Edit the recording
like it was text.

Core workflowFix a word in four clicks

1. Select the item in REAPER

   Repair transcribes it automatically (or reuses the cached transcript - selecting the same item later costs nothing). The transcript appears as clickable word chips with a search box for long material.

2. Select the words

   Left-click selects a word, Shift+click extends the range, right-click plays the word so you can confirm you have the right spot. The REAPER playhead also underlines the word it is passing through.

3. Choose the operation

   Replace swaps the selection for new text. Insert adds words at the cursor between chips. Delete removes the selection and closes the gap cleanly. Tab cycles the three modes.

   For Replace and Insert, Reasonate regenerates a few words on each side of the edit along with your new text, so the prosody flows through the splice instead of jumping. The panel shows exactly which words will be regenerated, and the estimated cost.

4. Apply and listen

   The regenerated phrase is spliced into the take at millisecond precision (text-to-audio alignment plus acoustic analysis pick the cleanest cut points, with crossfades in silences). The item is split around the edit; the new middle part carries the fix.

Whose voice?Voices and clones

The fix has to sound like the speaker. The header shows which voice will be used and where it comes from - for example Voice: Piotr Pro - from track casting. Reasonate resolves the voice in this order:

1. A voice you assigned to this speaker (see casting below),
2. a voice linked to this character in the project's Cast Registry,
3. the track's voice (as in Voice Replacement),
4. or, when there is nothing yet: Reasonate offers to train an instant clone from the item itself.

Training a clone takes seconds. For multi-speaker material a picker opens first: choose the speaker, and Reasonate pre-selects their cleanest regions (you can adjust and preview the training audio before spending anything). Re-clone retrains after you improve the source; Voice Isolator pre-cleans noisy material for both transcription and cloning.

Multiple voices in one fileSpeaker casting

When transcription detects several speakers, a banner offers Cast voices: a small table with each speaker, a 5-second sample button, a voice picker and a Train clone shortcut. Name the speakers - named characters go to the Cast Registry and follow you into Dubbing and TTS.

• Every edit automatically uses the voice cast for the speaker of the selected words.
• If the registry knows a better fit, a hint appears: "Selected words are spoken by X" with a one-click Use X's voice.
• Selected a range that spans two speakers? Repair refuses to guess: use Split at speaker boundary and fix one speaker at a time.

Fine controlPace and delivery

• Speed slider - nudge the generated phrase faster or slower before applying.
• Match pace (optional, off by default; Settings → Repair) - Reasonate measures the take's syllable rate, generates at a matching pace, and picks the better of two renders when the first misses. Natural pacing without the toggle is often already right; try the toggle on fast, dense reads.
• Voice settings - a collapsible section overrides stability and similarity for this one edit.