Broad Arabic speech for next-generation ASR. Access is restricted to NPL1.2 and the internal Quran Lab team. - Broad Arabic speech in segments of 2 to 20 seconds, cut at natural pauses (silences of 1.5 s or longer are removed), sized for CTC/transducer training and for LLM transcription. - Audio is 48 kHz Opus, exactly as the source streamed it (never re-encoded). Config tts holds studio Arabic podcasts and professional narration. Each recording is graded for TTS use on four 9 s windows inside its speech, and the grade is in metadata/recordings.parquet: Recordings with music under the speech (music score above 0.15) or under 75% speech also fail. Earlier broad recordings from the same…