@mengine/medeo-client 2.1.1-dsl.8 → 2.1.1-dsl.9

package/dist/{dsl-DbYqFFOb.js → dsl-COyzz7Fp.js}

@@ -1,4 +1,4 @@
-import { $ as assertId, A as hardenCidDescriptors, B as safeStringify, C as applySchemaToInferOptions, D as deepEqual, E as decodeNestedJsonValues, F as isObject, G as valuesEqual, H as stripUndefined, I as isObjectLike, J as getChildSchema, K as CID_KEY, L as isStateAndSchemaOfType, M as inferContainerTypeFromValue, N as insertChildToMap, O as defineCidProperty, P as isArrayLike, Q as assertEntityPatch, R as isTreeID, S as applyEncode, T as containerIdToContainerType, U as tryInferContainerType, V as schemaToContainerType, W as tryUpdateToContainer, X as assertCaptionContent, Y as getMapFieldSchema, Z as assertEntity, _ as isLoroTreeSchema, at as isRecord, b as validateSchema, c as isInvalidAtomic, ct as relationTables, d as getDefaultValue, dt as assertResourceIdentity, et as assertPhoneme, f as isContainerSchema, ft as defaultIdFactory, g as isLoroMovableListSchema, h as isLoroMapSchema, it as entityTables, j as hasTransform, k as getRootContainerByType, lt as validLanguage, m as isLoroListSchema, mt as resourceId, n as textStyles, nt as assertTimeAnchor, o as medeoDslLoroShape, ot as mergeFields, p as isListLikeSchema, pt as relationId, q as getChildContainerSchema, rt as atomicFields, s as captionContentCodec, st as patchRow, t as DEFAULT_UNIT_TIME_MS, tt as assertRelation, u as schema, ut as assertRelationIdentity, v as isRootSchemaType, w as cidsEqual, x as applyDecode, y as schemaContainsIgnoreField, z as isValueOfContainerType } from "./video-draft-types-Ce8NDenQ.js";
+import { $ as assertEntityPatch, A as getRootContainerByType, B as isValueOfContainerType, C as applyEncode, D as decodeNestedJsonValues, E as containerIdToContainerType, F as isArrayLike, G as tryUpdateToContainer, H as schemaToContainerType, I as isObject, J as getChildContainerSchema, K as valuesEqual, L as isObjectLike, M as hasTransform, N as inferContainerTypeFromValue, O as deepEqual, P as insertChildToMap, Q as assertEntity, R as isStateAndSchemaOfType, S as applyDecode, T as cidsEqual, U as stripUndefined, V as safeStringify, W as tryInferContainerType, X as getMapFieldSchema, Y as getChildSchema, Z as assertCaptionContent, _ as isLoroMovableListSchema, at as atomicFields, b as schemaContainsIgnoreField, c as isInvalidAtomic, ct as mergeFields, d as schema, dt as validLanguage, et as assertId, f as getDefaultValue, ft as assertRelationIdentity, g as isLoroMapSchema, gt as resourceId, h as isLoroListSchema, ht as relationId, it as assertTimeAnchor, j as hardenCidDescriptors, k as defineCidProperty, l as phonemeCodec, lt as patchRow, m as isListLikeSchema, mt as defaultIdFactory, n as textStyles, nt as assertPhoneticScript, o as medeoDslLoroShape, ot as entityTables, p as isContainerSchema, pt as assertResourceIdentity, q as CID_KEY, rt as assertRelation, s as captionContentCodec, st as isRecord, t as DEFAULT_UNIT_TIME_MS, tt as assertPhoneme, ut as relationTables, v as isLoroTreeSchema, w as applySchemaToInferOptions, x as validateSchema, y as isRootSchemaType, z as isTreeID } from "./video-draft-types-BnbfWrBx.js";
 import { Cursor, LoroDoc, LoroList, LoroMap, LoroMovableList, LoroText, LoroTree, isContainer, isContainerId } from "loro-crdt";
 import { Immer, setAutoFreeze } from "immer";
 //#region ../medeo-dsl/src/text.ts
@@ -104,42 +104,27 @@ function prepareTextEdits(text, edits) {
 			}
 			return edit;
 		}),
-		body: text
+		text
 	};
 }
-/** An empty body has no segments; otherwise each line, including a trailing empty one, is a segment. */
-function lines(body) {
-	if (body === "") return [];
-	const result = [];
-	let start = 0;
-	for (const text of body.split("\n")) {
-		result.push({
-			text,
-			start,
-			end: start + text.length
-		});
-		start += text.length + 1;
-	}
-	return result;
-}
-/** The language a whole line carries: one shared mark on every unit, else none. */
-function lineLanguage(text, line) {
-	if (line.start === line.end) return void 0;
-	const first = text.langs[line.start];
-	for (let i = line.start + 1; i < line.end; i++) if (text.langs[i] !== first) return void 0;
+/** The language a whole segment carries: one shared mark on every unit, else none. */
+function markedLanguage(text) {
+	if (!text.text.length) return void 0;
+	const first = text.langs[0];
+	for (let i = 1; i < text.text.length; i++) if (text.langs[i] !== first) return void 0;
 	return first;
 }
 /** Read language marks off a rich-text delta. */
 function langsFromDelta(delta) {
-	let body = "";
+	let text = "";
 	const langs = [];
 	for (const run of delta) {
-		body += run.insert;
+		text += run.insert;
 		const language = typeof run.attributes?.language === "string" ? run.attributes.language : void 0;
 		for (let i = 0; i < run.insert.length; i++) langs.push(language);
 	}
 	return {
-		body,
+		text,
 		langs
 	};
 }
@@ -178,16 +163,23 @@ const referenceFields = {
 	"phonetic-script-entities": ["phoneme"],
 	"caption-entities": ["selection"]
 };
+/** The segment a selection names: refused when its script or the segment is missing. */
+function requireSegment(rows, audioScriptId, segmentId) {
+	const script = requireEntity(rows, "audio-script-entities", audioScriptId);
+	const segment = Array.isArray(script.segments) ? script.segments.find((s) => isRecord(s) && s.id === segmentId) : void 0;
+	if (!segment) throw new Error(`Missing segment ${String(segmentId)} of ${String(audioScriptId)}`);
+	return segment;
+}
 /**
 * Whether a single-valued relation, stored as a field of the entity that owns it, may hold this
 * value among these rows. The owner need not be stored yet; an absent value unlinks it.
-* A caption selection's range is checked against the live body by whoever holds it.
+* A selection's run is checked against the segment's live text by whoever holds it.
 */
 function assertReferenceAdmitted(rows, table, id, field, value) {
 	assertId(id);
 	const key = `${table}.${field}`;
 	if (value === void 0) {
-		if (key === "clip-entities.source" || key === "phonetic-script-entities.phoneme") throw new Error(`Missing required relation: ${key}`);
+		if (key === "clip-entities.source") throw new Error(`Missing required relation: ${key}`);
 		return;
 	}
 	switch (key) {
@@ -213,26 +205,49 @@ function assertReferenceAdmitted(rows, table, id, field, value) {
 			break;
 		}
 		case "caption-entities.selection":
-			assertCaptionContent(value);
-			requireEntity(rows, "audio-script-entities", value.audioScriptId);
-			break;
-		case "phonetic-script-entities.phoneme":
-			assertPhoneme(value);
-			requireEntity(rows, "audio-script-entities", value.audioScriptId);
+		case "phonetic-script-entities.phoneme": {
+			if (key === "caption-entities.selection") assertCaptionContent(value);
+			else assertPhoneme(value);
+			const { audioScriptId, segmentId } = value;
+			requireSegment(rows, audioScriptId, segmentId);
 			break;
+		}
 	}
 }
-/** Both ends of a caption's selection exist: the caption and the script it selects from. */
+/** Both ends of a caption's selection exist: the caption and the segment it selects. */
 function assertCaptionEndpoints(rows, selection) {
 	requireEntity(rows, "caption-entities", selection.captionId);
-	requireEntity(rows, "audio-script-entities", selection.audioScriptId);
+	requireSegment(rows, selection.audioScriptId, selection.segmentId);
+}
+/** Both ends of a phoneme exist: the PhoneticScript and the segment it is written for. */
+function assertPhonemeEndpoints(rows, selection) {
+	requireEntity(rows, "phonetic-script-entities", selection.phoneticScriptId);
+	requireSegment(rows, selection.audioScriptId, selection.segmentId);
 }
-/** A caption selects a nonempty run of its script's current body, in UTF-16 offsets. */
-function assertCaptionSelection(rows, selection, body) {
+/**
+* A selection of a segment's current text: the whole segment, or a run of it in UTF-16 offsets
+* with both ends stated. What it selects is never empty.
+*/
+function assertSegmentRange(range, text) {
+	if (range.start === void 0 !== (range.end === void 0)) throw new Error("A run inside a segment states both its start and its end");
+	if (range.start === void 0 || range.end === void 0) {
+		if (!text.length) throw new Error("An empty segment cannot be selected");
+		return;
+	}
+	assertTextOffset(text, range.start);
+	assertTextOffset(text, range.end);
+	if (range.start >= range.end) throw new Error("Expected a nonempty range in the current segment");
+}
+/** A caption selects a segment of its script, or a nonempty run of it. */
+function assertCaptionSelection(rows, selection, text) {
 	assertCaptionEndpoints(rows, selection);
-	assertTextOffset(body, selection.start);
-	assertTextOffset(body, selection.end);
-	if (selection.start >= selection.end) throw new Error("Expected a nonempty range in the current body");
+	assertSegmentRange(selection, text);
+}
+/** A PhoneticScript is written for a segment of its script, or a nonempty run of it, as a Caption selects one. */
+function assertPhonemeSelection(rows, selection, text) {
+	assertPhonemeEndpoints(rows, selection);
+	assertSegmentRange(selection, text);
+	if (selection.phonemeScript !== void 0 && typeof selection.phonemeScript !== "string") throw new Error("phonemeScript is a string");
 }
 /** A caption's text comes from a selection of a script or from its own text, never both. */
 function assertCaptionSource(caption, selected) {
@@ -3898,11 +3913,21 @@ function rootMap(doc, name) {
 	const value = doc.getContainerById(id);
 	return value?.kind() === "Map" ? value : void 0;
 }
-function liveText(doc, scriptId, caption = false) {
-	const row = rootMap(doc, caption ? "caption-entities" : "audio-script-entities")?.get(scriptId);
-	if (!isContainer(row) || row.kind() !== "Map") return;
-	const body = row.get(caption ? "text" : "body");
-	return isContainer(body) && body.kind() === "Text" && !body.isDeleted() ? body : void 0;
+function asMap(value) {
+	return isContainer(value) && value.kind() === "Map" ? value : void 0;
+}
+function asText(value) {
+	return isContainer(value) && value.kind() === "Text" && !value.isDeleted() ? value : void 0;
+}
+/** The live Text a target names: a Caption's own, or a script segment's. */
+function liveText(doc, target) {
+	if ("captionId" in target) return asText(asMap(rootMap(doc, "caption-entities")?.get(target.captionId))?.get("text"));
+	const segments = asMap(rootMap(doc, "audio-script-entities")?.get(target.audioScriptId))?.get("segments");
+	if (!isContainer(segments) || segments.kind() !== "MovableList") return;
+	for (const item of segments.toArray()) {
+		const segment = asMap(item);
+		if (segment?.get("id") === target.segmentId) return asText(segment.get("text"));
+	}
 }
 function readContainNodes(nodes) {
 	const read = (nodes) => nodes.map((node) => ({
@@ -3912,19 +3937,41 @@ function readContainNodes(nodes) {
 	}));
 	return read(nodes);
 }
+/** What a selection holds now: its whole segment, or the run between its Cursors on that segment's Text. */
 function resolveContent(doc, content) {
-	const text = liveText(doc, content.audioScriptId);
-	if (!text || content.start.containerId() !== text.id || content.end.containerId() !== text.id) return;
+	const text = liveText(doc, {
+		audioScriptId: content.audioScriptId,
+		segmentId: content.segmentId
+	});
+	if (!text) return;
+	const whole = text.toString();
+	if (content.start === void 0 || content.end === void 0) return {
+		text: whole,
+		start: 0,
+		end: whole.length,
+		bodyId: text.id
+	};
+	if (content.start.containerId() !== text.id || content.end.containerId() !== text.id) return;
 	const start = doc.getCursorPos(content.start)?.offset;
 	const end = doc.getCursorPos(content.end)?.offset;
 	if (start === void 0 || end === void 0 || start > end) return;
 	return {
-		text: text.toString().slice(start, end),
+		text: whole.slice(start, end),
 		start,
 		end,
 		bodyId: text.id
 	};
 }
+function richTextOf(text) {
+	return {
+		bodyId: text.id,
+		plainText: text.toString(),
+		delta: text.toDelta().flatMap((run) => typeof run.insert === "string" ? [{
+			insert: run.insert,
+			...run.attributes ? { attributes: run.attributes } : {}
+		}] : [])
+	};
+}
 /** Optional relation fields read apart from their row: an unreadable one is dropped, the row is kept. */
 const optionalReferences = {
 	"clip-entities": ["anchor"],
@@ -3962,19 +4009,16 @@ function readMedeoDsl(doc, state) {
 		}
 		assertEntity(table, v);
 		assertResourceIdentity(table, id, v);
+		if (table === "phonetic-script-entities") assertPhoneticScript(void 0, v);
+		const texts = [];
+		if (table === "audio-script-entities") for (const segment of v.segments) texts.push([segment.id, liveText(doc, {
+			audioScriptId: id,
+			segmentId: segment.id
+		})]);
+		if (table === "caption-entities" && v.text !== void 0) texts.push([id, liveText(doc, { captionId: id })]);
+		for (const [, text] of texts) if (!text) throw new Error("Missing live LoroText");
 		entities[table][id] = v;
-		if (table === "audio-script-entities" || table === "caption-entities" && v.text !== void 0) {
-			const text = liveText(doc, id, table === "caption-entities");
-			if (!text) throw new Error("Missing live LoroText body");
-			richText[id] = {
-				bodyId: text.id,
-				plainText: text.toString(),
-				delta: text.toDelta().flatMap((run) => typeof run.insert === "string" ? [{
-					insert: run.insert,
-					...run.attributes ? { attributes: run.attributes } : {}
-				}] : [])
-			};
-		}
+		for (const [key, text] of texts) richText[key] = richTextOf(text);
 	} catch (error) {
 		delete entities[table][id];
 		diagnostic("invalid-entity", `${table}/${id}`, error);
@@ -4033,17 +4077,26 @@ function readMedeoDsl(doc, state) {
 		try {
 			const resolved = resolveContent(doc, caption.selection);
 			if (resolved) captionText[id] = resolved;
-			else diagnostic("invalid-caption-content", `caption-entities/${id}/selection`, "Cursor does not resolve in the current body");
+			else diagnostic("invalid-caption-content", `caption-entities/${id}/selection`, "Selection does not resolve in its current segment");
 		} catch (error) {
 			diagnostic("invalid-caption-content", `caption-entities/${id}/selection`, error);
 		}
 	}
+	const phoneticText = {};
+	for (const [id, phonetic] of Object.entries(entities["phonetic-script-entities"])) try {
+		const resolved = resolveContent(doc, phonetic.phoneme);
+		if (resolved) phoneticText[id] = resolved;
+		else diagnostic("invalid-phoneme", `phonetic-script-entities/${id}/phoneme`, "Selection does not resolve in its current segment");
+	} catch (error) {
+		diagnostic("invalid-phoneme", `phonetic-script-entities/${id}/phoneme`, error);
+	}
 	return freezeSnapshot({
 		...entities,
 		...relations,
 		"contain-relations": validTree(treeNodes),
 		richText,
 		captionText,
+		phoneticText,
 		diagnostics
 	});
 }
@@ -4056,6 +4109,19 @@ function freezeSnapshot(value) {
 }
 //#endregion
 //#region src/dsl/document/draft.ts
+function textKey(target) {
+	return "captionId" in target ? JSON.stringify(["caption", target.captionId]) : JSON.stringify([
+		"segment",
+		target.audioScriptId,
+		target.segmentId
+	]);
+}
+/** Rows holding a segment selection, by the field that holds it (D181). */
+const selectionField = {
+	"caption-entities": "selection",
+	"phonetic-script-entities": "phoneme"
+};
+const isSelectionTable = (table) => Object.hasOwn(selectionField, table);
 const entities = new Set(entityTables);
 const relations = new Set(relationTables);
 const reserved = new Set([
@@ -4088,9 +4154,10 @@ var TransactionDraft = class {
 	entities = /* @__PURE__ */ new Map();
 	relations = /* @__PURE__ */ new Map();
 	deleted = /* @__PURE__ */ new Set();
+	/** Texts this write edits, by {@link textKey}, as its edits so far leave them. */
 	textPreviews = /* @__PURE__ */ new Map();
-	/** Captions whose selection the draft wrote; they are replayed natively in order with text edits. */
-	dirtyCaptions = /* @__PURE__ */ new Set();
+	/** Selections the draft wrote, by table then row; they are replayed natively in order with text edits. */
+	dirty = /* @__PURE__ */ new Map();
 	removedCaptionTexts = /* @__PURE__ */ new Set();
 	treeChanged = false;
 	active = true;
@@ -4121,13 +4188,30 @@ var TransactionDraft = class {
 	close() {
 		this.active = false;
 	}
+	dirtySelections() {
+		return [...this.dirty].flatMap(([table, ids]) => [...ids].map((id) => [table, id]));
+	}
+	clearDirtySelections() {
+		this.dirty.clear();
+	}
+	/** Whether this write edits any segment of a script. */
+	usesScript(scriptId) {
+		const prefix = JSON.stringify(["segment", scriptId]).slice(0, -1) + ",";
+		return [...this.textPreviews.keys()].some((key) => key.startsWith(prefix));
+	}
 	wrap(value, path) {
 		if (!value || typeof value !== "object" || value instanceof Cursor) return value;
 		const cached = this.wrappers.get(value);
 		if (cached) return cached;
 		const proxy = new Proxy(value, {
 			get: (target, key) => this.run(() => {
-				if ((path[0] === "audio-script-entities" && key === "body" || path[0] === "caption-entities" && key === "text") && path.length === 2 && this.textPreviews.has(path[1])) return this.textPreviews.get(path[1]);
+				if (key === "text") {
+					const preview = path[0] === "caption-entities" && path.length === 2 ? this.textPreviews.get(textKey({ captionId: path[1] })) : path[0] === "audio-script-entities" && path[2] === "segments" && path.length === 4 ? this.textPreviews.get(textKey({
+						audioScriptId: path[1],
+						segmentId: String(target.id)
+					})) : void 0;
+					if (preview !== void 0) return preview;
+				}
 				if (!path.length && key === "contain-relations" && target[key] === void 0) target[key] = [];
 				return this.wrap(Reflect.get(target, key), [...path, String(key)]);
 			}),
@@ -4182,16 +4266,22 @@ var TransactionDraft = class {
 			if (path.length === 1 && (remove || next === void 0)) this.deleted.add(lifetime);
 			else if (this.deleted.has(lifetime)) throw new Error("Recreate an entity in a separate transaction");
 			if (table === "audio-script-entities") {
-				if (path.length === 2 && key === "body" && (initial || this.textPreviews.has(rowId))) throw new Error("Edit body through the Text adapter");
-				if (path.length === 1 && initial && isRecord(next) && next.body !== initial.body) throw new Error("Edit body through the Text adapter");
-				if (path.length === 1 && this.textPreviews.has(rowId)) throw new Error("Cannot delete or replace a body used in this transaction");
+				if ((path.length === 2 && key === "segments" || path[2] === "segments" || path.length === 1 && initial && isRecord(next) && !sameSegments(next.segments, initial.segments)) && (initial || this.usesScript(rowId))) throw new Error("Change segments with segment mutations and edit their text through the Text adapter");
+				if (path.length === 1 && this.usesScript(rowId)) throw new Error("Cannot delete or replace a Text used in this transaction");
+			}
+			if (isSelectionTable(table)) {
+				const field = selectionField[table];
+				if (path.length === 1 && isRecord(next) && Object.hasOwn(next, field) || path.length === 2 && key === field || path[2] === field) {
+					let ids = this.dirty.get(table);
+					if (!ids) this.dirty.set(table, ids = /* @__PURE__ */ new Set());
+					ids.add(rowId);
+				}
 			}
 			if (table === "caption-entities") {
-				if (path.length === 1 && isRecord(next) && Object.hasOwn(next, "selection") || path.length === 2 && key === "selection" || path[2] === "selection") this.dirtyCaptions.add(rowId);
 				const changesText = path.length === 1 || path.length === 2 && key === "text";
 				const nextText = path.length === 1 ? isRecord(next) ? next.text : void 0 : next;
 				if (changesText) {
-					if (this.textPreviews.has(rowId)) throw new Error("Cannot delete or replace a body used in this transaction");
+					if (this.textPreviews.has(textKey({ captionId: rowId }))) throw new Error("Cannot delete or replace a Text used in this transaction");
 					if (nextText !== void 0 && (this.removedCaptionTexts.has(rowId) || initial?.text !== void 0 && nextText !== initial.text)) throw new Error("Edit Caption text through the Text adapter");
 					if (initial?.text !== void 0 && nextText === void 0) this.removedCaptionTexts.add(rowId);
 				}
@@ -4208,7 +4298,7 @@ var TransactionDraft = class {
 			let fields = rows.get(rowId);
 			if (!fields) rows.set(rowId, fields = /* @__PURE__ */ new Set());
 			if (path.length === 1) for (const field of [...Object.keys(initial ?? {}), ...Object.keys(isRecord(next) ? next : {})]) {
-				if ((field === "body" && table === "audio-script-entities" || field === "text" && table === "caption-entities") && initial && isRecord(next) && next[field] === initial[field]) continue;
+				if (initial && isRecord(next) && (field === "segments" && table === "audio-script-entities" && sameSegments(next[field], initial[field]) || field === "text" && table === "caption-entities" && next[field] === initial[field])) continue;
 				fields.add(field);
 			}
 			else fields.add(path[2] ?? key);
@@ -4219,6 +4309,9 @@ var TransactionDraft = class {
 		}
 	}
 };
+function sameSegments(a, b) {
+	return JSON.stringify(a) === JSON.stringify(b);
+}
 //#endregion
 //#region src/dsl/document/draft-operations.ts
 function nodeIndex(nodes) {
@@ -4314,8 +4407,17 @@ function applyMutations(tx, mutations) {
 			case "delete-relation":
 				delete rows[m.table][m.id];
 				break;
+			case "insert-segment":
+				tx.insertSegment(m.audioScriptId, m.segmentId, m.text, m.index);
+				break;
+			case "move-segment":
+				tx.moveSegment(m.audioScriptId, m.segmentId, m.index);
+				break;
+			case "delete-segment":
+				tx.deleteSegment(m.audioScriptId, m.segmentId);
+				break;
 			case "edit-text":
-				tx.editText(m.audioScriptId, m.edits, m.expectedBodyId);
+				tx.editText(m.audioScriptId, m.segmentId, m.edits, m.expectedBodyId);
 				break;
 			case "edit-caption-text":
 				tx.editCaptionText(m.captionId, m.edits, m.expectedBodyId);
@@ -4323,6 +4425,9 @@ function applyMutations(tx, mutations) {
 			case "set-caption-content":
 				tx.setCaptionContent(m);
 				break;
+			case "set-phoneme":
+				tx.setPhoneme(m);
+				break;
 			default: changeTree(tx.draft["contain-relations"], m, tx.draft);
 		}
 	}
@@ -4375,12 +4480,15 @@ function applyTextEdits(text, edits) {
 }
 //#endregion
 //#region src/dsl/document/validation.ts
-/** A caption selection by the DSL's rules, which must also select a nonempty run of the live body. */
-function validateSelection(doc, state, captionId, value) {
-	assertReferenceAdmitted(state, "caption-entities", captionId, "selection", value);
-	const resolved = resolveContent(doc, value);
-	if (!resolved || resolved.start === resolved.end) throw new Error("Expected a nonempty range in the current body");
-	captionContentCodec.encode(value);
+/**
+* A selection the draft wrote whole, by the DSL's rules: its segment must exist and what it
+* selects must not be empty. A run's Cursors must lie on that segment's live Text.
+*/
+function validateSelection(doc, state, table, id, value) {
+	assertReferenceAdmitted(state, table, id, selectionField[table], value);
+	const selection = value;
+	if (!(selection.start === void 0 ? requireSegment(state, selection.audioScriptId, selection.segmentId).text : resolveContent(doc, selection)?.text)) throw new Error("Expected a nonempty selection of the current segment");
+	(table === "caption-entities" ? captionContentCodec : phonemeCodec).encode(selection);
 }
 /**
 * The DSL's admission of the whole draft, then what only this document can check: containment
@@ -4412,8 +4520,12 @@ function validateDraft(tracked) {
 }
 //#endregion
 //#region src/dsl/document/transact.ts
-function captionRow(doc, captionId) {
-	const row = doc.getMap("caption-entities").get(captionId);
+const codecs = {
+	"caption-entities": captionContentCodec,
+	"phonetic-script-entities": phonemeCodec
+};
+function entityRow(doc, table, id) {
+	const row = doc.getMap(table).get(id);
 	return isContainer(row) && row.kind() === "Map" ? row : void 0;
 }
 /** Serialization happens before preparing any writes, so invalid/cyclic audit data cannot partly commit. */
@@ -4434,8 +4546,8 @@ const nativeHookSupport = /* @__PURE__ */ new WeakMap();
 /**
 * Every write relies on the patched Mirror's `applyNative` (patches/loro-mirror@2.3.3.patch). An
 * unpatched Mirror ignores the option and commits only the draft, silently dropping Text, ranges
-* and Caption selections. The patch runs the hook even for an empty draft without committing, so
-* probe each Mirror class once on a scratch document and refuse before the first real write.
+* and selections. The patch runs the hook even for an empty draft without committing, so probe
+* each Mirror class once on a scratch document and refuse before the first real write.
 */
 function assertNativeHook(mirror) {
 	const Impl = mirror.constructor;
@@ -4469,57 +4581,75 @@ function runTransaction(doc, mirror, edit, options, setReject) {
 	mirror.setState((raw) => {
 		const tracked = new TransactionDraft(raw, before);
 		setReject((error) => tracked.fail(error));
-		const useBody = (scriptId, caption = false) => {
-			assertId(scriptId);
-			const entity = requireEntity(tracked.draft, caption ? "caption-entities" : "audio-script-entities", scriptId);
-			if (tracked.deleted.has(scriptId)) throw new Error("Cannot use a deleted body");
-			const body = entity[caption ? "text" : "body"];
-			if (typeof body !== "string") throw new Error("Missing Text body");
-			if (caption && tracked.removedCaptionTexts.has(scriptId)) throw new Error("Cannot edit a removed Text");
-			tracked.textPreviews.set(scriptId, body);
-			return body;
+		/** The current text of a Caption's own Text or a script segment, as this transaction leaves it so far. */
+		const useText = (target) => {
+			let text;
+			if ("captionId" in target) {
+				assertId(target.captionId);
+				const entity = requireEntity(tracked.draft, "caption-entities", target.captionId);
+				if (tracked.deleted.has(target.captionId)) throw new Error("Cannot use a deleted Text");
+				if (tracked.removedCaptionTexts.has(target.captionId)) throw new Error("Cannot edit a removed Text");
+				text = entity.text;
+			} else {
+				assertId(target.audioScriptId);
+				if (tracked.deleted.has(target.audioScriptId)) throw new Error("Cannot use a deleted Text");
+				text = requireSegment(tracked.draft, target.audioScriptId, target.segmentId).text;
+			}
+			if (typeof text !== "string") throw new Error("Missing Text");
+			tracked.textPreviews.set(textKey(target), text);
+			return text;
+		};
+		const textBefore = (target) => {
+			if ("captionId" in target) return before["caption-entities"][target.captionId]?.text;
+			return before["audio-script-entities"][target.audioScriptId]?.segments?.find((segment) => segment.id === target.segmentId)?.text;
 		};
-		const flushCaptions = () => {
-			for (const id of tracked.dirtyCaptions) {
-				const row = raw["caption-entities"][id];
+		const flushSelections = () => {
+			for (const [table, id] of tracked.dirtySelections()) {
+				const row = raw[table][id];
 				if (!row) continue;
-				const value = row.selection;
-				const old = before["caption-entities"][id]?.selection;
+				const field = selectionField[table];
+				const value = row[field];
+				const old = before[table][id]?.[field];
 				if (isInvalidAtomic(value)) {
 					if (isInvalidAtomic(old) && old.raw === value.raw) continue;
-					throw new Error("Invalid Caption selection");
-				}
-				if (value !== void 0) {
-					validateSelection(doc, raw, id, value);
-					useBody(value.audioScriptId);
+					throw new Error(`Invalid ${field}`);
 				}
+				if (value !== void 0) validateSelection(doc, raw, table, id, value);
 				steps.push({
-					type: "caption",
+					type: "selection",
+					table,
 					id,
 					value: copyInput(value)
 				});
 			}
-			tracked.dirtyCaptions.clear();
+			tracked.clearDirtySelections();
 		};
-		const editBody = (scriptId, edits, expectedBodyId, caption) => {
+		const editBody = (target, edits, expectedBodyId) => {
 			tracked.run(() => {
-				flushCaptions();
-				const body = useBody(scriptId, caption);
-				if (!bodies.has(scriptId)) bodies.set(scriptId, liveText(doc, scriptId, caption));
-				const native = bodies.get(scriptId);
+				flushSelections();
+				const text = useText(target);
+				const key = textKey(target);
+				if (!bodies.has(key)) bodies.set(key, liveText(doc, target));
+				const native = bodies.get(key);
 				if (expectedBodyId !== void 0 && native?.id !== expectedBodyId) throw new Error("Stale or missing Text body");
-				if ((caption ? before["caption-entities"][scriptId]?.text : before["audio-script-entities"][scriptId]?.body) !== void 0 && !native) throw new Error("Missing live Text body");
-				const prepared = prepareTextEdits(body, copyInput(edits));
-				tracked.textPreviews.set(scriptId, prepared.body);
+				if (textBefore(target) !== void 0 && !native) throw new Error("Missing live Text body");
+				const prepared = prepareTextEdits(text, copyInput(edits));
+				tracked.textPreviews.set(key, prepared.text);
 				steps.push({
 					type: "text",
-					scriptId,
-					caption,
+					target,
 					expectedBodyId: native?.id,
 					edits: prepared.edits
 				});
 			});
 		};
+		/** A script's segments as this transaction leaves them; never a Mirror-tracked write. */
+		const segmentsOf = (audioScriptId) => {
+			assertId(audioScriptId);
+			requireEntity(tracked.draft, "audio-script-entities", audioScriptId);
+			if (tracked.deleted.has(audioScriptId)) throw new Error("Cannot change a deleted AudioScript");
+			return raw["audio-script-entities"][audioScriptId].segments;
+		};
 		const tx = Object.freeze({
 			get draft() {
 				return tracked.run(() => tracked.draft);
@@ -4527,29 +4657,95 @@ function runTransaction(doc, mirror, edit, options, setReject) {
 			applyMutations(mutations) {
 				tracked.run(() => applyMutations(tx, mutations));
 			},
-			editText(scriptId, edits, expectedBodyId) {
-				editBody(scriptId, edits, expectedBodyId, false);
+			editText(audioScriptId, segmentId, edits, expectedBodyId) {
+				editBody({
+					audioScriptId,
+					segmentId
+				}, edits, expectedBodyId);
 			},
 			editCaptionText(captionId, edits, expectedBodyId) {
-				editBody(captionId, edits, expectedBodyId, true);
+				editBody({ captionId }, edits, expectedBodyId);
+			},
+			insertSegment(audioScriptId, segmentId, text, index) {
+				tracked.run(() => {
+					assertId(segmentId);
+					if (typeof text !== "string") throw new Error("Segment text is a string");
+					const segments = segmentsOf(audioScriptId);
+					if (segments.some((segment) => segment.id === segmentId)) throw new Error(`Segment already exists: ${segmentId}`);
+					const at = index ?? segments.length;
+					if (!Number.isSafeInteger(at) || at < 0 || at > segments.length) throw new Error("Invalid segment index");
+					segments.splice(at, 0, {
+						id: segmentId,
+						text
+					});
+				});
+			},
+			moveSegment(audioScriptId, segmentId, index) {
+				tracked.run(() => {
+					const segments = segmentsOf(audioScriptId);
+					const from = segments.findIndex((segment) => segment.id === segmentId);
+					if (from < 0) throw new Error(`Missing segment ${segmentId} of ${audioScriptId}`);
+					if (!Number.isSafeInteger(index) || index < 0 || index >= segments.length) throw new Error("Invalid segment index");
+					const [segment] = segments.splice(from, 1);
+					segments.splice(index, 0, segment);
+				});
+			},
+			deleteSegment(audioScriptId, segmentId) {
+				tracked.run(() => {
+					const segments = segmentsOf(audioScriptId);
+					const at = segments.findIndex((segment) => segment.id === segmentId);
+					if (at < 0) throw new Error(`Missing segment ${segmentId} of ${audioScriptId}`);
+					if (tracked.textPreviews.has(textKey({
+						audioScriptId,
+						segmentId
+					}))) throw new Error("Cannot delete a segment whose text this transaction edits");
+					segments.splice(at, 1);
+				});
 			},
 			setCaptionContent(selection) {
 				tracked.run(() => {
-					flushCaptions();
-					const { captionId, audioScriptId, start, end } = selection;
+					flushSelections();
+					const { captionId, audioScriptId, segmentId, start, end } = selection;
 					assertId(captionId);
-					assertCaptionEndpoints(raw, selection);
-					assertCaptionSelection(raw, selection, useBody(audioScriptId));
+					assertCaptionSelection(raw, selection, useText({
+						audioScriptId,
+						segmentId
+					}));
 					steps.push({
 						type: "range",
-						selection: {
-							captionId,
+						table: "caption-entities",
+						id: captionId,
+						range: {
 							audioScriptId,
+							segmentId,
 							start,
 							end
 						}
 					});
 				});
+			},
+			setPhoneme(selection) {
+				tracked.run(() => {
+					flushSelections();
+					const { phoneticScriptId, audioScriptId, segmentId, start, end, phonemeScript } = selection;
+					assertId(phoneticScriptId);
+					assertPhonemeSelection(raw, selection, useText({
+						audioScriptId,
+						segmentId
+					}));
+					steps.push({
+						type: "range",
+						table: "phonetic-script-entities",
+						id: phoneticScriptId,
+						range: {
+							audioScriptId,
+							segmentId,
+							start,
+							end
+						},
+						...phonemeScript === void 0 ? {} : { phonemeScript }
+					});
+				});
 			}
 		});
 		try {
@@ -4562,34 +4758,15 @@ function runTransaction(doc, mirror, edit, options, setReject) {
 				}
 				tracked.assertValid();
 				if (doc.getPendingTxnLength()) throw new Error("Native writes are not allowed during transaction preparation");
-				flushCaptions();
+				flushSelections();
 				for (const id of tracked.entities.get("caption-entities")?.keys() ?? []) {
 					const caption = tracked.draft["caption-entities"][id];
 					if (caption) caption.style ??= {};
 				}
 				validateDraft(tracked);
-				for (const step of steps) if (step.type === "text") requireEntity(raw, step.caption ? "caption-entities" : "audio-script-entities", step.scriptId);
-				const referenceIds = new Set(Object.entries(before["caption-entities"]).filter(([, caption]) => typeof caption === "object" && caption.selection !== void 0).map(([id]) => id));
-				const changedCaptions = /* @__PURE__ */ new Set();
-				for (const [id, fields] of tracked.entities.get("caption-entities") ?? []) if (fields.has("text")) changedCaptions.add(id);
-				for (const step of steps) {
-					if (step.type === "range") {
-						referenceIds.add(step.selection.captionId);
-						changedCaptions.add(step.selection.captionId);
-					}
-					if (step.type === "caption") {
-						changedCaptions.add(step.id);
-						if (step.value) referenceIds.add(step.id);
-						else referenceIds.delete(step.id);
-					}
-					if (step.type === "text" && step.caption) changedCaptions.add(step.scriptId);
-				}
-				for (const id of changedCaptions) if (raw["caption-entities"][id]) assertCaptionSource(raw["caption-entities"][id], referenceIds.has(id));
-				for (const step of steps) if (step.type === "range") assertCaptionEndpoints(raw, step.selection);
-				else if (step.type === "caption" && step.value) assertCaptionEndpoints(raw, {
-					captionId: step.id,
-					audioScriptId: step.value.audioScriptId
-				});
+				for (const step of steps) if (step.type === "text") if ("captionId" in step.target) requireEntity(raw, "caption-entities", step.target.captionId);
+				else requireSegment(raw, step.target.audioScriptId, step.target.segmentId);
+				assertSelectionsAfter(raw, before, tracked, steps);
 				for (const [id, fields] of tracked.entities.get("caption-entities") ?? []) {
 					const row = raw["caption-entities"][id];
 					if (fields.has("text") && row?.text !== void 0 && before["caption-entities"][id]?.text === void 0) {
@@ -4597,12 +4774,15 @@ function runTransaction(doc, mirror, edit, options, setReject) {
 						delete row.text;
 					}
 				}
-				for (const [id, fields] of tracked.entities.get("caption-entities") ?? []) {
-					const row = raw["caption-entities"][id];
-					if (!row || !fields.has("selection")) continue;
-					const old = before["caption-entities"][id];
-					if (old?.selection === void 0) delete row.selection;
-					else row.selection = old.selection;
+				for (const table of Object.keys(selectionField)) {
+					const field = selectionField[table];
+					for (const [id, fields] of tracked.entities.get(table) ?? []) {
+						const row = raw[table][id];
+						if (!row || !fields.has(field)) continue;
+						const old = before[table][id]?.[field];
+						if (old === void 0) delete row[field];
+						else row[field] = old;
+					}
 				}
 			});
 		} finally {
@@ -4612,45 +4792,104 @@ function runTransaction(doc, mirror, edit, options, setReject) {
 	}, {
 		...options,
 		applyNative: (doc) => {
-			const bodyFor = (scriptId, caption = false) => {
-				let text = bodies.get(scriptId);
+			const textFor = (target) => {
+				const key = textKey(target);
+				let text = bodies.get(key);
 				if (!text) {
-					text = liveText(doc, scriptId, caption);
-					bodies.set(scriptId, text);
+					text = liveText(doc, target);
+					bodies.set(key, text);
 				}
-				if (!text || text.isDeleted()) throw new Error("Prepared Text body is missing");
+				if (!text || text.isDeleted()) throw new Error("Prepared Text is missing");
 				return text;
 			};
 			for (const [id, value] of newCaptionTexts) {
-				const body = doc.getMap("caption-entities").get(id).setContainer("text", new LoroText());
-				if (value.length) body.insert(0, value);
-				bodies.set(id, body);
+				const text = doc.getMap("caption-entities").get(id).setContainer("text", new LoroText());
+				if (value.length) text.insert(0, value);
+				bodies.set(textKey({ captionId: id }), text);
 			}
-			for (const step of steps) if (step.type === "text") {
-				const text = bodyFor(step.scriptId, step.caption);
-				if (step.expectedBodyId && text.id !== step.expectedBodyId) throw new Error("Stale Text body");
-				applyTextEdits(text, step.edits);
-			} else if (step.type === "range") {
-				const { captionId, audioScriptId, start, end } = step.selection;
-				const text = bodyFor(audioScriptId);
-				const from = text?.getCursor(start, 0), to = text?.getCursor(end, 0);
-				const row = captionRow(doc, captionId);
-				if (!from || !to || !row) throw new Error("Unable to construct prepared range");
-				row.set("selection", captionContentCodec.encode({
-					audioScriptId,
-					start: from,
-					end: to
-				}));
-			} else {
-				const row = captionRow(doc, step.id);
-				if (step.value !== void 0) {
-					if (!row) throw new Error("Prepared Caption is missing");
-					row.set("selection", captionContentCodec.encode(step.value));
-				} else if (row?.get("selection") !== void 0) row.delete("selection");
+			for (const step of steps) {
+				if (step.type === "text") {
+					const text = textFor(step.target);
+					if (step.expectedBodyId && text.id !== step.expectedBodyId) throw new Error("Stale Text body");
+					applyTextEdits(text, step.edits);
+					continue;
+				}
+				const row = entityRow(doc, step.table, step.id);
+				const field = selectionField[step.table];
+				if (step.type === "range") {
+					const { audioScriptId, segmentId, start, end } = step.range;
+					const value = {
+						audioScriptId,
+						segmentId
+					};
+					if (start !== void 0 && end !== void 0) {
+						const text = textFor({
+							audioScriptId,
+							segmentId
+						});
+						value.start = text.getCursor(start, 0);
+						value.end = text.getCursor(end, 0);
+						if (!value.start || !value.end) throw new Error("Unable to construct prepared range");
+					}
+					if (step.phonemeScript !== void 0) value.phonemeScript = step.phonemeScript;
+					if (!row) throw new Error("Unable to construct prepared range");
+					row.set(field, codecs[step.table].encode(value));
+				} else if (step.value !== void 0) {
+					if (!row) throw new Error("Prepared selection owner is missing");
+					row.set(field, codecs[step.table].encode(step.value));
+				} else if (row?.get(field) !== void 0) row.delete(field);
 			}
 		}
 	});
 }
+/**
+* What only the whole write can say about selections, with the native steps counted in order:
+* each Caption keeps one source of words, each selection's segment still exists, and every
+* PhoneticScript the write touched is written for a segment and stores its text as a file.
+*/
+function assertSelectionsAfter(raw, before, tracked, steps) {
+	const selected = new Set(Object.entries(before["caption-entities"]).filter(([, caption]) => typeof caption === "object" && caption.selection !== void 0).map(([id]) => id));
+	const changedCaptions = /* @__PURE__ */ new Set();
+	for (const [id, fields] of tracked.entities.get("caption-entities") ?? []) if (fields.has("text")) changedCaptions.add(id);
+	const phonemes = /* @__PURE__ */ new Map();
+	for (const step of steps) {
+		if (step.type === "text") {
+			if ("captionId" in step.target) changedCaptions.add(step.target.captionId);
+			continue;
+		}
+		const value = step.type === "range" ? {
+			...step.range,
+			...step.phonemeScript === void 0 ? {} : { phonemeScript: step.phonemeScript }
+		} : step.value;
+		if (step.table === "caption-entities") {
+			changedCaptions.add(step.id);
+			if (value) selected.add(step.id);
+			else selected.delete(step.id);
+			if (value) assertCaptionEndpoints(raw, {
+				captionId: step.id,
+				...value
+			});
+		} else {
+			phonemes.set(step.id, value);
+			if (value) assertPhonemeEndpoints(raw, {
+				phoneticScriptId: step.id,
+				...value
+			});
+		}
+	}
+	for (const id of changedCaptions) if (raw["caption-entities"][id]) assertCaptionSource(raw["caption-entities"][id], selected.has(id));
+	const phonetics = new Set([...tracked.entities.get("phonetic-script-entities")?.keys() ?? [], ...phonemes.keys()]);
+	for (const id of phonetics) {
+		const row = raw["phonetic-script-entities"][id];
+		if (!row) continue;
+		const phoneme = phonemes.has(id) ? phonemes.get(id) : row.phoneme;
+		const old = before["phonetic-script-entities"][id];
+		assertPhoneticScript(old, {
+			...row,
+			phoneme: isRecord(phoneme) ? phoneme : void 0
+		});
+	}
+}
 //#endregion
 //#region src/dsl/document/medeo-dsl.ts
 const contentRoots = new Set([
@@ -4741,17 +4980,23 @@ var MedeoDsl = class MedeoDsl {
 	applyMutations(mutations, audit) {
 		this.transact((tx) => tx.applyMutations(mutations), audit);
 	}
-	editText(scriptId, expectedBodyId, edits, audit) {
-		this.transact((tx) => tx.editText(scriptId, edits, expectedBodyId), audit);
+	editText(scriptId, segmentId, expectedBodyId, edits, audit) {
+		this.transact((tx) => tx.editText(scriptId, segmentId, edits, expectedBodyId), audit);
 	}
 	editCaptionText(captionId, expectedBodyId, edits, audit) {
 		this.transact((tx) => tx.editCaptionText(captionId, edits, expectedBodyId), audit);
 	}
-	/** Local ranges exclude insertions at their start and include insertions at their end. K01 remains deferred. */
-	textRange(scriptId, start, end) {
+	/**
+	* A run inside one segment as native Cursors. Local ranges exclude insertions at their start and
+	* include insertions at their end. K01 remains deferred.
+	*/
+	textRange(scriptId, segmentId, start, end) {
 		this.assertActive();
-		const text = liveText(this.doc, scriptId);
-		if (!text) throw new Error("Missing AudioScript body");
+		const text = liveText(this.doc, {
+			audioScriptId: scriptId,
+			segmentId
+		});
+		if (!text) throw new Error("Missing AudioScript segment");
 		assertTextOffset(text.toString(), start);
 		assertTextOffset(text.toString(), end);
 		if (start >= end) throw new Error("Expected a nonempty range");
@@ -4759,6 +5004,7 @@ var MedeoDsl = class MedeoDsl {
 		if (!from || !to) throw new Error("Unable to construct text range");
 		return {
 			audioScriptId: scriptId,
+			segmentId,
 			start: from,
 			end: to
 		};
@@ -4820,16 +5066,28 @@ function voiceoverPlan(s, input, result) {
 		id: result.mediaId,
 		value: input.media
 	}];
-	if (input.script.body !== void 0) mutations.push({
-		type: "create-entity",
-		table: "audio-script-entities",
-		id: result.audioScriptId,
-		value: {
-			body: input.script.body,
-			..."defaultLanguage" in input.script && input.script.defaultLanguage !== void 0 ? { defaultLanguage: input.script.defaultLanguage } : {}
-		}
-	});
-	else if (!s["audio-script-entities"][result.audioScriptId]) throw new Error("Missing AudioScript");
+	let segmentIds;
+	if (input.script.segments !== void 0) {
+		if (result.segmentIds.length !== input.script.segments.length) throw new Error("Expected one id per segment");
+		const segments = input.script.segments;
+		segmentIds = result.segmentIds;
+		mutations.push({
+			type: "create-entity",
+			table: "audio-script-entities",
+			id: result.audioScriptId,
+			value: {
+				segments: segments.map((text, i) => ({
+					id: segmentIds[i],
+					text
+				})),
+				..."defaultLanguage" in input.script && input.script.defaultLanguage !== void 0 ? { defaultLanguage: input.script.defaultLanguage } : {}
+			}
+		});
+	} else {
+		const script = s["audio-script-entities"][result.audioScriptId];
+		if (!script) throw new Error("Missing AudioScript");
+		segmentIds = script.segments.map((segment) => segment.id);
+	}
 	const phoneticScriptId = input.script.phoneticScriptId;
 	if (phoneticScriptId !== void 0 && s["phonetic-script-entities"][phoneticScriptId]?.phoneme.audioScriptId !== result.audioScriptId) throw new Error("The PhoneticScript must name this AudioScript as its base");
 	validateWindow({
@@ -4912,6 +5170,8 @@ function voiceoverPlan(s, input, result) {
 		}, input.media.durationMs);
 		const previous = s["clip-entities"][clipId];
 		if (previous && !oldCaptions.includes(clipId)) throw new Error("Cannot overwrite another use’s Caption Clip");
+		const segmentId = segmentIds[caption.segment];
+		if (segmentId === void 0) throw new Error(`Missing segment ${caption.segment} of the voiceover's script`);
 		const previousSource = previous?.source;
 		const style = caption.style ?? (previousSource ? s["caption-entities"][previousSource]?.style : void 0);
 		mutations.push({
@@ -4923,8 +5183,11 @@ function voiceoverPlan(s, input, result) {
 			type: "set-caption-content",
 			captionId,
 			audioScriptId: result.audioScriptId,
-			start: caption.range.start,
-			end: caption.range.end
+			segmentId,
+			...caption.range === void 0 ? {} : {
+				start: caption.range.start,
+				end: caption.range.end
+			}
 		});
 		const anchor = {
 			to: result.clipId,
@@ -5075,30 +5338,76 @@ var DslEditor = class {
 			patch: { volume }
 		}], options);
 	}
-	createAudioScript(body, defaultLanguage, options) {
+	/** A script of these segments, in order; each segment gets its own id. */
+	createAudioScript(segments, defaultLanguage, options) {
 		const id = this.idFactory("audio-script");
+		const value = {
+			segments: segments.map((text) => ({
+				id: this.idFactory("segment"),
+				text
+			})),
+			...defaultLanguage === void 0 ? {} : { defaultLanguage }
+		};
 		this.write("createAudioScript", {
 			id,
-			body,
-			defaultLanguage
+			...value
 		}, [{
 			type: "create-entity",
 			table: "audio-script-entities",
 			id,
-			value: {
-				body,
-				...defaultLanguage === void 0 ? {} : { defaultLanguage }
-			}
+			value
 		}], options);
 		return id;
 	}
-	editAudioScript(scriptId, expectedBodyId, edits, options) {
-		this.dsl.editText(scriptId, expectedBodyId, edits, audit("editAudioScript", {
+	editAudioScript(scriptId, segmentId, expectedBodyId, edits, options) {
+		this.dsl.editText(scriptId, segmentId, expectedBodyId, edits, audit("editAudioScript", {
 			scriptId,
+			segmentId,
 			expectedBodyId,
 			edits
 		}, options));
 	}
+	/** Adds a segment at `index`, the end by default, and returns its id. */
+	insertScriptSegment(scriptId, text, index, options) {
+		const segmentId = this.idFactory("segment");
+		this.write("insertScriptSegment", {
+			scriptId,
+			segmentId,
+			text,
+			index
+		}, [{
+			type: "insert-segment",
+			audioScriptId: scriptId,
+			segmentId,
+			text,
+			...index === void 0 ? {} : { index }
+		}], options);
+		return segmentId;
+	}
+	/** `index` is measured after removing the segment. */
+	moveScriptSegment(scriptId, segmentId, index, options) {
+		this.write("moveScriptSegment", {
+			scriptId,
+			segmentId,
+			index
+		}, [{
+			type: "move-segment",
+			audioScriptId: scriptId,
+			segmentId,
+			index
+		}], options);
+	}
+	/** Captions and PhoneticScripts selecting the segment keep their selection, which then shows nothing (D72). */
+	deleteScriptSegment(scriptId, segmentId, options) {
+		this.write("deleteScriptSegment", {
+			scriptId,
+			segmentId
+		}, [{
+			type: "delete-segment",
+			audioScriptId: scriptId,
+			segmentId
+		}], options);
+	}
 	createCaption(input, options) {
 		const id = this.idFactory("caption");
 		this.write("createCaption", {
@@ -5115,18 +5424,16 @@ var DslEditor = class {
 		}, ...input.text !== void 0 ? [] : [{
 			type: "set-caption-content",
 			captionId: id,
-			audioScriptId: input.audioScriptId,
-			start: input.start,
-			end: input.end
+			...segmentRange(input)
 		}]], options);
 		return id;
 	}
-	setCaptionContent(captionId, audioScriptId, start, end, options) {
+	/** Show a segment of a script, or a run of it, in place of the Caption's own text. */
+	setCaptionContent(captionId, selection, options) {
+		const range = segmentRange(selection);
 		this.write("setCaptionContent", {
 			captionId,
-			audioScriptId,
-			start,
-			end
+			...range
 		}, [{
 			type: "patch-entity",
 			table: "caption-entities",
@@ -5135,9 +5442,7 @@ var DslEditor = class {
 		}, {
 			type: "set-caption-content",
 			captionId,
-			audioScriptId,
-			start,
-			end
+			...range
 		}], options);
 	}
 	setCaptionText(captionId, text, options) {
@@ -5263,6 +5568,7 @@ var DslEditor = class {
 			clipId: input.clipId ?? this.idFactory("clip"),
 			mediaId: resourceId("media", input.media.system, input.media.key),
 			audioScriptId: this.voiceoverScriptId(input.script),
+			segmentIds: input.script.segments?.map(() => this.idFactory("segment")) ?? [],
 			captionClipIds: input.captions.map((c) => c.clipId ?? this.idFactory("clip")),
 			captionIds: input.captions.map(() => this.idFactory("caption"))
 		};
@@ -5280,17 +5586,14 @@ var DslEditor = class {
 		return phonetic.phoneme.audioScriptId;
 	}
 	/**
-	* Write a PhoneticScript: create one for `audioScriptId`, or, given `phoneticScriptId`, replace its
-	* base and pronunciation together. A `phonemeScript` is stored as a Memota text file first, and
-	* the file's facts are written with it; an unchanged one keeps its file. Recordings already spoken
-	* from it keep their derived-from.
+	* Write a PhoneticScript: create one for a segment of a script, or a run of it, or, given
+	* `phoneticScriptId`, replace its selection and what to synthesize together (D181). A
+	* `phonemeScript` is stored as a Memota text file first, and the file's facts are written with
+	* it; an unchanged one keeps its file. Recordings already spoken from it keep their derived-from.
 	*/
 	async setPhoneticScript(input, options) {
 		const id = input.phoneticScriptId ?? this.idFactory("phonetic-script");
-		const phoneme = {
-			audioScriptId: input.audioScriptId,
-			...input.phonemeScript === void 0 ? {} : { phonemeScript: input.phonemeScript }
-		};
+		const range = segmentRange(input);
 		const file = input.phonemeScript === void 0 ? {} : await this.fileFor(id, input.phonemeScript);
 		const exists = Object.hasOwn(this.dsl.getSnapshot()["phonetic-script-entities"], id);
 		const cleared = {
@@ -5299,9 +5602,11 @@ var DslEditor = class {
 			storageKey: void 0,
 			digest: void 0
 		};
+		const phonemeScript = input.phonemeScript === void 0 ? {} : { phonemeScript: input.phonemeScript };
 		this.write("setPhoneticScript", {
 			id,
-			phoneme,
+			...range,
+			...phonemeScript,
 			...cleared,
 			...file
 		}, [exists ? {
@@ -5309,10 +5614,6 @@ var DslEditor = class {
 			table: "phonetic-script-entities",
 			id,
 			patch: {
-				phoneme: {
-					...phoneme,
-					phonemeScript: input.phonemeScript
-				},
 				...cleared,
 				...file
 			}
@@ -5320,10 +5621,12 @@ var DslEditor = class {
 			type: "create-entity",
 			table: "phonetic-script-entities",
 			id,
-			value: {
-				phoneme,
-				...file
-			}
+			value: file
+		}, {
+			type: "set-phoneme",
+			phoneticScriptId: id,
+			...range,
+			...phonemeScript
 		}], options);
 		return id;
 	}
@@ -5392,6 +5695,15 @@ var DslEditor = class {
 		this.dsl.applyMutations(mutations, audit(action, payload, options));
 	}
 };
+/** A selection as a write states it: a run only when both its ends are given. */
+function segmentRange(input) {
+	return {
+		audioScriptId: input.audioScriptId,
+		segmentId: input.segmentId,
+		...input.start === void 0 ? {} : { start: input.start },
+		...input.end === void 0 ? {} : { end: input.end }
+	};
+}
 function audit(action, payload, options) {
 	return {
 		semantic_op: action,
@@ -5467,35 +5779,39 @@ function spokenScripts(snapshot, recordingId) {
 /** Every AudioScript in the document with its PhoneticScripts, the Speech Clips saying it and the Caption Clips showing it. */
 function toAudioScriptViews(snapshot) {
 	const views = /* @__PURE__ */ new Map();
-	const bodies = /* @__PURE__ */ new Map();
-	for (const [id, script] of Object.entries(snapshot["audio-script-entities"])) {
-		const rich = snapshot.richText[id];
-		const text = rich ? langsFromDelta(rich.delta) : {
-			body: script.body,
-			langs: []
-		};
-		bodies.set(id, text.body);
-		views.set(id, {
-			id,
-			segments: lines(text.body).map((line) => {
-				const language = lineLanguage(text, line);
-				return language === void 0 ? { text: line.text } : {
-					text: line.text,
-					language
-				};
-			}),
-			sourceMediaIds: [],
-			phoneticScripts: [],
-			speeches: [],
-			captions: []
-		});
-	}
+	const segmentIndex = (scriptId, segmentId) => {
+		const index = snapshot["audio-script-entities"][scriptId]?.segments.findIndex((s) => s.id === segmentId) ?? -1;
+		return index < 0 ? void 0 : index;
+	};
+	for (const [id, script] of Object.entries(snapshot["audio-script-entities"])) views.set(id, {
+		id,
+		segments: script.segments.map((segment) => {
+			const rich = snapshot.richText[segment.id];
+			const language = rich ? markedLanguage(langsFromDelta(rich.delta)) : void 0;
+			return language === void 0 ? {
+				id: segment.id,
+				text: segment.text
+			} : {
+				id: segment.id,
+				text: segment.text,
+				language
+			};
+		}),
+		sourceMediaIds: [],
+		phoneticScripts: [],
+		speeches: [],
+		captions: []
+	});
 	for (const r of Object.values(snapshot["derived-from-relations"])) if (snapshot["media-entities"][r.sourceId]) views.get(r.entityId)?.sourceMediaIds.push(r.sourceId);
 	for (const [id, phonetic] of Object.entries(snapshot["phonetic-script-entities"])) {
-		const { phonemeScript } = phonetic.phoneme;
+		const { audioScriptId, segmentId, phonemeScript } = phonetic.phoneme;
 		const { digest } = phonetic;
-		views.get(phonetic.phoneme.audioScriptId)?.phoneticScripts.push({
+		const segment = segmentIndex(audioScriptId, segmentId);
+		const text = snapshot.phoneticText[id]?.text;
+		views.get(audioScriptId)?.phoneticScripts.push({
 			id,
+			...segment === void 0 ? {} : { segment },
+			...text === void 0 ? {} : { text },
 			...phonemeScript === void 0 ? {} : { phonemeScript },
 			...digest === void 0 ? {} : { digest }
 		});
@@ -5530,13 +5846,10 @@ function toAudioScriptViews(snapshot) {
 			continue;
 		}
 		const caption = snapshot["caption-entities"][source];
-		const shown = snapshot.captionText[source];
-		if (!caption?.selection || caption.text !== void 0 || !shown) continue;
-		const scriptId = caption.selection.audioScriptId;
-		const body = bodies.get(scriptId);
-		if (body === void 0) continue;
-		const segment = lines(body).findIndex((line) => shown.start >= line.start && shown.end <= line.end);
-		if (segment >= 0) views.get(scriptId)?.captions.push({
+		if (!caption?.selection || caption.text !== void 0 || !snapshot.captionText[source]) continue;
+		const { audioScriptId, segmentId } = caption.selection;
+		const segment = segmentIndex(audioScriptId, segmentId);
+		if (segment !== void 0) views.get(audioScriptId)?.captions.push({
 			clipId,
 			captionId: source,
 			segment
@@ -5644,7 +5957,7 @@ function toVideoDraft(snapshot) {
 					audio_storage_key: source.storageKey,
 					origin_speech_id: source.key,
 					volume: clip.volume,
-					audio_script: spokenText(snapshot, sourceId, captionIds),
+					audio_script: spokenText(snapshot, sourceId),
 					voice: voice ? {
 						id: voice.key,
 						name: voice.name
@@ -5730,29 +6043,17 @@ function requireWholeAudio(media, clip, fail) {
 	if (media.kind === "image" || clip.inMs !== 0 || clip.outMs !== media.durationMs || clip.rate !== 1) fail("VideoDraft audio has no trim or speed fields");
 }
 /**
-* The words a speech Clip says. A script voiced in several takes is said one part per take, and
-* the Captions following the Clip select that part: its words are the segments they select, in
-* script order. A Clip no Caption of its script follows says the whole script.
+* What a recording says: what the selection of the PhoneticScript it was spoken from holds now,
+* or, for one derived from a script directly, the whole script, one segment per line.
 */
-function spokenText(snapshot, recordingId, captionClipIds) {
-	const scripts = new Set(spokenScripts(snapshot, recordingId).map((spoken) => spoken.audioScriptId));
-	if (scripts.size > 1) projectionError("ambiguous-script", `media-entities/${recordingId}`, "Multiple scripts cannot be silently concatenated into audio_script");
-	const id = [...scripts][0];
-	if (!id) return "";
-	const body = snapshot.richText[id]?.plainText ?? "";
-	const segments = lines(body);
-	const said = /* @__PURE__ */ new Set();
-	for (const clipId of captionClipIds) {
-		const captionId = snapshot["clip-entities"][clipId]?.source;
-		const caption = captionId === void 0 ? void 0 : snapshot["caption-entities"][captionId];
-		const shown = captionId === void 0 ? void 0 : snapshot.captionText[captionId];
-		if (caption?.selection?.audioScriptId !== id || caption.text !== void 0 || !shown) continue;
-		segments.forEach((line, index) => {
-			if (shown.start < Math.max(line.end, line.start + 1) && shown.end > line.start) said.add(index);
-		});
-	}
-	if (!said.size) return body;
-	return [...said].sort((a, b) => a - b).map((index) => segments[index].text).join("\n");
+function spokenText(snapshot, recordingId) {
+	const spoken = spokenScripts(snapshot, recordingId);
+	if (new Set(spoken.map((s) => s.phoneticScriptId ?? s.audioScriptId)).size > 1) projectionError("ambiguous-script", `media-entities/${recordingId}`, "Multiple scripts cannot be silently concatenated into audio_script");
+	const [first] = spoken;
+	if (!first) return "";
+	if (first.phoneticScriptId !== void 0) return snapshot.phoneticText[first.phoneticScriptId]?.text ?? "";
+	const script = snapshot["audio-script-entities"][first.audioScriptId];
+	return script ? script.segments.map((segment) => segment.text).join("\n") : "";
 }
 //#endregion
 //#region src/dsl/projection/from-video-draft.ts
@@ -5894,16 +6195,16 @@ function prepare(input, options) {
 		let script = scriptByRecording.get(recordingId);
 		if (script && script.text !== speech.audio_script) fail(`part_library/${id}/audio_script`, "One recording cannot acquire conflicting source scripts during import");
 		if (!script) {
-			script = {
-				id: claim(mint("audio-script")),
-				text: speech.audio_script
-			};
+			script = importedScript(claim(mint("audio-script")), speech.audio_script, () => mint("segment"));
 			scriptByRecording.set(recordingId, script);
 			mutations.push({
 				type: "create-entity",
 				table: "audio-script-entities",
 				id: script.id,
-				value: { body: script.text }
+				value: { segments: script.segments.map(({ id, text }) => ({
+					id,
+					text
+				})) }
 			}, {
 				type: "set-relation",
 				table: "derived-from-relations",
@@ -5951,12 +6252,18 @@ function prepare(input, options) {
 			const text = caption.text;
 			if (typeof text !== "string") fail(`part_library/${captionId}/text`, "Expected Caption text");
 			const start = text.length ? script.text.indexOf(text, cursor) : -1;
-			if (start >= 0) {
-				cursor = start + text.length;
-				selections.set(captionId, {
+			const end = start + text.length;
+			const segment = script.segments.find((s) => start >= s.start && end <= s.start + s.text.length);
+			if (start >= 0 && segment) {
+				cursor = end;
+				selections.set(captionId, text === segment.text ? {
+					audioScriptId: script.id,
+					segmentId: segment.id
+				} : {
 					audioScriptId: script.id,
-					start,
-					end: cursor
+					segmentId: segment.id,
+					start: start - segment.start,
+					end: end - segment.start
 				});
 			} else diagnostics.push({
 				code: "caption-text-unmatched",
@@ -6067,5 +6374,22 @@ function milliseconds(value, path, positive = false, signed = false) {
 	if (typeof value !== "number" || !Number.isSafeInteger(value) || !signed && value < 0 || positive && value <= 0) projectionError("invalid-time", path, "Expected safe integer milliseconds in the valid range");
 	return value;
 }
+function importedScript(id, text, mintSegment) {
+	const segments = [];
+	let start = 0;
+	if (text.length) for (const line of text.split("\n")) {
+		segments.push({
+			id: mintSegment(),
+			text: line,
+			start
+		});
+		start += line.length + 1;
+	}
+	return {
+		id,
+		text,
+		segments
+	};
+}
 //#endregion
 export { DslEditor as a, DslProjectionError as c, createInitialMedeoDsl as i, toVideoDraft as n, MedeoDsl as o, toAudioScriptViews as r, resolveDslLayout as s, fromVideoDraft as t };