Dubs your subtitles on your own computer with Supertonic, a 99M-parameter ONNX model from Supertone. Its ten preset voices can each speak all 31 of its languages, and it runs on the CPU.

What it does

A voice per speaker

Choose Supertonic as the Dubbing engine and pick a voice for each speaker. Generate all speeches dubs that speaker's lines: all of them, a range, or the selection.

Any voice, any language

The voices aren't tied to a language. Supertonic reads each line in your project's language, or in a language-agnostic mode when it doesn't know that language.

Downloads once

The model downloads from Hugging Face the first time you dub: 200 to 400 MB, depending on the model you choose. After that it loads from disk.

Fits the subtitle

With Fit dub to subtitle duration on, each new clip is sped up or slowed down to last as long as its subtitle. The original clip is kept too.

Voices

Five male and five female presets, M1 to M5 and F1 to F5. Each keeps its character in every language; only the pronunciation changes. Try a line with a few of them before you dub a whole speaker.

10 voices.

Models

Keep the default, supertonic-3, unless your subtitles are only in English, Korean, Spanish, Portuguese or French. supertonic-2 covers those five with a smaller download (about 250 MB), and supertonic reads English only (about 200 MB). Change it with the Model option; the table lists the default.

Model	Download	Good for
supertonic-3	400 MB	31 languages, other text read language-agnostic

Where models are kept

In your home folder's .cache, in a folder named after the model: ~/.cache/supertonic3 for the default. That's outside the add-on, so updating or uninstalling the add-on keeps it. To keep models elsewhere, set SUPERTONIC_CACHE_DIR before you start Subtitld.

Options

Subtitld hands these settings to the add-on when it starts it. After a change, they apply once the add-on restarts: after five minutes without use, or when you restart Subtitld.

In Settings › Add-ons › Supertonic (on-device, ONNX, 99M) › Configure.

Model

Default: supertonic-3

Which of the three models to use (see Models above). Each one downloads the first time it's used.

Inference device

Default: cpu

The add-on is built with the standard onnxruntime, not onnxruntime-gpu, so CUDA can't use your graphics card. CPU works everywhere.

Quality (sampling steps)

Default: 8

More steps give cleaner audio and take longer: 5 for quick previews, 8 by default, 12 at most.

Language hint

Default: auto

This version of the add-on doesn't read this setting, so it always works as Auto: your project's language, or language-agnostic when Supertonic doesn't know it.

Custom voice JSON (optional)

Default: empty

A voice made with Supertone's Voice Builder. The add-on offers it as an eleventh voice, but Subtitld's Voice list shows only the ten presets, so you can't pick it yet.

Languages

Subtitld sends your project's language, and Supertonic uses its two-letter code. A project with no language, or with one Supertonic doesn't know, is read in language-agnostic mode (na). These are the default model's languages.

ar
bg
hr
cs
da
nl
en
et
fi
fr
de
el
hi
hu
id
it
ja
ko
lv
lt
pl
pt
ro
ru
sk
sl
es
sv
tr
uk
vi
na

Platforms

System	Download	Tested
Linux x86-64	78.1 MB	Start-up check in CI
macOS (universal)	49.1 MB	Start-up check in CI
Windows x86-64	45.6 MB	Start-up check in CI

Before every release, CI checks that each build starts and offers its ten voices. It doesn't synthesise speech there.

Troubleshooting

When a line fails, it's left without a dub. These are the reasons the add-on gives.

internalSupertonic couldn't load, or couldn't speak the line. The first line downloads the model (200 to 400 MB), so check your connection and your free disk space, then try again.
unsupported_voiceThe voice saved for this line isn't a Supertonic voice. Pick one from the Voice list again.
bad_paramsThe subtitle has no text to speak.

License

Add-on

MIT

Engine

MIT

supertonic (Python package)

Model

OpenRAIL-M

Supertonic-3 weights

Before you use the audio commercially, read the OpenRAIL-M terms on the Supertonic-3 model card; the add-on doesn't state the licenses of the two other models.

Supertonic (on-device, ONNX, 99M)