Open source AI Pronunciation trainer (aipronunciationtr.com)

An open‑source AI pronunciation trainer has been released that combines interactive sentence generation, live recording, and IPA feedback to help users practice and score their speech. In the web UI (best experienced in Chrome) you generate a new sentence via a right‑side bar, select language (upper left) and difficulty (lower left), then record with a mic button. Each sentence is accompanied by its IPA transcription, and the app gives a score (the guidance suggests aiming for 690 points/day), turning phonetic practice into a gamified routine. For the AI/ML community this is notable because it exposes a practical pipeline for linking user audio to phonetic transcriptions and automated scoring — useful for phonetics research, ASR fine‑tuning, pronunciation assessment models, and dataset creation. Being open source means researchers can inspect or adapt the scoring algorithm, integrate different acoustic models, swap language packs, or export annotated audio/IPA pairs for supervised training. The IPA output and per‑utterance scoring also make it a candidate tool for benchmarking pronunciation error detection, forced alignment quality, and model robustness across languages and difficulty levels.