Try Pro FREE for 2 weeksGet Started
Start for free

audio-take-qa — Traditional AI Model

Take QA listens to a generated voice take together with its transcript and lists the faults that read fine as text: a syllable said twice ("다, 다") that speech recognition merges into one word, a stray sound glued to the start of the take, a pause in the middle of a word, and words that differ from the script you expected. Run it before you spend more steps on a take. CPU.

Try in Studio

Details

  • CategoryTraditional
  • Year-1
  • License

Inputs & Outputs

TranscriptInputtranscript

Speech Recognition output for the take

SoundInputsoundoptional

The take's audio. Connect this or Video.

VideoInputvideooptional

A video carrying the take's audio. Connect this or Sound; if both are connected the sound is used.

ReportOutputstring

JSON report: ok, issues with times, and a one-line summary

Parameters

expected_scriptstringdefault

The text the take should say (optional). Differences are reported.

Optional script the take was meant to say. Words are compared after lower-casing and stripping punctuation; missing, extra and replaced words are reported as script_mismatch. Leave blank to skip. Max 5000 characters.

speech_dbfloatdefault -38range -80–0advanced

Level (dBFS) above which a 10 ms frame counts as voice.

Loudness threshold separating voice from room tone, in dBFS on a 10 ms RMS envelope (-80..0). Raise it for a noisy take whose room tone exceeds -38 dBFS.

peak_prominence_dbfloatdefault 4range 0.5–30advanced

How much a syllable peak must stand out (dB).

Prominence a loudness peak needs to count as a syllable (0.5-30 dB). Lower finds more repeats and more false alarms.

min_inner_pause_secfloatdefault 0.25range 0.05–5advanced

Shortest silence inside a word that is reported (s).

Silences inside one recognised word shorter than this are ignored (0.05-5 s).

lead_fragment_max_secfloatdefault 0.6range 0–5advanced

A pause this close to the start of the first word marks a lead fragment (s).

A pause inside the first word that starts within this many seconds of it is reported as lead_fragment rather than inner_pause (0-5 s).

Tags

  • audio
  • voice
  • tts
  • qa
  • transcript
  • repeat
  • stutter
  • veo

Alternatives in Video Editing

  • Audio Edit

    Trim a sound, fit it to a video or a length, dip or mute parts, and fade it.

  • Audio Finish

    Normalize a video's loudness to a broadcast target (2-pass loudnorm + peak limiter).

  • Audio Mix

    Mix narration, music and sound effects into one track, with ducking and loudness.

  • Audio Replace / Mix

    Replace or mix a clip's audio with a new sound, keeping the video as is.

  • Chroma Compose

    Cut a green-screen person out of one video and place them over another.

  • Hook Teaser (Cold Open)

    Prepend a cold-open teaser of each clip's hook moment to the front.

  • Ken Burns (Still to Video)

    Animate a still image into a video with a moving crop window.

  • Narration Mux (Cold Open)

    Prepend a voiceover cold-open (frozen first frame + narration) to a clip.

All AI models