A transcription project checklist keeps recordings, transcripts and the review process consistent, whether you are working on an interview, focus group, meeting, legal discussion or academic research. It also helps prevent avoidable mistakes: missing speakers, incorrect names, invented words, inconsistent timestamps and a final file that does not match the brief.
This guide contains 25 actual checks, arranged into five stages: project preparation, recording quality, transcription, quality assurance and delivery. Use the steps in order for a new project, or take the relevant checks into an existing workflow. Each check describes what to verify and why it matters.
If you need a transcription provider rather than a project-management guide, see TranscribeLingo’s transcription services. For advice on evaluating a provider, read our separate UK transcription service selection checklist.
The 25-step transcription checklist at a glance
| Stage | Checks | Purpose |
|---|---|---|
| 1. Preparation | 1–5: Purpose, style, language, format, security and deadline | Define what a correct finished transcript looks like |
| 2. Recording assessment | 6–10: Files, sound, speakers, glossary, duration and timestamps | Identify risks before transcription starts |
| 3. Production | 11–15: Method, speaker attribution, terms, overlaps, inaudible speech | Create an accurate, consistently marked draft |
| 4. Quality assurance | 16–20: Audio comparison, names and figures, timecodes, style, proofreading | Correct factual and formatting issues without changing meaning |
| 5. Delivery | 21–25: Layout, completeness, file formats, secure transfer, sign-off | Hand over a usable and correctly scoped result |
Stage 1: Project preparation (checks 1–5)
Agree the intended output before transcribing anything. Misunderstandings about transcript style or speaker identification can be harder to correct after hours of recordings have been processed.
1. Define the purpose and use of the transcript
Ask whether the material is for qualitative research, internal reference, publication, legal review, accessibility or translation. The intended use determines the level of detail, notation and quality-control needs. For instance, research interview transcripts may need hesitations and pauses, while internal meeting transcripts may use an agreed clean-verbatim approach.
Pass check: The brief names the audience, intended use and expected deliverable.
2. Choose the transcript style before work begins
Confirm whether the client wants strict/full verbatim, clean verbatim, intelligent verbatim or a separately labelled edited summary. Full verbatim may retain fillers, false starts and repetitions. Clean or intelligent styles may remove agreed disfluencies, but should not change what was said. Different clients use these labels differently, so capture the actual rules.
Pass check: You have an approved style description and an example. See verbatim vs intelligent verbatim transcription for a detailed comparison.
3. Confirm the language, dialect and speaker count
Record the spoken language or languages and any relevant regional varieties. Determine how many voices are expected and whether speakers are known, anonymised or to be labelled numerically. For bilingual recordings, clarify whether speech should be retained in its original language, translated, or supplied in two versions.
Pass check: Language handling and speaker identifiers are agreed; transcription and translation are quoted as separate tasks where appropriate.
4. Agree the document structure and formatting
Ask for the required Word, text, PDF or subtitle output; speaker-label convention; timestamp interval; paragraph breaks; punctuation standard; and any existing template. If the client requires UK English, apply that standard consistently to an English-language transcript while preserving names and spoken meaning.
Pass check: A short formatting specification or reference transcript is available to the transcriber and reviewer.
5. Set delivery, permissions and confidentiality requirements
Confirm the deadline, file-sharing method, access restrictions and who may handle the recording. Ask whether there is sensitive personal, medical or legal information, and whether extra handling restrictions or contractual terms apply. Plan enough time for transcription, independent review if required, client queries and delivery.
Pass check: The approved workflow and delivery requirements are recorded before transferring sensitive recordings.
Stage 2: Recording and resource checks (checks 6–10)
If the audio was captured on an iPhone, use our iPhone Voice Memos and Notes transcription guide to retrieve the recording and an automatic draft before applying this checklist’s quality checks.
An early file review prevents wasted work. Clear instructions cannot compensate for a corrupt recording, an absent audio channel or the wrong source file.
6. Confirm every file opens and is the correct version
Create a source-file inventory with filenames, file types, duration and a unique reference. Open the beginning, middle and end of each recording, and verify that the correct final files—not previews or partial downloads—were received. Check that audio is present throughout.
Pass check: Every expected recording appears in the inventory and plays to its documented end.
7. Check volume, clipping, distortion and speech clarity
Listen for very low speech, distortion, clipping, echo and sudden level changes. In stereo or multi-channel files, confirm that relevant voices are not present only on a channel the workflow is ignoring. Note where the recording quality makes accurate transcription uncertain.
Pass check: The team knows which passages need additional review and has agreed how to handle material that is genuinely unclear.
8. Identify background noise and overlapping speech
Mark stretches affected by traffic, music, microphone handling or people speaking simultaneously. Audio processing may help some sections, but heavy filtering can erase information; preserve the unmodified original and never invent words from an enhanced recording.
Pass check: Difficult sections have references for follow-up and realistic expectations are agreed. For recording preparation, see how to record quality audio for transcription.
9. Collect correct names, abbreviations and terminology
Request speaker names or approved pseudonyms, company names, technical abbreviations, product names and topic-specific glossaries. If speakers refer to acronyms or unusual terms, use the supplied reference material to check spelling—not to substitute an expected answer for what the audio says.
Pass check: A terminology sheet is ready, and any unverified spelling can be flagged for clarification.
10. Check duration, time references and timestamp rules
Confirm each file’s total playable duration and whether the transcript should use elapsed time from the file start, a recorded meeting clock or another specified reference. If several recordings form one interview, decide whether timestamps restart per file or remain continuous.
Pass check: The timestamp method is documented and can be followed consistently across all files.
Stage 3: Accurate transcription production (checks 11–15)
A draft transcript should prioritise what is audible, not what the transcriber assumes the speaker meant. Consistency is especially important in research, legal and multilingual work.
11. Select a suitable human or AI-assisted workflow
Choose manual transcription, automatic speech recognition (ASR) with human correction or another documented process based on language, audio quality, confidentiality requirements and the consequences of errors. Automated output is a draft to verify where accuracy matters, not proof that speech was captured correctly.
Pass check: A named person is responsible for accuracy, regardless of the tools used.
12. Apply consistent speaker labels
Use the agreed convention, such as INTERVIEWER and RESPONDENT, or Speaker 1 and Speaker 2. Keep the same identifier attached to the same voice throughout. Do not infer someone’s name or role solely from the subject of their speech. When identity cannot be resolved, use an agreed neutral label.
Pass check: A reader can track speaker turns without ambiguous switching.
13. Capture specialist terms, numbers and quotations carefully
Pay extra attention to dates, monetary amounts, medication names, case references, addresses, figures and negation. Check difficult spellings using the approved glossary or a suitable authoritative reference. Do not ‘correct’ a speaker’s factual error unless the project explicitly requests annotations, and keep any annotations separate from spoken words.
Pass check: High-risk words and figures have been checked against the actual recording.
14. Handle overlapping speakers and non-speech events
Where people talk simultaneously, represent what can genuinely be heard and mark overlap using the client’s preferred convention. If relevant, note laughter, crying or a significant interruption without over-describing the scene. Avoid implying a speaker said a complete phrase when part of it was obscured.
Pass check: The transcript distinguishes confirmed speech from overlapping or non-speech material.
15. Mark genuinely inaudible material rather than guessing
Use a consistent, approved marker such as [inaudible 00:06:14] for a passage that cannot reliably be understood after a reasonable replay. Where only one word is unclear, preserve the intelligible words around it. Use unintelligible, overlapping speech or another label only if the client defines that convention.
Pass check: No invented or unsupported words have been inserted into uncertain passages.
Example: how a correctly marked transcript might look
This fictional example shows speaker labels, elapsed time and an inaudible marker. It is not a quotation from an actual client recording and is not a universal formatting rule.
| Time | Speaker | Transcript |
|---|---|---|
| 00:00:08 | INTERVIEWER | Could you describe what happened at the meeting on Monday? |
| 00:00:14 | RESPONDENT | I sent the draft to the project manager, but the figures had not been checked. |
| 00:00:21 | INTERVIEWER | Which figures needed checking? |
| 00:00:25 | RESPONDENT | The final total was [inaudible 00:00:28], so I asked for the original spreadsheet. |
Stage 4: Transcript quality assurance (checks 16–20)
The highest-value checks compare the written transcript with the recording. A spelling-only proofread will not detect missing statements, wrong speaker attribution or reversed meanings.
16. Compare the complete transcript against the audio
Carry out an audio-following review according to the agreed risk and quality plan, not only a visual read-through. Confirm that words, speaker turns and significant passages are present. Give high-risk or unclear sections additional attention, and use a separate reviewer where the project calls for it.
Pass check: Every file has a documented review status, and critical discrepancies have been resolved or clearly flagged.
17. Double-check names, dates, quantities and negation
Check proper names, numeric values, percentages, currency, dates, doses and statements containing not, never or similar negatives. These errors may substantially alter meaning. Validate against the audio first and approved supporting material second. If the audio itself is uncertain, record that uncertainty.
Pass check: Key terms and figures are verified without silently rewriting the speaker.
18. Verify timestamps and speaker transitions
Sample timestamps against playback at the start, middle and end; also check points where files join or speakers interrupt each other. Look for timestamp drift, incorrect sequence, duplicated intervals and labels carried over from a different speaker.
Pass check: Time markers correspond to the declared source and speaker assignments are consistent.
19. Apply the agreed verbatim and editing rules
Review whether fillers, repetitions, grammar adjustments, punctuation, false starts and non-speech events match the client’s style. Intelligent transcription may improve readability, but should not recast the speaker’s meaning, attribute an opinion to another person or turn uncertainty into certainty.
Pass check: All sections follow one documented editorial convention.
20. Proofread the transcript without altering meaning
Check spelling, paragraph breaks, obvious typos, inconsistent terminology, accidental duplicate lines and broken tables. Recheck any proposed substantive edit against the recording. A second set of eyes can help, but even a second reviewer needs access to the correct audio and requirements.
Pass check: The final copy is readable, internally consistent and faithful to the source. See common transcription errors and how to prevent them.
Stage 5: Final checks and delivery (checks 21–25)
The last stage makes sure the work is useful, complete and deliverable—not merely accurate within the main body text.
21. Inspect formatting in the actual output file
Open the supplied document in its intended format rather than judging only the editor view. Check title, file reference, headings, tables, speaker labels, line spacing, page numbers if requested, and mobile or print readability if relevant. Remove accidental editing notes and tracked changes only as agreed.
Pass check: The delivered file follows the approved template and contains no unintended formatting artefacts.
22. Reconcile every file and transcript section
Compare the transcript set with the original source-file inventory. Ensure all agreed recordings, segments and requested language versions are present. Confirm that recording names, respondent IDs, interview dates and section order match the supplied project details. Do not claim a recording is complete if its ending is missing.
Pass check: There is a one-to-one reconciliation of expected source materials and delivered outputs, accounting for any intentional combined files.
23. Verify the final file format and naming convention
Produce the formats agreed in advance: for example, editable DOCX, PDF for review, plain text or caption files where separately requested. Check opening and rendering, filenames, version numbers and whether a transcript should include a bilingual layout. Keep a clearly identified final version.
Pass check: Each file opens, has a meaningful name and matches the requested format.
24. Check the approved delivery channel and access controls
Confirm that the recipient and file-transfer method are correct and approved for the material’s sensitivity. Make sure links, permissions, expiry settings and password-sharing arrangements are suitable. Avoid placing confidential recordings or transcripts in publicly accessible folders. Follow the agreed retention and deletion terms.
Pass check: The authorised recipient can access the complete final files without exposing them to unintended parties.
25. Complete final sign-off and record any outstanding queries
Use a final sign-off record listing the source files received, output files delivered, transcript style, review outcome, agreed exceptions and any remaining genuinely inaudible sections. Ask the designated project contact to resolve questions that prevent acceptance, and record any amendments as a new version.
Pass check: A reviewer can identify what was delivered, what was checked, who approved it and what remains unresolved.
The important point is that the unclear phrase remains visibly uncertain. An editor should not insert a plausible number simply because it would make the sentence read smoothly. If a different timestamp or notation style is agreed, use that style instead.
Final transcription QA sign-off template
Copy these fields into a delivery checklist or job-tracking system. Mark each one Pass, Fail or Not applicable; add the reviewer and date where the project requires an audit trail.
| Item | What to record |
|---|---|
| Project and source inventory | Project ID; file names; number of files; total playable duration |
| Delivery scope | Languages; verbatim style; speakers; timestamps; output format |
| Source completeness | All expected files present; no truncated or missing endings |
| Audio-accuracy review | Reviewer; review date; exceptions and follow-up actions |
| High-risk details | Names, figures, dates and specialist terms checked |
| Inaudible or overlapping speech | Locations recorded using agreed notation |
| Formatting review | Speaker labels, template and output file checked |
| Delivery and access | Approved recipient and file-sharing method verified |
| Acceptance | Final version, approval status and any open questions |
Transcription checklist mistakes to avoid
- Assuming automated transcription is a finished transcript. Good ASR output can still omit negatives, confuse speakers or mishear figures.
- Using “verbatim” without defining it. Clients may mean full verbatim, clean verbatim or another style.
- Inferring missing speech from context. An intelligible-looking sentence is not evidence for words that cannot be heard.
- Changing speech into meeting minutes. A summary needs separate authorisation and should be labelled accordingly.
- Missing the end of a recording. File checks should include the end, not only the first few minutes.
- Overwriting uncertainty during proofreading. A final read should not erase a legitimate inaudible marker.
- Sharing sensitive files without checking recipients. Delivery is part of the quality workflow.
Frequently asked questions about transcription quality
How do I know whether a transcript is accurate?
Accuracy is judged by comparing the text with the original recording using agreed transcription rules. Check completeness, speaker attribution, names, numbers, negation and uncertainty markers. A polished-looking transcript may still contain factual errors if it has not been audio-checked.
Do all transcription projects need 25 separate checks?
The checklist is a planning and QA framework, not a claim that every project needs the same amount of work. Apply proportionate checks based on the material’s sensitivity, length, subject matter and intended use; document any not-applicable items.
Should I choose full verbatim or clean verbatim?
Choose based on the purpose. Full verbatim retains more features of how something was said and may suit certain research or evidential needs; clean verbatim is easier to read when pauses and fillers are not necessary. The client should approve exactly which changes are allowed.
How should inaudible words be marked?
Use an agreed marker, for example [inaudible 00:06:14], ideally with a time reference. Do not replace a genuinely uncertain word with an unsupported guess. Preserve the surrounding intelligible text.
Is proofreading alone enough for transcript quality assurance?
No. Proofreading finds typography and readability problems; it cannot reliably identify a wrong speaker, missing sentence or incorrect number without listening to the audio. An audio-following review remains important.
How long should transcription QA take?
That depends on recording clarity, speaker count, subject complexity, the review depth and the amount of editing required. Build review time into the project schedule rather than treating it as an optional final minute.
Can transcription include translation or subtitles?
Yes, but they are distinct deliverables. Translation changes the output language; subtitles require timing and display constraints in addition to the words. Specify each service, language pair and format before work starts.
Need a professional transcription team?
A reliable project begins with a clear specification, suitable handling of the recordings and a defined review plan. Explore our professional transcription services or request a quote with the duration, number of speakers, language, topic, transcript style and deadline. We can assess the scope before confirming a quote and available turnaround.
If you are working with interviews or focus groups, you may also find our interview transcription guide useful. The goal is to produce a transcript that can be traced to its source recording, reviewed consistently and handed over in the agreed format.
