HEAIN-SD-1.0: Policy-Governed Subtitling and Dubbing for Thai on Hybrid Edge AI Nodes

Abstract

Automatic dubbing systems now produce intelligible speech, but professional subtitling and dubbing remain constrained by three problems that are seldom addressed together. Expressive delivery has to be directed by people who read the picture; profane or censorable content has to obey rules that differ by country and destination platform; and the recordings that make a system better are the personal data of identifiable voices. This paper introduces HEAIN-SD-1.0, an Application Profile that specifies how a Thai-centered subtitling and dubbing module runs as Layer 3 (Application AI) on HEAIN Core Protocol and the Hybrid Edge AI Node architecture. Four contributions are described. First, a policy-governed autonomy ladder (L0–L3): the system may learn everything about delivery and nothing about what is permitted, because content policy is a separately owned, fail-closed registry selected per destination. Second, a four-part decomposition of dubbed voice, in which timbre comes from the source performance, accent from the standard of professional dubbing teams in the target country, and emotion and rhythm from a producer's direction, realized as retrieval-prompted synthesis over a library of consented professional deliveries. Third, a protocol profile with a session model, a CORE/EDGE zone split, and an uplink whitelist that keeps media on the processing node and sends only records, scores, and small embeddings to the control node. Fourth, an evaluation plan with quality gates fixed before any experiment. The paper reports no experimental results: the profile is an unimplemented design, and every quantity marked as an estimate is a planning value. Keywords: Automatic Dubbing, Subtitling, Thai Language, Voice Cloning, Tonal Languages, Human-in-the-Loop, Content Policy, Edge Computing, Consent Management, Protocol Specification