VideoText workflow guide
A practical formatting reference for Scribie transcribers. Covers the mandatory [MM:SS] timestamp format, speaker label syntax without brackets, full verbatim requirements, and every notation you'll need.
Choose Rev, GoTranscript, TranscribeMe, Scribie, or a custom client specification. Each guideline has distinct rules for verbatim level, speaker label format, timestamp intervals, inaudible notation, and maximum paragraph length.
Clean verbatim removes fillers, false starts, and repetitions for readability. Full verbatim preserves all spoken content. Applying the wrong level is the most common single reason transcripts fail marketplace QA — it cannot be fixed without re-reading the source audio.
A single speaker must have exactly one label format from the first occurrence to the last. Mixed formats (JOHN SMITH / John / J. Smith) require a find-and-replace pass across the full document before any other formatting work.
Rev-style: timestamps every 2 minutes or at each speaker change. GoTranscript: job-dependent — bold [00:00:00] either every 2 minutes or at every speaker change, with no timestamping on the qualification test. TranscribeMe: per-speaker-turn timestamps. Paragraph limits range from 8 lines (Rev) to roughly 500 symbols (GoTranscript).
Verify inaudible notation consistency (all [inaudible] or all [INAUDIBLE], never mixed), check bracket format for crosstalk sections, confirm verbatim level is consistent throughout, and validate that paragraph length does not exceed the client's maximum.
Clean verbatim text. Speaker names in CAPITAL LETTERS followed by a colon. Timestamps in [HH:MM:SS] format at 2-minute intervals and at speaker changes. No paragraph longer than 8 lines. Inaudible sections marked [inaudible]. False starts and fillers removed.
Clean verbatim on most jobs, full verbatim where the job says so. Bold speaker labels followed by a colon and a single space. [inaudible 00:00:00] and [unintelligible 00:00:00] for unclear audio, [crosstalk] for overlapping speech. Timestamps only where the job calls for them, in bold [00:00:00]. Paragraph breaks at topic shifts, under roughly 500 symbols.
Structured document with consistent heading-style speaker labels, paragraph breaks, correctly formatted timestamps, and clean or full verbatim as specified. Formatted for direct delivery — no manual cleanup required before attaching to the client email.
Reduce revision risk before submitting marketplace jobs. A pre-delivery formatting check catches the label inconsistency, timestamp format error, or verbatim level mismatch that triggers a QA rejection and unpaid revision.
Create a consistent formatting baseline across a team of transcriptionists working on the same client account — so reviewers spend time on content accuracy, not fixing label formats.
Turn raw AI-generated transcripts into readable interview documents: speaker structure, consistent punctuation, accurate paragraph breaks, and timestamps that let readers verify context in the source recording.
Page 1: "JOHN SMITH: Thanks for joining us." Page 3: "John: So as I was saying..." Page 7: "Speaker 1: Right, exactly." — All three refer to the same speaker. Label drift of this kind fails QA at every major transcript marketplace.
[00:05:12] on page 2, (00:05:12) on page 4, [5:12] on page 6, and 00:05:12 on page 8 — all in the same document. Most style guides require a single format, and inconsistency triggers automatic rejection on automated QA systems.
"[inaudible]", "[INAUDIBLE]", "[unclear]", and "[crosstalk]" appearing in the same file for the same type of audio problem. Rev requires "[inaudible]" in lowercase brackets; GoTranscript requires a full timestamp inside the tag and keeps [crosstalk] separate. Neither accepts a mix of styles.
Pages 1–4 are clean verbatim (fillers removed). Pages 5–8 switch to full verbatim (fillers preserved). This happens when an ASR tool applies different post-processing across transcript segments — and it fails QA because the verbatim level is supposed to be consistent throughout.
Clean verbatim by default. Speaker names in ALL CAPS followed by a colon. Timestamp format [HH:MM:SS] every 2 minutes and at speaker changes. Maximum paragraph: 8 lines. Inaudible: [inaudible]. Crosstalk: [crosstalk]. False starts and filler words removed.
Clean verbatim by default, full verbatim where the job specifies it. Bold speaker label, colon, single space (Speaker 1:, Interviewer:), with a ? prefix when the speaker is uncertain. [inaudible 00:00:00] for speech that cannot be heard and [unintelligible 00:00:00] for speech heard but not understood; [crosstalk] for overlap. Timestamps are job-dependent: bold [00:00:00] every 2 minutes or at every speaker change. Paragraphs under roughly 500 symbols. Spell out zero through nine.
Strict verbatim — all fillers and false starts preserved. Timestamps required at each speaker turn. Speaker format "SPEAKER NAME:" in caps. Specific bracket format for technical notation. Paragraph breaks only at speaker changes.
Many agencies specify hybrid rules that don't map cleanly to any standard guide — for example, Rev-style speaker labels with GoTranscript-style clean verbatim and no timestamps at all. These combinations must be documented explicitly because the formatter cannot infer them.
Each Scribie paragraph starts with [MM:SS] Speaker N: — timestamp in square brackets, then speaker label with colon, then the text. New paragraph for every speaker change and for every 4–6 lines within a single speaker turn. Maximum 4–6 lines per paragraph.
Use [inaudible] — square brackets, lowercase, no timestamp inside the notation itself (unlike GoTranscript). Place it exactly where the inaudible word(s) occur. The paragraph-level timestamp tells the client approximately when it happens.
No. The verbatim level is set per job by the client. If the job is full verbatim, all text must be full verbatim. If it's non-verbatim (clean), all text must be clean. You cannot mix approaches within a single transcript.
In Scribie full verbatim, false starts are included with an em dash: "I — I was thinking about" or with a comma: "I, I was thinking about." Both are acceptable. The key is capturing the false start exactly as it occurred in the audio.
Use [music], [background noise], [applause], or [laughter] — square brackets, lowercase. Include sound effects when they are significant to understanding the content. Minor consistent background noise is not marked for every instance, only when it first appears or changes significantly.