Source record

Published by arXiv cs.HC (T1 source). The original is at https://arxiv.org/abs/2607.24756.

Pipeline notes

The summary and note below are generated by the signal pipeline — they are Beyond Desk’s reading, not quotations from the source.

This paper derives a failure-mode-based evaluation framework for AI-generated oral history visualization. It compares a multi-agent scene-decomposition pipeline with a single summarization pipeline across 82 diaspora interviews, finds that scene planning and narrative preservation frequently conflict, and proposes a routing protocol that selects between systems based on source narrative strength.

Empirical comparison of multi-agent versus single-pipeline approaches for transforming testimony into image sequences, with a concrete routing heuristic grounded in 82 interviews.

Cited by

No citations on record.