文字起こし 学術向け と大学院研究
Thesis interviews, lecture recordings, conference talks, focus groups, fieldwork audio — speaker-labeled transcripts at $2 per hour. 100+ languages for cross-cultural research. Plain-text exports drop straight into NVivo, Atlas.ti, MAXQDA, and Dedoose.
質的研究向けに設計
Speaker labels separate interviewer from participant automatically. Plain-text export with timestamps is the standard input for every major qualitative analysis tool.
フィールドワーク向けの 100 以上の言語
Cross-cultural research, area-studies fieldwork, multilingual focus groups — all transcribe in their source languages. No English-bias paraphrasing.
大学院生向けの予算
$2/時間は 2026 年の AI 文字起こしサービスの中で確認できる最も安い料金です。1 時間 × 20 件の典型的な論文インタビューで合計 $40 — 人間文字起こしの 1 時間がそれより高くつきます。
IRB に沿ったデータ取り扱い
EU data centres, 90-day audio retention from last sign-in, no model training on customer audio. Disclose TranscribeCat as a non-PHI processor in your IRB protocol where applicable.
From research recording to coded transcript in 3 steps
録音をアップロード
Drop interview audio, lecture recordings, or focus group files. MP3, WAV, M4A, MP4 supported. Up to 500 MB / 10 hours per file.
文字起こし
Auto language detection, speaker labels, verbatim transcript with timestamps. A 1-hour interview is ready in 4–6 minutes; a 90-min focus group in 6–10 minutes.
分析用にエクスポート
Plain-text export drops directly into NVivo, Atlas.ti, MAXQDA, or Dedoose. Word export includes speaker labels and timestamps for human-readable archive.
The qualitative research workflow — from interview to NVivo coding
Most thesis-level qualitative work follows the same pipeline:
- Conduct interviews (typically 30–90 minutes, 12–30 participants for a thesis, 8–20 for a journal article).
- Transcribe verbatim. Used to be done by hand, taking 4–6 hours per interview hour. Now transcribed in 4–6 minutes per interview hour.
- Anonymise / pseudonymise. Replace participant names with codes (P01, P02) before importing to qualitative software. Find-and-replace in the editor or in the exported text.
- Import to qualitative analysis software (NVivo, Atlas.ti, MAXQDA, or Dedoose). Plain text is the universal format. Word and SRT also accepted by most.
- Code thematically. Apply codes, develop themes, run queries. The transcript is the foundation; the coding is the analysis.
Where TranscribeCat fits: step 2. We replace 80% of the manual transcription burden, reduce per-interview cost from $50–$150 (human transcription) to $2, and produce output formatted for the rest of the pipeline.
What we don't replace: the human verification pass. AI transcription is 90–96% accurate on clean interview audio, which means a 60-minute interview has roughly 200–500 words that need review. Spot-check during the anonymisation step (step 3) and you catch most issues for free.
IRB, data residency, and what to disclose in your protocol
Most research institutions require disclosure of any third-party data processors in the IRB protocol. The exact wording varies, but the relevant facts about TranscribeCat:
- Data location: audio processed in EU data centres (Frankfurt region). The EU adequacy framework applies for participants in EU member states; participants outside the EU should be informed that audio leaves their jurisdiction during transcription.
- Retention: 90 days from your last sign-in, then deleted automatically. You can also delete individual transcripts and the source audio at any time from the dashboard.
- Subprocessors: OpenAI for the transcription pass (operates under a zero-retention agreement, audio not used for training), Stripe for billing, Clerk for authentication, Modal for compute. Full list at /trust.
- No model training on customer audio. We don't use research recordings to improve our models. Neither does our subprocessor (zero-retention).
- Not HIPAA-certified. If your research involves protected health information (PHI), TranscribeCat is not currently the right vendor — pursue a HIPAA-BAA-providing service for that work.
Standard disclosure language for an IRB protocol: "Research interviews will be transcribed using TranscribeCat (transcribecat.com), an EU-based AI transcription service. Audio is processed in EU data centres and automatically deleted within 90 days of last account access. TranscribeCat does not train models on participant audio."
What academic transcription costs
$2 per hour — typical thesis-scale projects:
$24
12 × 1-hour interviews (small qual study)
$40
20 × 1-hour interviews (typical thesis)
$100
50 × 1-hour interviews (PhD-scale)
比較: 人間による学術文字起こしは音声 1 分あたり $1.50〜$3、1 時間あたり $90〜$180。20 インタビューの論文を $1.50/分で行うと $1,800。$2/時間なら同じ論文が $40 で済みます。
Frequently asked questions
Does this work for NVivo, Atlas.ti, MAXQDA, Dedoose?+
Yes — plain-text export is the universal format for every major qualitative analysis tool. Word export with speaker labels also works in NVivo and MAXQDA. The transcript is import-ready; just anonymise participant names first.
Is the accuracy good enough for academic work?+
Realistic accuracy on clean interview audio: 93–96% on English, comparable on most major European languages, slightly lower on languages with less training data (some smaller African or indigenous languages). The remaining 4–7% is what your verification pass catches. Some PIs require 100% human-transcription for publication; many accept AI-transcribed-then-human-reviewed.
How do I anonymise participants?+
After transcription, use the in-app editor to find-and-replace participant names with codes (Participant 1 → P01). Do this before exporting to your qualitative software so the codes propagate forward. Renaming speakers (Speaker A → P01) is one click per speaker.
Can I get a refund if a recording was unusable?+
Yes — if transcription fails or finds no speech, you get an automatic refund to your original payment method. No support ticket needed.
What about non-English fieldwork?+
100+ languages with auto-detection, including Arabic, Mandarin, Hindi, Spanish, French, Portuguese, Norwegian, Swedish, Danish, Polish, Greek, and many more. The transcript stays in the source language. Code-switching (mixing two languages within one interview) is handled.
Can my whole research group share an account?+
Yes — set up an Organisation at /org/settings. One billing card pays for all uploads by all members. Per-member usage is tracked. No per-seat fees. Useful for departmental research budgets and grant accounting.
How does this differ from /for/researchers?+
Significant overlap. /for/researchers focuses on the research-method side (interview design, focus group facilitation, fieldwork). /for/academia focuses on the institutional context (IRB, qualitative analysis software, thesis budgets). Pick whichever resonates — same engine, same pricing.
関連する研究・学術リソース
Start transcribing your research
$2 per interview, NVivo-ready exports, EU data residency. Reclaim a year of grad-student transcription time.
文字起こしを始めるNo card needed to sign up.