An accordion-repair loft dumps a week of bench recordings with no typed lines and labels the dump “the Riva set.” What should happen before treating it as ASR customization data?

A week of accordion-repair bench recordings with no typed lines is not yet a Riva ASR customization set. Reject it as an ASR pair until each usable clip has a matching transcript (or is set aside as unused audio). Volume alone does not qualify an unlabeled dump as a Riva ASR corpus. Converting the dump into CLIP image–caption pairs does not train ASR on speech. Redeploying Riva with a new Helm chart does not magically turn unpaired audio into transcripts. Exam caveat: unpaired audio is a collection gap, not customization data. Operational check: inventory clips, require transcripts for the ASR subset, quarantine unlabeled audio, and only then label the corpus “the Riva set.” Quarantine unlabeled accordion dumps until transcripts exist; do not name them Riva sets. Quarantine unlabeled accordion dumps until transcripts exist; do not name them Riva sets. Unlabeled accordion dumps are not Riva sets until usable clips have transcripts or are quarantined. Volume alone never qualifies an unpaired accordion dump; quarantine first, transcribe second, label as Riva set third. Close q138 with a municipal evidence note that records why the correct mechanism fits this civic multimodal scenario, why each of the three distractors fails by a distinct concept, and which operational check was observed in logs or dataset cards before promotion. File that note with the run packet so later maintenance cannot silently revive a rejected framing.