Skip to content
VoiceQC

Catch script-transcript mismatches

This tutorial picks up from the Quickstart. You have a Session with at least one transcribed audio file. Now you’ll attach a script (the lines the actor was supposed to deliver) and let VoiceQC automatically flag the takes that don’t match.

By the end you’ll have:

  • A script CSV uploaded to your Session.
  • The Mismatch detection rule turned on.
  • One or more Session items auto-tagged with a Property when the transcription differs from the script.

Before you start

Section titled “Before you start”

A Session with audio

Have at least one WAV uploaded and transcribed in a Session. If you haven’t, do the Quickstart first.

A spreadsheet tool

Excel, Numbers, Google Sheets, or any text editor will do — you’ll write a small CSV.

Prepare the script CSV

Section titled “Prepare the script CSV”

The script CSV has two columns: the audio filename and the line of text VoiceQC should compare the transcription against.

filename,script_text
hero-line-001.wav,"This is the line the actor was supposed to deliver."
hero-line-002.wav,"I'll be back."
hero-line-003.wav,"Line that spans
multiple lines wrapped in quotes."

A few rules:

  • Header row is optional. The parser auto-detects a header if the first line contains both the words filename and script (case-insensitive).
  • Filenames must match the audio files you uploaded. Including the .wav extension is recommended but not required; if your original had special characters, use the sanitized version of the name.
  • Multi-line scripts are supported — wrap them in double quotes.
  • Empty rows are skipped. Trailing blank rows from Excel exports are ignored.
  • Both columns are required on every data row.

Save the file as .csv.

Upload the script

Section titled “Upload the script”

  1. Open the Session where your audio lives.

  2. Find the Upload Scripts CSV zone (next to the audio drop zone).

  3. Drag your .csv file onto it, or click Choose File to pick it from your file manager. (Click Format Info if you want to see the format reference without leaving the page.)

  4. The script uploads. Each Session item whose filename matches a row in the CSV gets its Script field filled in.

You can upload the script before or after the audio — the order doesn’t matter. Mismatched filenames (rows that don’t match any uploaded audio) are stored and applied if and when that audio arrives later.

Turn on the Mismatch rule

Section titled “Turn on the Mismatch rule”

Now tell VoiceQC to compare each transcription to its script and flag the ones that disagree.

  1. Open the Auto-rules panel for the Session (look for Auto-rules in the Session toolbar).

  2. Switch to the Mismatch detection tab.

  3. Toggle the rule from Disabled to Enabled at the top of the panel.

  4. Skim the Assigned Property section — this is the Property VoiceQC will apply when a mismatch is detected. The default is Status / Mismatch. Change it via Change Property if you’d prefer a different tag.

  5. Leave the Comparison Pipeline at its defaults for now (ignore case, punctuation, and whitespace; convert numbers to words off). These are sensible starting points; you can tune them later. See Configure a Mismatch rule for the full reference.

  6. Click Save Changes at the bottom right.

Run the rule against your Session

Section titled “Run the rule against your Session”

The rule is now active for new items as they arrive. To check existing items in this Session, run a one-time evaluation:

  1. In the same panel, click Re-evaluate Session at the bottom left.

  2. A toast appears in the bottom right: Evaluating session items…

  3. Wait. When evaluation completes the toast turns green: Evaluation complete! with a count of Properties added and removed.

See the results

Section titled “See the results”

Close the panel and look at the Session items table:

  • Items whose transcription matched their script: no change (or any previous Mismatch tag is removed).
  • Items whose transcription didn’t match: a new Property chip appears, set to whatever you picked in Assigned Property (default Status / Mismatch).

Click into an item to compare the script (what you uploaded) and the transcription (what VoiceQC heard) side by side.

What did you just do?

Section titled “What did you just do?”
  • Wrote a script CSV and uploaded it to your Session.
  • Turned on the Mismatch rule and accepted sensible defaults.
  • Ran a one-time re-evaluation to apply the rule to existing items.
  • Saw VoiceQC auto-tag the items where the take diverged from the script.

Where to next

Section titled “Where to next”

Tune the comparison

The Configure a Mismatch rule how-to walks through the normalization toggles, synonyms, and how to test before re-running.