This article presents ALLIES corpus along with protocols. ALLIES is a French meta-corpus of almost 500 hours of speech and 1048 files. In addition to the Train set, 3 Test sets have been designed to evaluate segmentation tasks. A first contribution is the homogenization of speaker names in the whole corpus and the complete segmentation and transcription work done on the FullTest partition regarding overlap, music, noise, speakers, segmentation and transcription. This corpus will be included in ELRA catalogue in 2024. As part of the work done during JSALT 2023, this article also provides as a second contribution, several baseline results for speech segmentation, speaker diarization, speech recognition and speaker change detection from the transcription.