Continuous recording and word pronunciation proficiency (0–1)

Returns the original CEFR and fluency fields, metadata, validity check and billing behavior. cefr_score stays 0–5 and fluency_score stays 0–10, from fluency_cefr-v2-final. Top-level check controls CEFR usability and billing; insufficient_audio is free. For compatibility, numeric legacy estimates are present on invalid checks but must not be treated as valid grades. Additive proficiency.score and words[].score are 0–1 pronunciation estimates from continuous-pilot-v2; unavailable or unscorable continuous values are null. proficiency.continuous_check reports pronunciation validity independently, and continuous_metadata holds alignment-based speech duration, word count and whether a reference was supplied. Legacy speech_seconds remains VAD-based and transcript_used includes automatic transcripts. Supply transcript to skip ASR, or omit it for Resonance transcription. Existing account billing terms apply. V2 was evaluated on short English read speech; its scores are not converted to CEFR. Longer recordings use word-weighted window scores, not separately validated long-form ratings.