About Services Tools Blog Contact us

Local Text to Speech

Turns a script into speech on the studio Mac with a choice of stock voices, and splits a long script into numbered clips so an editor can replace one line at a time.

  • Text-to-Speech
  • Cross-sector
  • Tool
  • Productised
  • Local/private
  • 2026

The problem

Reading a script aloud takes time, and sending it to an online voice service sends the text off the premises. A single long recording is also hard to fix, because one wrong line in a video edit means redoing the whole take.

What it does

You give it text, a stock voice and an optional speed or quality setting. Kokoro speaks the text on the Mac, with a British voice by default and nothing sent over the internet. It saves an audio file and prints where it is. A long script is split into numbered clips, so one line can be recorded again without redoing the rest. It uses stock voices and does not clone a person.

Outcomes

  • A narration file is made with no hosted speech service.
  • The default voice is British, and other installed stock voices are on a list.
  • An editor swaps one numbered clip and keeps the rest.
  • A short script has already been rendered to an audio file on disk.

Simplified mockup

Interactive preview with sample data. Nothing is sent to a server.

Speak a line

Sample rate
24 kHz
Channels
Mono
Network
None
on this machine
Script

One sentence for a single WAV. A longer script belongs on the Chunks tab, one clip per line.

Voice and speed

British stock voice. Used when the command does not name another.

Ready to synthesise

Ordered clips

8 lines
Clips
0
Total
0
Replaced
0
line 05 only

Each Play button plays that line and pauses the others. The files stay separate, so an edit can swap one sentence.

spring-rail.txt
#LineSecClip
01 The spring rail goes out on Thursday. 3.0 ready
02 Wide-fit court at the front of the rail. 2.8 ready
03 Navy leather, low block heel. 2.6 ready
04 Stone trainer sits one peg along. 2.9 ready
05 Hold on the size strip, six and seven. 3.0 Replaced
06 Ankle boot closes the run. 2.6 ready
07 Elastic sides, black finish. 2.6 ready
08 Cut on the street window. 2.2 ready

Installed stock voices

Names on this list are stock voices shipped with the model. Nothing here is cloned from a person.

British default in use

Default when a command omits the voice. British English, even pace.

The Speak player and the eight clips use this voice.

Narrator A installed

A second stock voice, a little brighter, for alternate reads.

Installed on the machine. Not rendered on this page.

Narrator B installed

A third stock voice, lower and slower at speed 1.00.

Installed on the machine. Not rendered on this page.

Where it runs
Apple GPU when present CPU fallback WAV path printed

A long script is split on line boundaries. Replacing line 05 writes a new clip and leaves 01 to 04 and 06 to 08 untouched.

How it works

  1. Take the text typed in, from a file or piped in.
  2. Pick an installed stock voice and a speed or quality setting.
  3. Run Kokoro on the Apple graphics chip when available, otherwise on the main processor.
  4. Save an audio file and print where it is.
  5. For a long script, split it into sensible pieces and number each clip.

Pipeline

flowchart LR
  Text[Script] --> Voice[Stock voice]
  Voice --> Synth[Kokoro]
  Synth --> Chunk[Ordered clips]
  Chunk --> Wav[WAV file]
  Wav --> Play[Playback]

Sample output

The spring rail goes out on Thursday.
Wide-fit court at the front of the rail.
Navy leather, low block heel.
Stone trainer sits one peg along.
Hold on the size strip, six and seven.
Ankle boot closes the run.
Elastic sides, black finish.
Cut on the street window.