With the release and hype around Drama-3 by Fish Audio, I've decided to challenge myself & create my own TTS model in my spare time.It is 1/10th the size of Drama 3, can be just as expressive, and can take instructions about emotion and pacing for fine-grained output.Obviously it's not perfect, but from my experience still better than whatever is out there at this size. Burned 300 bucks on this thing so enjoy.Demo: https://vojtechcekal.com/hlasP.S.: it's a Breeze v2 finetune (best open weight base model out there overall) so their weird license applies (unfortunately). Might finetune on a more permissive base in the future.
Comments URL: https://news.ycombinator.com/item?id=50205300
Points: 1
# Comments: 0
Moonshot AI released Kimi K3.1 Pro, a 2.8T-parameter, 104B-active MoE model with 1M-token context. It scores 65.7% on Terminal-Bench 4.0, costs $2/$10 per million…
Meta released Muse Spark Open, an 816B-parameter open-weight MoE model with 42B active parameters and a 1M-token context. It leads open models on MCP-Atlas and runs on 2…
A new update to the Calibre ebook manager is out and it brings more natural sounding text-to-speech (TTS) voices, as well as AVIF image support and better handling of…
Discussion (0)
No comments yet. Start the conversation!