Joe Rocca's picture

Joe Rocca

rocca

·

rocca27

AI & ML interests

None yet

Recent Activity

reacted to hexgrad's post with 👍 35 minutes ago

I wrote an article about G2P: https://hf.co/blog/hexgrad/g2p G2P is an underrated piece of small TTS models, like offensive linemen who do a bunch of work and get no credit. Instead of relying on explicit G2P, larger speech models implicitly learn this task by eating many thousands of hours of audio data. They often use a 500M+ parameter LLM at the front to predict latent audio tokens over a learned codebook, then decode these tokens into audio. Kokoro instead relies on G2P preprocessing, is 82M parameters, and thus needs less audio to learn. Because of this, we can cherrypick high fidelity audio for training data, and deliver solid speech for those voices. In turn, this excellent audio quality & lack of background noise helps explain why Kokoro is very competitive in single-voice TTS Arenas.

liked a Space 2 days ago

lllyasviel/iclight-v2

liked a Space 2 days ago

deepseek-ai/Janus-Pro-7B

View all activity

Organizations

rocca's activity

upvoted 2 collections about 2 months ago

Granite 3.1 Language Models

A series of language models with 128K context length trained by IBM licensed under Apache 2.0 license. • 8 items • Updated Dec 18, 2024 • 52

DeepSeek-VL2

5 items • Updated 1 day ago • 52

upvoted a collection 2 months ago

InternVL2.5

Better than InternVL 2.0 • 18 items • Updated 27 days ago • 81

upvoted 2 collections 3 months ago

LLM2CLIP

LLM2CLIP makes SOTA pretrained CLIP modal more SOTA ever. • 10 items • Updated 29 days ago • 55

glm-4-voice

3 items • Updated Oct 25, 2024 • 3

upvoted 2 collections 4 months ago

v4

18 items • Updated Oct 20, 2024 • 32

Resource Mirrors

Mirrors of some well-known sites/resources or something • 43 items • Updated Dec 31, 2024 • 6

upvoted a collection 6 months ago

magnum-v2

12 items • Updated Aug 23, 2024 • 7

upvoted a collection 7 months ago

InternLM2.5

14 items • Updated Sep 14, 2024 • 71

upvoted an article 9 months ago

Article

LLM Comparison/Test: Llama 3 Instruct 70B + 8B HF/GGUF/EXL2 (20 versions tested and compared!)

By

•

Apr 24, 2024

• 61

upvoted a collection 10 months ago

Meta Llama 3

This collection hosts the transformers and original repos of the Meta Llama 3 and Llama Guard 2 releases • 5 items • Updated Dec 6, 2024 • 711

upvoted a paper 11 months ago

The Era of 1-bit LLMs: All Large Language Models are in 1.58 Bits

Paper • 2402.17764 • Published Feb 27, 2024 • 608

upvoted 4 papers over 1 year ago

Example-based Motion Synthesis via Generative Motion Matching

Paper • 2306.00378 • Published Jun 1, 2023 • 6

Segment Anything Meets Point Tracking

Paper • 2307.01197 • Published Jul 3, 2023 • 35

DragonDiffusion: Enabling Drag-style Manipulation on Diffusion Models

Paper • 2307.02421 • Published Jul 5, 2023 • 34

DreamDiffusion: Generating High-Quality Images from Brain EEG Signals

Paper • 2306.16934 • Published Jun 29, 2023 • 31