For macOS
Give every idea its own voice
AudioColor is a local AI voice studio for Mac. Go from a few lines of text to a finished multi-character production — then send it straight into Final Cut, Resolve or Premiere.
Requires macOS 13.3 or later · Speech is synthesized on your Mac
Runs locally
Text and audio are never uploaded to synthesize
7 languages
Chinese, English, Japanese, Korean, French, German, Spanish
8 export formats
WAV, SRT, CSV, markers, FCPXML, Premiere XML and more
Four ways to make a voice
Each one feeds the same shared voice library, so anything you make in one place is available everywhere else.
01
Natural text to speech
Type or paste text, pick a voice, language, speed and tone, then render clear, natural speech. Long passages split automatically and queue in the background with visible progress.
02
Multi-character script dubbing
Set the speaker, tone, speed, language and target duration for every single line, then render the whole script in one batch. Paste text or import TXT, Markdown, DOCX and EPUB.
03
Design custom voices
Describe the age, texture, accent, character and intended use in plain language. AudioColor creates a new voice, saves it to your library, and makes it available everywhere.
04
Clone with permission
Record or import 10–30 seconds of clear speech to create a reusable voice and preview it immediately. Only clone your own voice, or one you have explicit permission to use.
From a document to a finished timeline
AudioColor is built around the way voice work actually gets done — long text, many characters, and an editor waiting at the end of it.
- 1
Import the document
Paste text, or bring in TXT, Markdown, DOCX and EPUB. Everything is normalized to Markdown on disk so long books stay workable.
- 2
Cast the characters
Optional automatic casting reads the source, proposes a cast, and assigns a designed voice to each character before writing the script.
- 3
Tune line by line
Every line carries its own speed, language, tone and target duration. Set a segment longer than the audio and choose where the voice sits inside it.
- 4
Render in the background
Jobs queue with a concurrency limit you control. Long runs checkpoint their progress, so an interrupted render resumes instead of restarting.
- 5
Export to your editor
Numbered WAVs, a full mix, subtitles, markers and a ready-made timeline — dropped into Final Cut, Resolve or Premiere with the cuts already in place.
Built for editing workflows
One render, every file your timeline needs. Numbered clips stay in sync with the subtitles, markers and timing sheet that ship alongside them.
Numbered WAV clips
One file per line, in order
Full mix
The whole render as a single track
SRT subtitles
Timed against the render
CSV timing sheet
Every clip with in and out points
Audition markers
Drop straight into a session
FCPXML timeline
Final Cut Pro and DaVinci Resolve
Premiere XML
XMEML timeline for Premiere Pro
Voice library
Tags, favorites and search across every voice
Local synthesis, clear boundaries
Speech is generated on your own machine. Nothing about the way AudioColor works requires you to hand your script to a server.
Synthesis happens on your Mac
Your text and the audio it produces are not uploaded in order to be synthesized. The voice model runs on your own hardware.
Models download once
The first time you use a voice model, AudioColor downloads it from Hugging Face or ModelScope. After that it works offline.
Automatic casting is opt-in
If — and only if — you turn on automatic casting, the relevant document text is sent to the model provider you configured yourself: Anthropic, OpenAI or DeepSeek.
No analytics, no accounts
There is no telemetry, no tracking SDK and no sign-up. Projects, tasks and voices live in a database on your machine.
Questions
Does AudioColor work offline?
Yes, once the voice models are downloaded. The first run needs a connection to fetch the model files, and the optional automatic casting feature needs one because it calls a third-party model provider. Everything else — generating speech, designing voices, cloning, exporting — runs with no network at all.
Whose voice am I allowed to clone?
Your own, or a voice you have explicit permission to use. Voice cloning is a real-world consent question, not just a technical one, and misusing someone's voice can be illegal where you live. AudioColor gives you the tool; the permission is on you.
Can I use the audio commercially?
The audio you generate is yours to use, including commercially. Cloned voices are the exception worth repeating: you need permission from the person whose voice it is.
Which editors does it export to?
FCPXML for Final Cut Pro and DaVinci Resolve, XMEML for Premiere Pro, plus WAV clips, a full mix, SRT subtitles, a CSV timing sheet and Audition markers that work in just about anything else.
Which languages can it speak?
Chinese, English, Japanese, Korean, French, German and Spanish. The app's own interface is available in the same seven languages and follows your system setting by default.
What do I need to run it?
A Mac running macOS 13.3 or later, and enough free storage for the voice models — they are large, so leave room before the first download.
Start with one line of text
Download AudioColor and hear your first generated voice in a couple of minutes.
One-time model download on first use.