text-to-text-transfer-transformer VS tortoise-tts

Google released the T5 paper about 5 years ago:

https://arxiv.org/abs/1910.10683

This included full model weights along with a detailed description of the dataset, training process, and ablations that led them to that architecture. T5 was state-of-the-art on many benchmarks when it was released, but it was of course quickly eclipsed by GPT-3.

Following GPT-3, it became much more common for labs to not release full details or model weights. Prior to that, it was common practice from Google (BERT, T5), Meta (BART), OpenAI (GPT1, GPT2) and others to release full training details and model weights.

[2] - https://arxiv.org/abs/1910.10683

T5-style relative encodings https://arxiv.org/abs/1910.10683

11. T5: Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer" (2020) - https://arxiv.org/abs/1910.10683 (Google Research)

You can find a link to the paper here: https://arxiv.org/abs/1910.10683

The Text-to-Text Transformer is a type of neural network architecture that is particularly well-suited for natural language processing tasks involving the generation of text. It was introduced in the paper "Attention is All You Need" by Vaswani et al. and has since become a popular choice for many NLP tasks, including language translation, summarization, and text generation

Found relevant code at https://github.com/google-research/text-to-text-transfer-transformer + all code implementations here

I use Tortoise TTS. It's slow, a little clunky, and sometimes the output gets downright weird. But it's the best quality-oriented TTS I've found that I can run locally.

https://github.com/neonbjb/tortoise-tts

neonbjb/tortoise-tts

EDIT: It appears to be the exact same issue as this

It was primarily being used to train TTS models (see https://github.com/neonbjb/tortoise-tts), which largely fit into a single GPUs memory. So, for data parallelism, x8 PCIe isn't that much of a concern.

Previously TortoiseTTS was associated with PlayHT in some way, although the exact connection is a bit vague [0].

From the descriptions here it sounds a lot like AudioLM / SPEAR TTS / some of Meta's recent multilingual TTS approaches, although those models are not open source, sounds like PlayHT's approach is in a similar spirit. The discussion of "mel tokens" is closer to what I would call the classic TTS pipeline in many ways... PlayHT has generally been kind of closed about what they used, would be interesting to know more.

I assume the key factor here is high quality, emotive audio with good data cleaning processes. Probably not even a lot of data, at least in the scale of "a lot" in speech, e.g. ASR (millions of hours) or TTS (hundreds to thousands). As opposed to some radically new architectural piece never before seen in the literature, there are lots of really nice tools for emotive and expressive TTS buried in recent years of publications.

Tacotron 2 is perfectly capable of this type of stuff as well, as shown by Dessa [1] a few years ago (this writeup is a nice intro to TTS concepts). With the limit largely being, at some point you haven't heard certain phonetic sounds before in a voice, and need to do something to get plausible outcomes for new voices.

[0] Discussion here https://github.com/neonbjb/tortoise-tts/issues/182#issuecomm...

[1] https://medium.com/dessa-news/realtalk-how-it-works-94c1afda...

Tortoise GitHub repo - Source code, documentation, and usage guide

TorToiSe (https://github.com/neonbjb/tortoise-tts) produces the best quality speech of any freely available model. However, its long inference times makes it impractical for voice chatbots like Gdansk.