A public snapshot of conversational speech research code, synthetic speech data, model checkpoints, manuscripts, and evaluation records. Upload is in progress. Volumes preserve directory paths beneath a chosen restore root. Runtime environments, credentials and Git internals are excluded. Files held for content review are not yet included. Existing upstream licenses still apply; no new blanket license is asserted. Each volume is an independent tar.zst archive and includes a manifest with per-file SHA256 hashes. Verify archive hashes from the upload index, then extract into an empty restore directory. Do not extract directly over an existing installation. Absolute filesystem links require…