Pendrokar/open_tts_tracker

above models sorted by the amount of capabilities; #legend

Cloned the GitHub repo for easier viewing and embedding the above table as once requested by @reach-vb: https://github.com/Vaibhavs10/open-tts-tracker/issues/30#issuecomment-1946367525

---

🗣️ Open TTS Tracker

A one stop shop to track all open-access/ source Text-To-Speech (TTS) models as they come out. Feel free to make a PR for all those that aren't linked here.

This is aimed as a resource to increase awareness for these models and to make it easier for researchers, developers, and enthusiasts to stay informed about the latest advancements in the field.

This repo will only track open source/access codebase TTS models. More motivation for everyone to open-source! 🤗

Some of the models are also being battle tested at TTS arenas:

• 🏆 TTS Arena - Battle tab allows to choose 2 candidates and compare them
• 🤗🏆 TTS Spaces Arena - Uses online HuggingFace Spaces, which have Gradio API enabled

Name	GitHub	Weights	License	Fine-tune	Languages	Paper	Demo	Issues
AI4Bharat	Repo	Hub	MIT	Yes	Indic	Paper	Demo
Amphion	Repo	Hub	MIT	No	Multilingual	Paper	🤗 Space
Bark	Repo	Hub	MIT	No	Multilingual	Paper	🤗 Space
EmotiVoice	Repo	GDrive	Apache 2.0	Yes	ZH + EN	Not Available	Not Available	Separate GUI agreement
F5-TTS	Repo	Hub	MIT	Yes	ZH + EN	Paper	🤗 Space
Fish Speech	Repo	Hub	CC-BY-NC-SA 4.0	Yes	Multilingual	Not Available	🤗 Space
Glow-TTS	Repo	GDrive	MIT	Yes	English	Paper	GH Pages
GPT-SoVITS	Repo	Hub	MIT	Yes	Multilingual	Not Available	Not Available
HierSpeech++	Repo	GDrive	MIT	No	KR + EN	Paper	🤗 Space
IMS-Toucan	Repo	GH release	Apache 2.0	Yes	ALL*	Paper	🤗 Space, 🤗 Space*
MahaTTS	Repo	Hub	Apache 2.0	No	English + Indic	Not Available	Recordings, Colab
MaskGCT (Amphion)	Repo	Hub	CC-BY-NC 4.0	No	Multilingual	Paper	🤗 Space
Matcha-TTS	Repo	GDrive	MIT	Yes	English	Paper	🤗 Space	GPL-licensed phonemizer
MeloTTS	Repo	Hub	MIT	Yes	Multilingual	Not Available	🤗 Space
MetaVoice-1B	Repo	Hub	Apache 2.0	Yes	Multilingual	Not Available	🤗 Space
Neural-HMM TTS	Repo	GitHub	MIT	Yes	English	Paper	GH Pages
OpenVoice	Repo	Hub	MIT	No	Multilingual	Paper	🤗 Space
OverFlow TTS	Repo	GitHub	MIT	Yes	English	Paper	GH Pages
Parler TTS	Repo	Hub	Apache 2.0	Yes	English	Not Available	🤗 Space
pflowTTS	Unofficial Repo	GDrive	MIT	Yes	English	Paper	Not Available	GPL-licensed phonemizer
Pheme	Repo	Hub	CC-BY	Yes	English	Paper	🤗 Space
Piper	Repo	Hub	MIT	Yes	Multilingual	Not Available	🤗 Space	GPL-licensed phonemizer
RAD-MMM	Repo	GDrive	MIT	Yes	Multilingual	Paper	Jupyter Notebook, Webpage
RAD-TTS	Repo	GDrive	MIT	Yes	English	Paper	GH Pages
Silero	Repo	GH links	CC BY-NC-SA	No	Multilingual	Not Available	Not Available	Non Commercial
StyleTTS 2	Repo	Hub	MIT	Yes	English	Paper	🤗 Space	GPL-licensed phonemizer
Tacotron 2	Unofficial Repo	GDrive	BSD-3	Yes	English	Paper	Webpage
TorToiSe TTS	Repo	Hub	Apache 2.0	Yes	English	Technical report	🤗 Space
TTTS	Repo	Hub	MPL 2.0	No	Multilingual	Not Available	Colab, 🤗 Space
VALL-E	Unofficial Repo	Not Available	MIT	Yes	NA	Paper	Not Available
VITS/ MMS-TTS	Repo	Hub / MMS	Apache 2.0	Yes	English	Paper	🤗 Space	GPL-licensed phonemizer
WhisperSpeech	Repo	Hub	MIT	No	Multilingual	Not Available	🤗 Space, Recordings, Colab
XTTS	Repo	Hub	CPML	Yes	Multilingual	Paper	🤗 Space	Non Commercial
xVASynth	Repo	Hub	GPL-3.0	Yes	Multilingual	Not Available	🤗 Space	Base model trained on non-permissive datasets

• Multilingual - Amount of supported languages is ever changing, check the Space and Hub which specific languages are supported
• ALL - Supports all natural languages; may not support artificial/contructed languages

Also to find a model for a specific language, filter out the TTS models hosted on HuggingFace: https://huggingface.co./models?pipeline_tag=text-to-speech&language=en&sort=trending

---

Legend

For the #above TTS capability table. Open the viewer in another window or even another monitor to keep both it and the legend in view.

• Processor ⚡ - Inference done by
  • CPU (CPUs = multithreaded) - All models can be run on CPU, so real-time factor should be below 2.0 to qualify for CPU tag, though some more leeway can be given if it supports audio streaming
  • CUDA by NVIDIA™
  • ROCm by AMD™, also see ONNX Runtime HF guide
• Phonetic alphabet 🔤 - Phonetic transcription that allows to control pronunciation of words before inference
  • IPA - International Phonetic Alphabet
  • ARPAbet - American English focused phonetics
• Insta-clone 👥 - Zero-shot model for quick voice cloning
• Emotion control 🎭 - Able to force an emotional state of speaker
  • 🎭 <# emotions> ( 😡 anger; 😃 happiness; 😭 sadness; 😯 surprise; 🤫 whispering; 😊 friendlyness )
  • 🎭👥 strict insta-clone switch - cloned on sample with specific emotion; may sound different than normal speaking voice; no ability to go in-between states
  • 🎭📖 strict control through prompt - prompt input parameter
• Prompting 📖 - Also a side effect of narrator based datasets and a way to affect the emotional state
  • 📖 - Prompt as a separate input parameter
  • 🗣📖 - The prompt itself is also spoken by TTS; ElevenLabs docs
• Streaming support 🌊 - Can playback audio while it is still being generated
• Speech control 🎚 - Ability to change the pitch, duration, etc. for the whole and/or per-phoneme of the generated speech
• Voice conversion / Speech-To-Speech 🦜 - Streaming support implies real-time S2S; S2T=>T2S does not count
• Longform synthesis 📜 - Able to synthesize whole paragraphs, as some TTS models tend to break down after a certain audio length limit

Example if the proprietary ElevenLabs were to be added to the capabilities table:

Name	Processor ⚡	Phonetic alphabet 🔤	Insta-clone 👥	Emotional control 🎭	Prompting 📖	Speech control 🎚	Streaming support 🌊	Voice conversion 🦜	Longform synthesis 📜
ElevenLabs	CUDA	IPA, ARPAbet	👥	🎭📖	🗣📖	🎚 stability, voice similarity	🌊	🦜	📜 Projects

More info on how the capabilities table came about can be found within the GitHub Issue.

train_data Legend

Legend for the separate TTSDS Datasets (train_data viewer GitHub)

• 🌐 Multilingual
  • The ISO codes of languages the model is capable off. ❌ if English only.
• 📚 Training Amount (k hours)
  • The number of hours the model was trained on
• 🧠 Num. Parameters (M)
  • How many parameters the model has, excluding vocoder and text-only components
• 🎯 Target Repr.
  • Which output representations the model uses, for example audio codecs or mel spectrograms
• 📖 LibriVox Only
  • If the model was trained on librivox-like (audiobook) data alone
• 🔄 NAR
  • If the model has a significant non-autoregressive component
• 🔁 AR
  • If the model has a significant autoregressive component
• 🔡 G2P
  • If the model uses G2P (phone inputs)
• 🧩 Language Model
  • If an LM-like approach is used (next token prediction)
• 🎵 Prosody Prediction
  • If prosodic correlates such as pitch or energy are predicted
• 🌊 Diffusion
  • If diffusion is used (outside vocoder)
• ⏱️ Delay Pattern
  • If a delay pattern is used for audio codes (see Lyth & King, 2024)

Please create pull requests to update the info on the models!

---