Mode 02 - Voice Replacement

Same performance,
different voice.

Voice Replacement converts existing recordings to another voice using ElevenLabs speech-to-speech. Timing, phrasing and emotion stay; the voice changes. Perfect for recasting a narrator, anonymizing a speaker, or hearing your script in the voice you could not book.

InputAudio items on tracks
Output[AI] track per source
EngineElevenLabs STS
Batch3 conversions at a time
Source audioNever modified

Core workflowFrom recording to new voice

  1. Give the track a voice

    Every track in the table has a Pick voice... button. It opens the voice picker: search, filter, star favorites, and press ▶ to hear a preview. Double-click to assign.

    Optionally type a role for the track (like "narrator" or "guest"). Roles power Casts: save the current role-to-voice mapping once, apply it to any future project with two clicks.

  2. Select items and convert

    Select one or more audio items in REAPER (Cmd+click for multiple), then press Convert selected (N) or Cmd+Enter. A confirmation dialog shows exactly what will happen first: total audio length, estimated cost, and which voices are used.

  3. Let the batch run

    Conversions run in the background, three at a time. The interface stays responsive; progress lives in the batch dialog and in the footer activity chips. Cancel drops the queue but lets in-flight conversions finish, so you never pay for audio you do not receive.

  4. Audition and compare

    Click any item row and use the audition strip: ▶ Original vs ▶ AI. If an item has several takes, cycle them right there, or select the item in REAPER and press Tab.

Batch confirmation dialog showing total audio length, estimated API cost and the voices that will be used
Nothing is spent silently: the batch dialog shows duration, cost estimate and voices before you commit.
REAPER timeline with the source item and the converted AI item on a child track, next to the Reasonate table showing the source track and its read-only AI output track
After conversion: the AI take lands on a linked child track. The source item is untouched.
Non-destructive

Output goes to a separate [AI] track, created as a folder child of the source track (a flat layout is available in Settings). The [AI] rows in the table are read-only on purpose - they always tell you which source they belong to.

IteratingRe-convert, variants, takes

Reasonate tracks what each conversion was made from. If you convert the same item again with the same voice and settings, it is a free cache hit and gets skipped. Change the voice, the item audio or the voice settings, and it re-renders - the new result is appended as a new take, so nothing you liked is lost.

Variants

Want options? Select one item and press Variants.... Reasonate renders N takes (you choose how many) with different random seeds; each take is a slightly different read. Cycle takes with Tab and keep the winner.

Per-track voice settings

The ... menu on each track opens Voice settings: stability, similarity, style and speaker boost overrides for that track only. Useful when one voice needs a steadier read than the rest.

Your libraryFinding and managing voices

Voice picker

Wherever you assign a voice - a track's Pick voice..., a TTS speaker, a dubbing character - the same picker opens: search, gender / accent / category filters, star favorites (with a favorites-only filter), and Play previews. Double-click a name to apply it.

Pick voice modal with search, gender, accent and category filters, favorites column and Play preview per voice
One picker everywhere: search, filter, star, preview, double-click.

Two header tools keep your voice collection in shape, in every mode:

Voice Library

Library browses the public ElevenLabs collection (thousands of voices) with search and filters for gender, age, language, category and use case. Play previews a voice; Add imports it into your account, where it immediately shows up in Pick voice... everywhere.

Voice Library browser with search, filters and a paginated list of public voices with Play and Add buttons
The public library: filter, preview, import.

Voice Manager

Voice Manager lists only voices you own - clones and imports - with rename, delete and batch delete. Dubbing and Repair create clones as you work, so a periodic cleanup keeps your voice slots free.

Voice Manager listing owned cloned voices with rename and delete actions and a batch delete button
Everything you own in one list - rename clones, delete leftovers.

CaptureRecord straight into the workflow

You can record source material without leaving Reasonate. In a track's ... menu choose ● Record: the track arms, an input meter appears in the row, and an optional pre-roll counts you in. Press stop, and the fresh item is ready to convert.

Input, pre-roll length and monitoring live in Settings → Voice Replacement.

CleanupVoice Isolator

Noisy source? Two options:

  • Per track - toggle Voice Isolator in the track's ... menu. Every conversion on that track first strips background noise, then converts the cleaned audio. The row shows a small dot when active.
  • One-off - the Clean voice button in the audition strip runs the isolator on the selected item without converting anything, so you can hear the cleanup by itself.
Cost

Voice Isolator is a separate ElevenLabs call (1000 credits per minute of audio). Its results are cached like everything else, so re-running a conversion does not clean the same audio twice.

Edge casesLong items and limits

ElevenLabs speech-to-speech accepts up to 5 minutes per call. Longer items are handled for you: Reasonate splits them at silences, converts the chunks, and lays the results out as a grouped set of items that plays seamlessly. If you re-convert, finished chunks come from cache and only missing ones are rendered.

Items over 290 seconds with a modified playrate or take FX cannot be auto-chunked yet and are rejected with a clear message.

Bigger sessionsRoles, filters, colors

  • Filter bar - with several roles or six plus tracks, chips with role counts and a search box appear above the table. Filtering never hides anything silently: a caption always says "showing N of M".
  • Status colors - items are tinted by state (queued, converting, done, error...) so a glance at the timeline tells you where the batch is. Pick your own colors anytime; Reasonate respects manual colors and stops auto-tinting that item.
  • Stats strip - the row above the table counts tracks, voiced tracks and items, with status pills for anything pending or failed.