Langcraft offers a developer‑first REST API that converts audio recordings into IPA phonemes with millisecond timestamps, providing per‑phoneme confidence and articulatory similarity scores. The service performs forced alignment to a reference transcript and returns structured JSON that highlights substitutions, insertions, deletions, and fluency metrics, enabling real‑time, automated pronunciation feedback for edtech, assessment, speech‑therapy, and corporate training applications.
Funding
Funding not disclosed
Founders
Product
Problem
Language learning platforms, oral assessment tools, and speech‑therapy applications need detailed, automated analysis of spoken input to provide immediate pronunciation feedback, but manual grading is time‑consuming and existing speech‑to‑text services lack phoneme‑level precision and confidence metrics.
Solution
Langcraft offers a developer‑first REST API that converts audio recordings into International Phonetic Alphabet (IPA) phonemes with millisecond‑accurate timestamps. The service performs forced alignment to a supplied reference transcript, generating per‑phoneme confidence scores and articulatory similarity metrics that identify substitutions, insertions, and deletions. By returning structured JSON containing canonical and produced phonemes, similarity scores, and fluency metrics, the API enables applications to deliver real‑time, granular pronunciation feedback without human raters. The solution supports common audio formats (WebM, WAV, MP3, M4A) and is designed for high‑volume integration in edtech, assessment, speech‑therapy, and corporate training contexts.
Target Audience
Primary customers are edtech developers, oral proficiency exam platforms, speech‑therapy software providers, and corporate training teams that require scalable, phoneme‑level speech analysis.
Features
- Accurate speech‑to‑IPA transcription with start and end timestamps for each phoneme
- Forced alignment to reference text for precise mispronunciation detection
- Confidence and articulatory similarity scores for automated pronunciation assessment
- JSON output detailing canonical vs. produced phonemes, substitution/insertion/deletion events, and fluency metrics
- Native support for WebM, WAV, MP3, and M4A audio uploads
- Simple REST endpoint (`POST https://api.langcraft.world/transcribe`) for easy integration