Transcript
Allowed formats: .txt, .srt, .vtt, .docx
Or drop a file here (same formats as above)
Everything we format is taken from this box — upload fills it for you; you can edit before running Step 3.
Style guide
Step 3
Run formatting
We'll apply your rules and return a review-ready diff and flagged sections.
This page formats an existing transcript to client style rules — Rev, GoTranscript, TranscribeMe, or Scribie presets you can edit. It does not transcribe video. Paste text or jump from Video to Transcript, pick a preset, and export client-ready copy. Files are deleted after processing. Free: tool-specific free access, no credit card.
Professional transcription agencies and marketplace platforms use a structured QA pipeline before accepting deliverables. This tool automates stages 2 through 5 — the most time-consuming and error-prone steps for freelancers.
Raw transcript produced from audio — either AI-generated (Whisper, Otter, etc.) or a rough human first pass. At this stage the text is accurate but unformatted: speaker labels are inconsistent, filler words are present, punctuation is uneven.
Apply the client style guide: normalize speaker labels to the required format ([Speaker Name]: vs SPEAKER 1: vs Speaker 1:), correct number formatting (spell out one–nine, numerals for 10+), apply capitalization rules, normalize ellipsis and dash usage. VideoText automates this pass entirely based on the selected preset.
Remove or retain filler words (um, uh, like, you know) based on the platform spec. Non-verbatim platforms (Rev, GoTranscript, Scribie) require removal unless fillers are meaningful. Full verbatim platforms retain every word. False starts and stutters are cleaned in this stage. VideoText applies the correct ruleset automatically per preset.
Every speaker transition must be labeled. No speaker label should appear inconsistently — mixing "Speaker 1" and "SPK 1" for the same person is a common rejection reason on Rev and GoTranscript. VideoText flags all speaker label inconsistencies in the QA compliance report before export.
When a GoTranscript job requires timestamping, stamps are every 2 minutes or at every speaker change (per job instructions). TranscribeMe requires them every paragraph. VideoText validates that required timestamps are present, correctly formatted ([00:02:00] not 00:02:00), and fall on speaker turns rather than mid-sentence. Missing or malformed timestamps are flagged before export.
A human reviewer reads through the formatted transcript against the original audio. Focus areas: proper nouns the AI may have misheard, inaudible sections that need [inaudible 00:00:00] or [unintelligible 00:00:00] tags, and any segments flagged as below the QA confidence threshold. VideoText surfaces exactly which segments need manual review — no need to scan the whole document.
Export in the required format: plain TXT for most marketplaces, DOCX for agency clients, SRT for caption deliverables. The QA compliance score documents that style guide rules were applied systematically — useful for agencies that need to demonstrate process to clients.
These are realistic examples of what a raw AI-generated transcript looks like, what it looks like after formatting, and what the final client-ready output contains. Every transformation shown here is applied automatically.
Raw AI output
spk_0: um so like the the thing that we uh found was that you know the results were uh really quite significant i mean we we ran it three times spk_1: yeah yeah and and the second run was i think more more compelling right
After formatting pass
[Speaker 1]: The thing that we found was that the results were really quite significant. We ran it three times. [Speaker 2]: Yeah, and the second run was more compelling, right?
Client-ready (Rev spec)
[Speaker 1]: The thing we found was that the results were really quite significant. We ran it three times. [Speaker 2]: Yeah, and the second run was more compelling, right?
✓ Fillers removed ✓ Labels formatted ✓ False starts cleaned
Raw AI output
Speaker 1: so we launched in Q3 and uh the response was better than we thought [00:02:14] Speaker 1: the first 30 days we had 4,000 signups which is uh which is above projections
After formatting pass
Speaker 1: We launched in Q3 and the response was better than we thought. [00:02:14] Speaker 1: The first 30 days we had 4,000 signups, which is above projections.
Client-ready (GoTranscript spec)
Speaker 1: We launched in Q3 and the response was better than we thought. [00:02:14] Speaker 1: The first 30 days we had 4,000 signups, which is above projections.
✓ Fillers removed ✓ Timestamps validated ✓ Numerals correct
Raw AI output
spk_0: I- I wasn't at the location on that date. I mean, uh, I was definitely not there.
Full verbatim output
WITNESS: I-- I wasn't at the location on that date. I mean, uh, I was definitely not there.
✓ Fillers retained ✓ False starts formatted ✓ Label normalized
Every platform has a distinct style guide. Submitting to the wrong format is the leading cause of transcript rejection and reduced QA scores. This table documents the key formatting rules for each platform as of 2026.
| Rule | Rev | GoTranscript | TranscribeMe | Scribie |
|---|---|---|---|---|
| Verbatim type | Non-verbatim (clean) | Non-verbatim (clean) | Non-verbatim (clean) | Non-verbatim (clean) |
| Filler words (um, uh) | Remove | Remove | Remove | Remove |
| False starts | Remove | Remove | Remove | Remove |
| Speaker label format | [Speaker Name]: | Speaker 1: | [Speaker Name] | Speaker 1: |
| Timestamps | Not required | Every 2 minutes | Every paragraph | Not required |
| Timestamp format | N/A | [00:02:00] | [00:01:30] | N/A |
| Numbers 1–9 | Spell out | Spell out | Numerals | Spell out |
| Numbers 10+ | Numerals | Numerals | Numerals | Numerals |
| Inaudible sections | [inaudible] | [inaudible] | [inaudible] | [inaudible] |
| Crosstalk / overlap | [crosstalk] | [crosstalk] | [crosstalk] | [crosstalk] |
| Profanity | Transcribe as spoken | Transcribe as spoken | Bleep optional | Transcribe as spoken |
| QA strictness | High — automated scoring | Medium — manual review | High — automated scoring | Medium — manual review |
| Typical rejection cause | Retained fillers, wrong label format | Missing timestamps | Numeral inconsistency | Retained fillers |
Style guide details verified against platform documentation. Rules may be updated by platforms without notice — always check the current platform guidelines for final submissions.
The QA compliance score is not a spell-checker. It runs a structured checklist against the selected style guide, detects specific rule violations, and surfaces exactly which segments need manual review — not the entire document.
The validator scans every speaker-attributed line and checks that: (1) the label format matches the selected preset exactly, (2) the same speaker is not referred to by two different label strings, and (3) no speaker turn is missing a label. Inconsistent labels — the most common QA rejection reason on Rev and GoTranscript — are flagged individually with the problematic label shown.
For non-verbatim presets, the engine scans for retained filler words (um, uh, like, you know, kind of, sort of) and flags any that were not removed during the formatting pass. Fillers inside quoted speech or technical terms are excluded from flagging. Each flagged filler is linked to the segment it appears in for one-click navigation.
For formats that require timestamps (GoTranscript: every 2 min or speaker change when job requires; TranscribeMe: every paragraph), the validator checks that: timestamps are present at required intervals, the format matches the platform spec exactly ([HH:MM:SS] vs HH:MM:SS), and timestamps fall on speaker turn boundaries rather than mid-sentence. Missing or misplaced timestamps are flagged by position in the transcript.
Some editing decisions create conflicts: a speaker label that appears in both short and long form within the same document, a number rendered as both a word and a numeral within a paragraph, or punctuation that violates the selected preset. The validator surfaces these as "needs review" flags rather than treating them as outright errors, since some conflicts are intentional.
When the input is a subtitle file (SRT or VTT), the validator additionally checks: no subtitle cue exceeds 42 characters per line, no more than 2 lines per cue, no overlapping timestamps between adjacent cues, and correct decimal separator in timestamps (comma for SRT, period for VTT). Subtitle-specific failures are surfaced separately from transcript-level formatting issues.
The automated check score summarizes formatting checks and detected preservation discrepancies. Even 100% does not verify meaning or readiness for client delivery. Compare the original and formatted transcript; changed numbers, negation, names or missing spans require review before export.
If you work on Rev, GoTranscript, TranscribeMe, or Scribie — or deliver transcripts for agencies and direct clients — this tool was designed around your specific workflow problems.
Experienced transcriptionists spend 20–40% of their working time on formatting tasks that have nothing to do with listening accuracy: normalizing speaker labels, removing fillers, fixing number formats, adding timestamps. These are mechanical rules applied identically to every document. This tool handles all of them in a single pass so you can focus on the parts that require human judgment — catching mishears, marking inaudibles, handling ambiguous speech.
A rejected submission on Rev or GoTranscript does not just cost the time to fix and resubmit. It impacts your quality score, which affects job availability and pay rates on the platform. The most common rejection causes — retained filler words, wrong speaker label format, missing timestamps — are all checkable before submission. The QA compliance score tells you whether these issues exist before you click submit.
Agency clients and direct clients often have their own style guides that differ from marketplace standards. Some clients want timestamps every paragraph. Some require INTERVIEWER: / RESPONDENT: rather than speaker numbers. Some have custom verbatim rules for specific terminology. Upload your client's style guide as a PDF, DOCX, or TXT file and the tool will parse it into editable rule cards. Adjust any card before running the formatter.
The highest-output professional transcription workflow combines AI first-pass transcription with style guide formatting: (1) transcribe audio with VideoText — the full transcript, including speaker labels, is generated in minutes; (2) paste the output into this formatter; (3) select your platform preset; (4) review flagged segments only. This workflow cuts total time-per-file by 60–80% compared to transcribing manually from scratch, while keeping human review in the loop for the segments that need it.
If you deliver SRT or VTT files alongside or instead of plain transcripts — for captioning clients, media companies, or platforms — this tool handles the subtitle formatting pass as well. Paste your SRT or VTT content directly: the formatter preserves timestamp structure, applies verbatim rules to caption text, validates line length against broadcast standards, and exports in caption-ready format. The QA report includes subtitle-specific checks for overlapping timestamps and character limits.
Start with the guides that already get search traffic — GoTranscript, Scribie, Verbit, TranscribeMe, and Rev — then format your file to those rules in one click on this page.
