- Published on
I fine-tuned a text-to-speech model to sound like Finland's former president using his 1975 New Year's speech and a MacBook M4. Spoiler: it mostly worked, but Finnish pronunciation needs GPU love.
Fine-tuning
All Posts
- llm (3)
- fine-tuning (2)
- llm-training (2)
- best-practices (1)
- claude (1)
- gpt-oss-120b (1)
- gpt-oss-20b (1)
- music (1)
- startup (1)
- innovation (1)
- venture (1)
- bandstack (1)
- verotus (1)
- demokratia (1)
- blockchain (1)
- kryptovaluutta (1)
- julkinen-talous (1)
- kansalaisvaikuttaminen (1)
- automaatio (1)
- text-to-speech (1)
- qwen3 (1)
- machine-learning (1)
- macos (1)
- Published on
A comprehensive breakdown of LLM training costs, covering GPU infrastructure requirements, cost analysis for 20B and 120B parameter models, deployment considerations, and practical optimization strategies to reduce expenses while maintaining performance.