Clean Verbatim Transcription: Rules, Examples and Practice

A transcript can look neat and still be wrong. Removing a hesitation is one decision; changing what the person meant is another. Clean verbatim transcription asks you to make limited, defined edits, and those limits come from the assignment's written guideline rather than your instinct for polished prose.

This guide explains the distinction through original examples, including the same fictional utterance in full and clean styles. It is written for contributors doing the work, not for buyers choosing a transcription service. The examples use explicitly stated practice rules. A real project may make different choices about repetitions, corrections, contractions, sound events and punctuation.

What clean verbatim does and does not mean

Clean verbatim generally aims to make speech easier to read by removing selected disfluencies. It should still represent the speech rather than become a summary, translation or rewritten argument. Full verbatim retains more of the delivery. The precise treatment varies: a guide may distinguish filler sounds from conversational words, or preserve a false start because it adds information.

Rev's explanation of clean verbatim describes lightly editing a transcript for readability without paraphrasing. Its published style reference separates verbatim and non-verbatim expectations. These are examples of a provider's rules, not a replacement for the guideline attached to your task.

Before typing, decide which instructions govern this particular file. If the brief says clean verbatim but its examples keep hesitations, do not silently choose a different convention. Check for a project-specific instruction, updated rule or clarification route. Record the answer in your own checklist using the task's rule names.

The same utterance in raw, full and clean form

The following line is an invented script, not a recording or a real contributor test. Our practice convention removes “um” and immediate accidental word repetitions, but keeps a correction and the speaker's original contractions. The full version preserves every scripted spoken word; the clean version follows only the editing permissions just stated.

The raw script supplies the fictional words. Its punctuation is a reading aid, not something you can hear directly. A real transcriber must decide punctuation from speech and the applicable rules. The full and raw rows match here because the script contains no additional sound events to represent.

Under our stated clean convention, “Thursday. No, Friday” remains because the correction is part of the information. “It's for Lina, not Lena” also remains. Do not replace the whole passage with “I sent Lina the folder on Friday.” That sentence summarises and rearranges the speaker's wording. Another project may specify a different correction policy, so check rather than treating this example as a universal answer.

Five decisions that need a written rule

First, check fillers. In “Um, the parcel is outside,” our practice rule allows removal of “um.” In “It looks like the smaller one,” “like” expresses a comparison and must stay. A find-and-delete operation on every instance of a word cannot make that distinction.

Second, check repetition. The repeated “I” in our main example is an accidental restart under the practice convention. In “It was very, very close,” the repetition can express emphasis. Do not flatten emphasis automatically. GoTranscript's published guidelines, for example, distinguish emphatic repetition from the repetitions removed in its clean style. That is a useful reminder to read the rule in context.

Third, check contractions and informal speech. Do not assume clean means formal. “I don't know” does not become “I do not know” unless the rule permits that change. Keep the speaker's grammar rather than correcting “We was outside” to “We were outside.” A spelling policy and a grammar-rewriting policy are different instructions.

Fourth, check responses and sound events. A short “yes” may answer a question; a similar sound may be an acknowledgement while someone else continues. A laugh may matter under one scheme and be omitted under another. Follow the task's definitions and notation rather than inventing a tag or deleting every short response.

Fifth, check numbers, names and punctuation. A brief might request digits, words or a specific date format. Apply that formatting without changing the value. Research a supplied term only through permitted sources; context may help identify it, but a plausible spelling is not evidence that the word was spoken.

Handle unclear words without inventing an answer

Replay the whole phrase before isolating a difficult word. Check whether the issue is the source, your output device or an unfamiliar term. A short rewind and a permitted speed adjustment can help you examine the sound. A second pair of available headphones may help identify a playback fault, but no device can establish a missing word from context alone.

Imagine the fictional line “Put it by the [unclear] door.” If the key word is unresolved, do not decide it must be “blue” because the earlier sentence mentioned a blue folder. Use the project's approved uncertainty notation and time position if required. The bracketed placeholder here explains the problem; it is not a tag to copy into a real assignment.

Some schemes distinguish speech masked by noise from audible speech you cannot understand. Others use one tag. Do not import a provider's label into another project's transcript. If the audio cannot support the requested deliverable, follow the task's escalation or rejection process. Also avoid deleting an unclear fragment merely to make the clean version read smoothly.

Our audio quality checker inspects technical properties of a local file. It cannot identify the missing word or judge whether the transcript is correct. For playback preparation, use the transcription equipment checklist before changing the source audio or buying anything.

Build a task-specific transcription rules checklist

Write a short working sheet before starting a batch. Include the guideline version or date so that a later update does not leave you applying an old rule. The following questions are a structure for your notes; answers must come from the actual project.

1. Which words stay? Record filler, repetition, false-start, correction and contraction rules with one allowed example for each.
2. Which labels are required? Note speaker naming, uncertainty, language changes and sound-event conventions.
3. How is text formatted? Check numbers, punctuation, paragraphing and whether any time references are required.
4. Which tools are permitted? Confirm whether automatic transcripts, external research, downloading and local storage are allowed.
5. What is the final review? Record completeness checks, the submission format and the official route for a question you cannot resolve.

Download the plain-text transcription style worksheet to practise these decisions. It contains only original fictional text and blank fields. Do not put confidential client excerpts into a personal worksheet if the project does not permit local notes or storage.

Practise, compare and review in separate passes

Start with our free transcription practice exercises. Read the exercise-specific rule before listening. Make a first attempt without opening the reference, then compare the result. Its automated word comparison is useful feedback for that reference; it does not certify compliance with every real project's style or evaluate a contributor for work.

Review in two passes. First, listen against the text for missing words, wrong words, speaker attribution and altered meaning. Then check the requested style: fillers, repetitions, labels, numbers and formatting. Separating the passes makes it easier to see whether a mistake came from hearing the audio or applying the rule.

Keep an error note such as “removed a meaningful like” or “changed a spoken contraction,” using invented material if the real task restricts notes. Practise that one decision again instead of simply retyping the full reference until the score rises. If you consult the answer first, label the next attempt as guided practice rather than an independent accuracy result.

Apply the method to a real project

Read the brief, eligibility, consent and review conditions before accepting a task. A transcription minute describes the recording's duration, not the time needed to listen, type, research permitted terms and correct the work. Use the freelance task rate calculator with your own payout, total time and applicable fees when evaluating an offer.

Joining Spirelight is free. Projects open in waves, and matching depends on active requirements such as language, location and equipment. Practice does not guarantee qualification, an invitation or income. Read how speech AI projects work and use the language and country directory for relevant guidance. The actual invitation supplies the writing and payment terms that govern the work.