- Published on
I benchmarked OpenAI gpt-4o-transcribe and Deepgram Nova-3 on real Korean/English kitchen audio — clean and noisy. Here are the CER, WER, latency, and cost numbers so you can pick the right model without guessing.
Speechtotext
All Posts
- ai (9)
- hackathon (5)
- buildinpublic (5)
- web3 (3)
- blockchain (3)
- community (3)
- meetup (2)
- multilingual (2)
- defi (1)
- llm (1)
- student-builders (1)
- bay-area-tech (1)
- developers (1)
- accountability (1)
- agents (1)
- voice-ai (1)
- networking (1)
- family (1)
- prompting (1)
- architecture (1)
- devtools (1)
- claudecode (1)
- devops (1)
- vibecoding (1)
- asr (1)
- korean (1)
speechtotext (1)