Multilingual Text to Speech
Tomiinek/Multilingual_Text_to_SpeechReviewed by AI
Checked by RepoPicks Radar on
What is Multilingual Text to Speech?
This is the code behind a 2020 research paper on multilingual text to speech (a Tacotron 2 model). It can train and run a single voice model for many languages, mix languages in one sentence and copy a speaker's voice. It is meant for researchers and developers: you clone the repository, install Python packages, download datasets and model files, and run scripts from the terminal. The repository is archived, so it gets no more fixes. The authors also link Google Colab notebooks and a page with audio samples.
Multilingual Text to Speech is free and open source (MIT) and can replace ElevenLabs and Murf.
How easy is it to install Multilingual Text to Speech?
E · developers only · Linux
On Linux you clone the repository, install old pinned Python packages (torch 1.13.1, librosa 0.7.0, numba 0.48), download datasets and checkpoint files by hand and run training or synthesis scripts; there is no installer and no end-user app, so it is an E. The linked Colab notebooks are the only click-and-try option, and they are from 2020.
Coming soon: install in 1 click with the RepoPicks app.
Multilingual Text to Speech on GitHub
- Stars
- 843on Nov 10, 2026
- Downloads
- 14,70019 this week
- Growth
- —no gain in the last reading
Stars over time
Facts
- Runs on
- Linux
- You need to install
- Python, Graphics card
- To build from source
- Python, Graphics card
- Disk space
- 900 MB–4.2 GB · estimate
- Accounts or keys
- None
- Cost
- Free
- License
- MIT (Permissive)
- Can replace
- ElevenLabs, Murf
- Last change on GitHub
- Oct 10, 2023
Multilingual Text to Speech risk signals (7)
- Attentionstale releases
Last release on 2020-05-26 (2359 days ago)
automatic check
- AttentionDependencies with known vulnerabilities
Pinned dependencies with known vulnerabilities (OSV): pillow: GHSA-22hg-r758-rrw9, GHSA-24rc-8hj2-v98c, GHSA-254f-3936-288h; torch: GHSA-3749-ghw9-m3mg, GHSA-53q9-r3pm-6pq6, GHSA-5pcm-hx3q-hm94
automatic check
- Attentionold pinned dependencies
requirements.txt pins 2019-2020 versions (torch==1.13.1, librosa==0.7.0, numba==0.48, tensorboard==2.0.2, phonemizer==2.1), which usually fail to install on current Python versions.
AI review
- InfoRepository archived
The repository is archived (read-only, no longer maintained)
automatic check
- Infovoice cloning without misuse notice
The repository description and topics advertise "voice cloning" (topics: voice-cloning, code-switching) and the README offers a checkpoint "supporting code-switching or voice-cloning", with no notice about consent or misuse of someone else's voice.
AI review
- Inforesearch code not an app
README.md is written around reproducing a paper ("source code, and parameters for the paper One Model, Many Languages"); the usage section starts with downloading the CSS10 and Common Voice datasets and training the model.
AI review
- Infodataset licenses vary
README.md: training needs the CSS10 dataset (Apache License 2.0) and a cleaned Common Voice set (Creative Commons CC0), downloaded separately from the code.
AI review
Evidence is shown in its original language.
What Radar could not check about Multilingual Text to Speech
- Whether the pinned 2020 packages still install on a current Python, and whether the code runs at all today.
- Whether the Google Colab demo notebooks linked in the README still work.
- How the generated voices actually sound compared with current text-to-speech services.
- Whether synthesis works at a usable speed without a graphics card.
- Real disk space needed once the datasets and the WaveRNN vocoder weights are downloaded.