Speech-to-Digits API – 95% accuracy on spoken numbers

Show HN: Speech-to-Digits API – 95% accuracy on spoken numbers

Speech-to-Digits API – 95% accuracy on spoken numbers operates in show hn launch.

Profile

Hi HN,I built EchoEntry (https://echoentry.ai) – a speech-to-text API optimized specifically for digits.The problem: Generic STT APIs struggle with numbers. "One oh five" becomes "105" sometimes, "15" other times. For healthcare apps, warehouse systems, or IVR, this inconsistency breaks workflows.My solution: Fine-tuned Whisper-small on 1-999 spoken numbers across 5 English accents. Gets 95% accuracy on 1-3 digit numbers.Tech stack: - Custom Whisper model (1.7GB) - FastAPI backend - Deployed on 8GB Linode - FFmpeg for audio processingTry it now (two commands, no signup):# Download test audio curl -O https://echoentry.ai/test_audio.wav# Test the API curl -X POST https://api.echoentry.ai/v1/tr

Within the show hn launch space, the company focuses on its core offering rather than a broad product line.

The company is listed at the Show HN launch stage of development.

Website: echoentry.ai