Katsuya (@kn) header
Katsuya (@kn) profile picture

Co-founder & CTO at @Bluenote_AI

SF
Joined March 2011
Asked Codex to internationalize our app and localize it into Japanese before bed last night. Woke up to complete Japanese support this morning 🇯🇵 What would have taken a few days was done overnight.
14
40
12
545
206,516
Domain-specific agents are driving massive productivity gains. Success requires industry-tailored UX design, evaluation frameworks, guardrails, and post-training.
1
1
3
3,302
Talking to animals (or babies) using AI is happening fast…
The wildest company pitch I've ever heard? 🤔 I told @katelin_cruse it's gotta be an app that allows us to communicate with babies before they can talk... or even dolphins 👶🐬 And someday, we might actually get there thanks to AI ↓
2
3
6
2,741
Replying to @XDevelopers
My dashboard doesn't update - would you mind looking into this?
277
This aligns with my experience in translating from English to Japanese. For instance, translating culturally specific references, such as "Hot Vax Summer," poses a significant challenge for machine translation models. 👇
2
691
Replying to @runwayml
Love it! Added AI Narration using zeebra.ai/narration 🚀
5
12
2
72
19,478
More test on Elevenlabs vs OpenAI TTS - Elevenlabs is more engaging which is important for entertainment use cases. - OpenAI TTS occasionally doesn't follow the exact text (similar to tortoise/bark) For video translation, Elevenlabs generally sounds better for now. OpenAI is still great for use cases such as LLM chat and accessibility because of the quality vs cost tradeoff.
1
1
12
1,232
Here's a quick test of OpenAI vs. Elevenlabs TTS: - In English, Eleven sounds more animated and has studio quality (44.1kHz vs. 24kHz); however, OpenAI is also very good and would make sense for many use cases, especially given its 10x cheaper price point. - In Japanese, OpenAI personally sounds better and makes fewer mistakes in Kanji readings. Eleven makes quite a few Kanji mistakes but hallucinates less than OpenAI.
1
2
13
1,285
This will be the second biggest breakthrough consumer product after mobile - by @seevua
5
461
We'd love to help. ✉️ [email protected] We can produce high quality translation using advanced AI + professional services. Optimized for each platform: We generate multiple audio tracks for 1) YouTube multilingual audio tracks (speech aligned with video), and 2) Podcast (optimized for audio-only listening experience). I don't think lipsync is the right solution for you since YT doesn't support multilingual video tracks but available if interested.
1
1
13
2,901