Clones each speaker's voice from a sample of the video and speaks your subtitles in English or Chinese, on your own computer. It has no built-in voices, and its model is for non-commercial use only.

What it does

Clone only

One voice, Custom voice. Each speaker gets their own, copied from a sample Subtitld cuts from the video. It works best with 6 to 15 seconds.

Transcribes the sample

When a sample comes without its transcript, F5-TTS writes one with Whisper. The first time, that means downloading Whisper too; after that it is cached.

English and Chinese

The model was trained on English and Chinese speech. It has no language setting and reads each line as written.

Loads once

The first line waits for the model to download and load; the rest reuse it until the add-on has been idle for five minutes.

Non-commercial model. The F5-TTS weights are under CC-BY-NC-4.0, which doesn't permit commercial use. The add-on's code is MIT, but that doesn't change the weights' terms. For commercial work, Qwen3-TTS also clones voices, and its weights are Apache-2.0.

Voices

In the Speakers tab, choose F5-TTS as a speaker's Dubbing engine; its only Voice is Custom voice. Each line comes out as long as the model makes it; turn on Fit dub to subtitle duration for the speaker to stretch or squeeze it to its subtitle.

1 voice.

The sample for the clone

For each speaker, Subtitld cuts a sample from the project's audio:

their longest subtitle between 3 and 12 seconds;
if none fits, the first 12 seconds of a longer one;
if all are shorter than 3 seconds, several in a row, joined until they reach 3.

F5-TTS works best with 6 to 15 seconds, so give each speaker at least one line between 6 and 12 seconds. If you have separated music from voice for this video, the sample comes from the voice alone. Default voice reference in the add-on's settings is used only when no sample can be cut, for example with no video open.

Subtitld keeps each speaker's sample and cuts it again only if the audio file changes. After you retime a speaker's lines, delete their sample to get a fresh one; its name ends with the speaker's name.

System	Folder
Linux	`~/.cache/subtitld/clone-refs`
macOS	`~/Library/Caches/subtitld/clone-refs`
Windows	`%LOCALAPPDATA%\subtitld\subtitld\Cache\clone-refs`

Models

Both download the first time you generate a line: the F5-TTS model, and the Vocos vocoder that turns its output into sound. Whisper comes on top, the first time F5-TTS has to transcribe a sample.

Model	Download
f5-tts-v1-base	1.5 GB
vocos-mel-24khz	80 MB

Options

The add-on reads these when it starts, so a change applies once it has been idle for five minutes, or after you restart Subtitld. Skip auto-extracted reference transcript is the exception: Subtitld reads it for every line.

In Settings › Add-ons › F5-TTS (offline + voice clone) › Configure.

Inference device

Default: cpu

F5-TTS picks the device itself when it loads; the add-on doesn't pass this setting on. The released builds are made with PyTorch's CPU package.

Default voice reference (6-15 s, mono WAV recommended)

Default: empty

A clip to copy when Subtitld can't take a sample of the speaker from the video, for example with no video open. A sample from the video always comes first. Aim for 6 to 15 seconds.

Reference clip transcript

Default: empty

The words spoken in the default clip above. The add-on uses it with any sample that arrives without its own transcript, samples from the video included, so leave it empty unless you rely on the default clip. Empty, F5-TTS transcribes each sample itself.

Skip auto-extracted reference transcript

Default: off

Turn it on to keep the subtitle text from going with the sample as its transcript; F5-TTS then transcribes the sample itself with Whisper. Useful when your subtitles paraphrase what is said.

Languages

F5-TTS doesn't take a language: it reads each line as written. Models for other languages exist, but this add-on doesn't load them.

en
zh-cn

Platforms

System	Download	Tested
Linux x86-64	647.8 MB	Built and started in CI
macOS (universal)	433.8 MB	Built and started in CI
Windows x86-64	433.2 MB	Built and started in CI

Each release is built on Linux, macOS and Windows, and CI checks that the program starts and lists its voice. CI doesn't generate speech.

Troubleshooting

bad_paramsThere is nothing to work with: the subtitle has no text, or there is no sample to copy. Subtitld couldn't cut one (no video open, or no subtitles for that speaker) and no default clip is set, or the default clip's file has moved. Give the subtitle text, open the video, or set Default voice reference to a file that exists.
internalThe model failed to load or to speak the line; the message says which. A failed download on first use ends here too: check your connection and free disk space, and try again. If the message mentions ffmpeg, install ffmpeg so it is on your PATH; F5-TTS needs it to read the sample.

License

Add-on

MIT

Engine

MIT

f5-tts

Models

CC-BY-NC-4.0

F5TTS_v1_Base weights; non-commercial only

Not for commercial work: the weights' CC-BY-NC-4.0 license forbids it, even though the add-on's code is MIT.

F5-TTS (offline + voice clone)

What it does

Voices

The sample for the clone

Models

Options

Languages

Platforms

Troubleshooting

License

Versions

Other Voices add-ons