of initial coding, plus additional time for theme refinement and the writing that follows.

Aim for 30 to 60 codes across your full dataset. Over-coding is a common masters-level mistake. Doctoral projects can support 150 or more codes because the volume of data can carry that level of granularity. At 10 to 15 interviews, finer distinctions usually reflect the data less accurately, not more.

For thematic analysis specifically (the most common analytic approach at masters level), the thematic analysis from transcripts guide walks through Braun and Clarke's phases in practical terms. For other analytic approaches, coding qualitative interviews covers the range from inductive to deductive.

Time and Budget Math for a Typical Masters Thesis

For a 12-interview project:

Total post-collection work: 125 to 200 hours. Spread across 4 to 5 months, that is 7 to 12 hours per week. Tight but tractable if you start transcribing as you go rather than saving it all for the end.

Financial cost using AI transcription: under $30 for the entire project. Compared to human transcription at current rates, the savings for a 12-interview project are around $1,300 to $1,400. That difference matters on a student budget.

If you want a plain single-file upload without setting up an account, ConvertAudioToText's audio-to-text tool handles one-off interview files without friction.

Building Your Findings Chapter From Transcripts

The findings chapter of a masters thesis is built from your coded and themed transcripts.

Introduce each theme with a few sentences framing what it captures and why it matters for your research question.

Present evidence. Two to four direct quotes per theme, drawn from at least three different participants. Quotes from a single participant dominate a findings chapter in a way that looks thin to supervisors and examiners.

Analyze. Your interpretation of what the evidence shows. This is the move that distinguishes a thesis from a content summary. The evidence itself is not the finding; what you make of it is.

Connect to existing work. How your finding relates to, extends, or complicates the literature your introduction reviewed.

For a 30 to 50 page findings chapter, you might include 30 to 70 direct quotes. Verify every quote against the audio before it appears in your final draft. Transcription errors in quoted material are the kind of thing an external examiner can notice, and the correction process is embarrassing.

What Your Supervisor Will Ask About Transcription

Thesis supervisors and oral defense panels ask predictable questions about data handling. Clean answers to these questions mark the difference between a confident defense and an anxious one.

"How did you ensure participants were accurately quoted?" Describe your review process: open transcript next to audio, scrub through correcting errors, verify every direct quote before including it in the chapter.

"How did you protect participant confidentiality?" Describe your anonymization protocol step by step: consistent pseudonyms, encrypted key file, generalization of identifying details, when files will be deleted.

"How did you code your data?" Name your approach (inductive, deductive, or hybrid), describe how your code book developed over the analysis, and if you did any inter-coder reliability checking, mention it even if it was informal.

The discipline of the methods document you wrote before the first interview pays back here. You can describe what you did specifically because you planned it and followed it.

Three Mistakes That Cost Masters Projects Time

Postponing all transcription until interviews are complete. This is the most common mistake. Transcribing as you go keeps you analytically engaged with your data and prevents the timeline compression that makes the final months miserable.

Relying on AI transcripts without review. Errors propagate into your coded material and then into your quotes and analysis. The review investment of 4 to 8 hours saves considerably more time than it costs.

Over-coding for the scale of the project. More codes do not mean better analysis. They mean harder synthesis. For a 10 to 15-interview thesis, 30 to 60 codes is enough.

Common Questions

How long does transcription review take for a masters thesis?

Plan 15 to 30 minutes of careful review per hour of audio. For 10 interviews averaging 45 minutes each, that is roughly 4 to 8 hours of review work total. Doing it in batches of 3-4 interviews at a time keeps the work from compressing into a crisis at the end.

Can I use AI transcription for IRB-approved research?

Usually yes, but you must check three things against your specific IRB approval: whether the service trains AI models on your audio (most IRBs require no), how long the service retains your files, and where data is stored if your protocol specifies a geographic restriction. If any of these is uncertain, contact your IRB office directly before uploading participant audio.

Which QDA tool is best for a 5-15 interview thesis?

It depends on what your institution provides. Check your university library first because NVivo and MAXQDA site licenses are common and would give you free access. If your institution has nothing, Taguette is completely free and handles text tagging well at this scale. Dedoose runs about $14.95 per month and adds mixed-methods capability. For 10 or fewer interviews, Microsoft Word with comments is genuinely workable even if it feels low-tech.

How many codes should a masters thesis have?

Aim for 30 to 60 codes across your full dataset. Doctoral projects can support 150 or more codes because they have more data and more time to develop higher-order categories. A masters thesis with 8-15 interviews rarely benefits from finer granularity, and over-coding makes theme development harder, not easier. Start broad, refine as you read, and merge codes that point to the same underlying idea.

Sources

• Rev pricing (human transcription): https://www.rev.com/pricing
• Rev human per-minute rate ($1.99): https://brasstranscripts.com/blog/rev-ai-pricing-per-minute-2025-better-alternative
• NVivo student licensing: https://www.usercall.co/post/nvivo-software-pricing-how-much-does-it-really-cost-in-2025
• ATLAS.ti student licenses: https://atlasti.com/student-licenses
• MAXQDA pricing page: https://www.maxqda.com/pricing
• Taguette (free, open-source): https://www.taguette.org/
• Dedoose pricing ($14.95/month): https://delvetool.com/blog/ultimate-guide-comparing-qualitative-coding-software