Prepare an Audio Transcription Sample You Can Check

An audio transcription service needs a defined text style, speaker labels, timestamp rules and a way to handle speech you cannot identify. Before accepting a file, listen to a representative portion and confirm the intended output. Audio duration alone does not tell you how long the work will take.
This guide explains a proposed practice workflow for ordinary transcription. The sample uses your own invented recording. It does not establish competence for legal, medical or other specialist records.
Separate the requested outputs
A transcript represents speech as text. Captions also need to work with the timing and presentation of a video; a summary selects and condenses information. Clarify which deliverable the client needs before describing all three as the same service.
Agree on whether the transcript should retain fillers and false starts, how speakers will be labelled and where timestamps should appear. Ask for the actual style instructions and any approved names or terminology. A client requesting readable text may still expect the speaker's meaning and wording to remain intact.
State the file format and intended use. A document for reviewing an interview differs from text prepared for publication. If the client wants editing beyond the agreed transcription style, identify that as a separate decision rather than silently rewriting the recording.
Follow a defined accuracy standard
Rev's public transcription guide says to preserve what was spoken, attribute speech to the right speaker and avoid adding or paraphrasing content. It also explains research for unfamiliar terms and timestamped inaudible notation for speech that cannot be identified. These are Rev's instructions; another client's guide may set different formatting conventions.
Source: Rev's style guide references. Read 2026-10-26. This article discusses the transcription section, not the separate caption and subtitle rules on the same page.
For your practice sample, write a short style sheet before transcribing. Decide how to show a speaker change and an unresolved word. Apply the same choices throughout the file so the reader does not have to infer whether different notation means a different problem.
Do not replace an unclear word with a plausible guess. Replay the surrounding phrase, check an authorized reference when appropriate and preserve uncertainty if it remains. Knowing what a speaker probably intended is not the same as hearing the words that were spoken.
Make a recording you have permission to use
Record a short original conversation or monologue using invented, non-sensitive content. If another person participates, obtain their permission for the practice recording and its use. A public video is not automatically free material for a portfolio sample.
For an invented conversation about arranging a community reading group, include a change of speaker and a few ordinary names you can verify from your own script. Keep the script as a checking aid, then put it aside while transcribing. Do not create a misleading claim that the audio came from a real client.
Listen once before starting the detailed pass. Note overlapping speech, background noise and any passage you cannot make out. Check the beginning, middle and end rather than assuming the first clear minute represents the whole file. This is a preparation exercise, not a guarantee that a future recording will be workable.
Keep audio time separate from working time
Track your actual time for listening, typing, replaying, terminology checks, formatting and final review. Record the audio duration separately. If you need to estimate a new assignment later, compare its difficulty with material you have actually measured, rather than assuming a universal transcription speed.
For a quote, confirm what happens if the file contains much more overlapping speech or poor sound than the sample suggested. Ask whether the client can supply a clearer recording or a list of names. If the agreed deadline cannot accommodate a reliable review, resolve that before accepting the work.
Include communication, secure file handling and revisions in the scope. A price per audio minute, when offered by a client or platform, is a unit of recorded material. It is not automatically a payment per minute of your labour or proof of the rate you will achieve.
Check marketplace access separately
Rev's freelancer page describes a skills assessment and a short sample as part of applying for transcription, caption or translation work. Reading that page does not establish acceptance, availability of suitable assignments, country eligibility or an individual's ability to receive payment.
Source: Rev's freelancer information. Read 2026-10-26. The page's promotional workload or earnings language is not a forecast.
Before choosing any marketplace, inspect its current role requirements, permitted locations, payment arrangements and work rules. Distinguish a public application page from an open account or available paid assignment. Keep the practice sample useful even if a particular platform is unavailable to you.
Review the transcript against the recording
Read while listening from start to finish. Check missing or repeated words, negations, speaker changes, names and numbers. A small change to one of those can alter meaning even when the paragraph remains fluent. Use timestamps to return to a specific uncertainty instead of searching the entire recording again.
Next, inspect the document without the audio. Confirm speaker labels and formatting are consistent, timestamps follow the requested pattern, and any unresolved notation is visible. This reading catches presentation problems; it does not replace listening to the source.
If you use an automated draft, treat it as unverified text and follow the client's rules on tool use. Obtain authorization before sending recordings to another service. Do not let fluent machine output persuade you to keep words you cannot hear or speaker identities you cannot establish.
Deliver the agreed file and open questions
Return the version and format specified in the brief, with a short note identifying unresolved passages and any agreed limitations. Avoid describing the result as certified or error-free unless you have a valid basis for that exact claim.
Keep the delivered copy separate from your working draft so later corrections are traceable. Agree on how feedback will refer to timestamps and what counts as a correction versus a new request, such as turning the transcript into an edited article.
Use the agreed retention and deletion arrangements for audio and text. Do not reuse customer material as a portfolio example without permission. A transcript can contain the same confidential information as the recording even though it is easier to search or copy.
Check cloud recordings and recipient access
If the client authorizes Word's Transcribe feature, Microsoft says recorded or uploaded audio is saved in OneDrive's Transcribed Files folder. Sharing the Word document does not, by default, let its recipient play that audio in the Transcribe pane. Microsoft describes sharing the recording separately to enable playback. A transcript recipient and an audio recipient therefore need separate decisions in your delivery brief.
Microsoft also says its transcription service does not store the audio and transcription results after processing. That statement does not mean the recording saved in OneDrive has been deleted: the same guidance tells you where to delete that file. Keep the agreed retention plan specific to the stored recording and delivered text. Source: Microsoft's Transcribe guidance, read 2026-11-16. This documented example applies to Word's Transcribe feature.
For the invented reading-group exercise, decide whether the reviewer should receive text alone or also hear the recording. If an authorized cloud workflow is used, check those two permissions with non-sensitive practice material before offering the same handover to a client. Include this check and the agreed cleanup work when estimating your service time; do not assume that sending a document finishes the assignment.
Use the sample to choose your next step
Your practice result should show whether you can maintain accuracy, consistent labels and a useful uncertainty record. If the difficult passages remain unresolved, improve the listening setup or choose a narrower scope before offering that kind of work.
Compare remote side hustles for other work models. If you prefer checking an existing written document, read proofreading services; if you work between languages, see translation project preparation. None of these sample workflows establishes client demand or a paid outcome.