What's this about?
AuK uses text instructions and audio context for speech generation, content edits, enhancement, separation and acoustic changes. Code and weights are open.
What exactly is speech editing?
It alters an existing recording—changing words, voice qualities or background acoustics—without rerecording the whole clip, much like editing text while preserving the speaker's identity.
Why is it worth attention?
The breadth and four-step Flash version are useful for local research. Author benchmarks look strong, but a Chinese-language hands-on test heard a foreign accent, illustrating the gap between aggregate scores and language-specific quality.