ssmdconvert speech freeze

Freeze selected eligible automatic speech decisions into explicit SSMD semantics in a destination document or book bundle.

ssmdconvert speech freeze chapter.ssmd -o frozen.ssmd
ssmdconvert speech freeze BOOK.ssmdbook.zip --chapters 2-4 -o frozen.ssmdbook.zip

Fresh EPUB conversion applies this freeze automatically: ssmdconvert book convert novel.epub -o novel.ssmdbook. Use --speech-semantics preserve to request the historical structural-only result. The conversion-time aliases come from ttsready/spokenform, do not alter visible source text, and do not override author-owned annotations. A resolvable language is required for freeze; pass --language or use preserve when none is available.

freeze uses the same fail-closed mapping, validation, and atomic writer as speech annotate. It preserves visible text, does not duplicate authoritative SSMD substitutions, refuses unresolved effective language, and never overwrites its input. Existing output requires --force.

Dirty directory workspaces may be read, but the output must be a new path outside the workspace. Selected chapters are materialized in the output; unselected chapters remain unchanged. Bundle type is inferred from .ssmdbook or .ssmdbook.zip, or can be set with --format directory|zip.

To preview without writing, run speech audit. For the complete source, context, cache, and projection workflow, see SSMD analysis workflows.