Classify emotion and delivery with Resonance-2

Returns six signed opposite axes, 19 unipolar scores, all 31 category scores, and calibrated selected labels. Opposite labels cannot both be selected. Send 0.1–120 seconds of audio, at most 30 MiB. Longer files use nonoverlapping 20-second windows and average logits before calibration. Digital silence abstains with zero scores; there is no forced top label. Use regime=f1 (default) or regime=precision as a query parameter. Same shared plan minutes and affect tariff as Resonance 1: $0.0100 per audio minute for legacy metered accounts, one-second minimum. Requires approved organization access; existing accounts retain access. Request at /v1/models/oruk-resonance-2/access.