Here’s an early preview of ElevenLabs Music.
All of the songs in this thread were generated from a single text prompt with no edits.
Title: It Started to Sing
Style: “Pop pop-rock, country, top charts song.”
Excited to share what I've been working on for the past 6 months: Text-to-Audio Diffusion.
More samples and info here:
Enjoy!
🔊 Prompt: "Dubstep Insane Drop Remix (Deluxe Edition), 2 of 4"
The world's most advanced TTS system is now accessible with the new Elevenlabs Python API - a few lines of code, and you're good to go.
Streaming? Custom voices? All there. Enjoy! 🗣️
GitHub:
Colab:
@elevenlabsio
now supports low-latency input streaming. This allows you to listen to LLMs in real-time as the text is being generated.
🔊 "A one-sentence relaxing speech"
Drag & drop with 50'000 objects using instanced meshes. In the process of making this even more declarative - this shouldn't be possible in a web browser.
@reactjs
+
@threejs_org
+ R3F + React Spring
Working on Graphire: a fully declarative and unopinionated react graph library. It supports force layouts in both 2D and 3D using SVGs or
@threejs
with R3F.
repo:
sandbox:
Draggable graph nodes with graphire, use-gesture, and react-three-fiber. Graphire makes it super simple to build diagrams, networks, or flow interfaces.
graphire:
sandbox:
Working on a new unconditional voice generation architecture with diffusion models in
@PyTorch
. I'm really enjoying using
@weights_biases
reports to document the progress. Here are a few weird but cool generated voice samples:
A declarative interface for general/force graphs that uses instancing. The integration between functional D3 and declarative React interfaces has always been uncomfortable - we need react-d3-fiber 😅
libs:
@threejs
, react-three-fiber, d3-force-3d.
box:
We are delighted to start the year by announcing a set of upcoming product launches and our $80 million Series B funding round. We are grateful for the continued trust from
@a16z
,
@natfriedman
&
@danielgross
who co-led the round.
Listen & Read:
@borisdayma
I think it refers to something like this, but with text: . The idea is that you concatenate the usual mean token with all other N text embedding, then you attend on this N+1 sequence and output only the last token as sentence embedding.
Added new Extract feature to surgeon-pytorch, allowing you to extract sub-graphs of
@PyTorch
models using symbolic tracing – much better than Inspect (
@rasbt
).
Easily get any intermediate function output, and change input nodes.