Hacker News

Show HN: KVoiceWalk – Voice cloning for Kokoro TTS using random walk algorithms

Hacker News - Wed, 05/21/2025 - 11:07am

I was blown away by Kokoro and what it managed to do with such little space. I became curious if it would be possible to create new voices by direct manipulation of the style tensors. After many failed attempts I finally landed on a method that properly scores the similarity of two audio segments that works well enough to random walk similar voices for Kokoro. I plan on using this scoring as part of a genetic algorithm, but wanted to baseline test it with this code.

The scoring mechanism using Resemblyzer to calculate similarity to target audio and similarity to another segment of audio it generates itself, self similarity. This self similarity was key in keeping the model stable and the audio consistent across inputs. But it was not enough to prevent over fitting to Resemblyzer.

I had to create a third metric which uses a normalized difference of a variety of audio features compared to the target features. Summing those I get a feature similarity metric which is useful in keeping audio quality from degrading too much and prevents over fitting.

The last challenge was weighting the score while keeping it flexible enough to explore the complex text to speech style space. Using a weighted harmonic mean allowed for back sliding on some metrics for significant improvement in others, which reduced stagnation and worked well enough for the random walk to work.

The results are fairly good. I would say it ends up in the uncanny valley of similarity rather than producing a proper clone of the target voice. It sounds like it might be the target voice, but does well enough to improve similarity from 70% to around 90%. There are probably limitations to the architecture of Kokoro in how close it can possibly sound to other voices, but there is probably some more progress to be made using a more advanced genetic algorithm.

Check out the code, make some new voices, and let me know if you have any ideas on ways to improve.

Comments URL: https://news.ycombinator.com/item?id=44052295

Points: 1

# Comments: 0

Categories: Hacker News

Ask HN: Why ChatGPT/Claude can't update projects?

Hacker News - Wed, 05/21/2025 - 11:04am

ChatGPT/Claude Projects are a great way to group files/chats and use them for long-running projects. But seems like there is no way to add/update files in projects. Why?

One of the basic project management needs is to keep track of its stages and generate some artifacts on each stage. Claude can generate artifacts, but can't add them to the project. ChatGPT can't either. Even if you add a markdown file with a simple TODO list, and want to update completed tasks, "projects" can't do it.

Is there a reason for such a limitation? Or am I missing something obvious?

Comments URL: https://news.ycombinator.com/item?id=44052262

Points: 1

# Comments: 0

Categories: Hacker News

CodeX but with Claude Code

Hacker News - Wed, 05/21/2025 - 10:12am

Article URL: https://cloudcoding.ai/chat

Comments URL: https://news.ycombinator.com/item?id=44051654

Points: 1

# Comments: 1

Categories: Hacker News

Show HN: I made an app that solves homework step-by-step from a photo

Hacker News - Wed, 05/21/2025 - 10:08am

Hi Everyone!!

At Lumination, we help educational institutions and students with ethical AI tools, guiding them from implementation to impact in an evolving learning landscape. We mix Pedagogy, Innovation & Ethics to craft enlightening, customized learning experiences, and tools to drive meaningful, educational outcomes.

Comments URL: https://news.ycombinator.com/item?id=44051616

Points: 1

# Comments: 0

Categories: Hacker News

Pages