watsonmusic

8 ms·

1.
by watsonmusic 1y ago
it's not oss
2.
by watsonmusic 1y ago
bonus usage
3.
by watsonmusic 1y ago
11labs is facing a real competitor
4.
by watsonmusic 1y ago
genius
5.
by watsonmusic 1y ago
this model is superb
6.
by watsonmusic 1y ago
Microsoft is cool
7.
by watsonmusic 1y ago
yes the best
8.
by watsonmusic 1y ago
one of the best models built by Microsoft
9.
by watsonmusic 1y ago
https://github.com/microsoft/VibeVoice
10.
by watsonmusic 1y ago
https://huggingface.co/microsoft/VibeVoice-1.5B
11.
Microsoft releases VibeVoice, generates 90-minute, 4-speaker audio(microsoft.github.io)
3 points by watsonmusic 1y ago|3 comments
12.
by watsonmusic 1y ago
VibeVoice is a novel framework designed for generating expressive, long-form, multi-speaker conversational audio, such as podcasts, from text. It addresses significant challenges in traditional Text-to-Speech (TTS) systems, particularly in
13.
by watsonmusic 1y ago
cannot wait seeing how it goes beyond the current llm training pipeline
14.
by watsonmusic 1y ago
it could be adaptive. only high-value tokens were allocated with more compute
15.
by watsonmusic 1y ago
A new scaling paradigm finally comes out!
16.
by watsonmusic 1y ago
14b model performs comparably with 32b size. the improvement is huge
17.
by watsonmusic 2y ago
negative values can enhance the expressibility
18.
by watsonmusic 2y ago
not all hallucinations are creativity Imaginate that for a RAG application, the model is supposed to follow the given documents
19.
by watsonmusic 2y ago
the model is supposed to learn this
20.
by watsonmusic 2y ago
The modification is simple and beautiful. And the improvements are quite significant.
21.
by watsonmusic 2y ago
that would be huge!