@mengine/medeo-client 1.2.1-alpha.0 → 1.2.1-alpha.2

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
@@ -1,4 +1,3 @@
1
- import { A as MoveVideoClipsInput, J as ChangeSpeechScriptInput, K as DeleteBgmInput, N as MoveVideoClipsByAnchorInput, O as ReplaceVideoClipContentInput, Q as AdjustVideoClipVolumeInput, R as MoveSpeechesInput, S as SetBgmInput, T as ReplaceVideoClipSequenceInput, V as DeleteVideoClipsInput, W as DeleteSpeechesInput, Y as ChangeSpeechVoiceInput, b as SetCaptionStyleInput, ct as AddSpeechesInput, et as AdjustVideoClipDurationInput, g as SetVideoClipSpeedShiftInput, it as AdjustBgmVolumeInput, nt as AdjustSpeechVolumeInput, ot as AddVideoClipsInput, v as SetCaptionVisibilityInput } from "./index-InEWX9rk.js";
2
1
  import { InferInputType } from "loro-mirror";
3
2
  import { z } from "zod";
4
3
  import { LoroDoc, PeerID } from "loro-crdt";
@@ -546,6 +545,8 @@ interface TrackItem {
546
545
  part_id: string;
547
546
  time_position: TrackItemTimePosition;
548
547
  fallback_abs_ms?: number | undefined;
548
+ /** Presentation window for this placement; currently used by caption resize. */
549
+ duration_override_ms?: number | undefined;
549
550
  }
550
551
  interface Track {
551
552
  id: string | undefined;
@@ -553,19 +554,6 @@ interface Track {
553
554
  is_hidden: boolean | undefined;
554
555
  items: TrackItem[] | undefined;
555
556
  }
556
- /** The linear speed multiplier of a `speed_shift`, defaulting to 1 (original). */
557
- declare function speedOf(speedShift: SpeedShift | undefined): number;
558
- /**
559
- * A video clip's effective timeline duration, derived from authoritative facts
560
- * (RFC 02, `reference/16` §4, reference/17 §5): the trim window
561
- * `play_out - play_in` divided by the speed multiplier, rounded to integer ms.
562
- * `play_in` / `play_out` are optional in the IDL but every write path sets them
563
- * (defaulting to the whole media), and the legacy-ingest projection backfills the
564
- * window from a legacy `duration_ms` — so an authoritative clip always carries a
565
- * trim window and there is no stored `duration_ms` to fall back to. A clip with
566
- * neither bound yields 0. Speeds the clip up (>1× → shorter) or down (<1×).
567
- */
568
- declare function effectiveVideoClipDurationMs(clip: VideoClipPart): number;
569
557
  /**
570
558
  * Authoritative part union: a flat discriminated union over the four part kinds.
571
559
  * Deliberately *not* the IDL's namespace union (no `$unknown` / `visit`): engine
@@ -949,35 +937,623 @@ type TrackDraft = NonNullable<NonNullable<VideoDocumentDraft['tracks']>[number]>
949
937
  /** A single track item in a draft. */
950
938
  type TrackItemDraft = NonNullable<NonNullable<TrackDraft['items']>[number]>;
951
939
  //#endregion
952
- //#region src/editor/id-gen.d.ts
940
+ //#region src/editor/schemas/add-speeches.d.ts
941
+ /**
942
+ * Add speeches (and their captions). TTS runs upstream; the stable speech /
943
+ * caption parts arrive materialized (see `speech-assets.ts`). The op writes the
944
+ * parts and each speech's `{ mode:'anchored', anchorPartId, offsetMs }` fact
945
+ * verbatim — no write-time host-picking, no cascade (RFC 02 §4). The projection
946
+ * derives absolute positions on read.
947
+ */
948
+ declare const addSpeechesInputSchema: z.ZodObject<{
949
+ speeches: z.ZodArray<z.ZodObject<{
950
+ speech_id: z.ZodString;
951
+ anchor_part_id: z.ZodNullable<z.ZodString>;
952
+ offset_ms: z.ZodNumber;
953
+ audio_storage_key: z.ZodString;
954
+ duration_ms: z.ZodNumber;
955
+ audio_script: z.ZodString;
956
+ volume: z.ZodNumber;
957
+ voice: z.ZodObject<{
958
+ id: z.ZodString;
959
+ name: z.ZodString;
960
+ }, z.core.$strip>;
961
+ origin_speech_id: z.ZodString;
962
+ caption_ids: z.ZodArray<z.ZodString>;
963
+ }, z.core.$strip>>;
964
+ captions: z.ZodArray<z.ZodObject<{
965
+ caption_id: z.ZodString;
966
+ speech_part_id: z.ZodString;
967
+ text: z.ZodString;
968
+ start_ms: z.ZodNumber;
969
+ duration_ms: z.ZodNumber;
970
+ }, z.core.$strip>>;
971
+ }, z.core.$strip>;
972
+ type AddSpeechesInput = z.infer<typeof addSpeechesInputSchema>;
973
+ //#endregion
974
+ //#region src/editor/schemas/add-video-clips.d.ts
953
975
  /**
954
- * Part-id generation, aligned with the online ecosystem.
976
+ * Add video clips to a track. Each clip's duration facts are separated so a
977
+ * single number is never overloaded (RFC 02 / `reference/16` §0b):
955
978
  *
956
- * The authoritative online producers — agent-harness (`@harness/shared`
957
- * `genObjId`) and director.v2 (`common/obj_id.py` `gen_obj_id`) — both mint part
958
- * ids as `` `${prefix}_${ulid()}` ``, and real captured drafts use exactly that
959
- * shape (`clip_…` / `spe_…` / `cap_…` / `bgm_…`, each a 26-char ULID). The engine
960
- * previously emitted `vc_<base36 timestamp><6 random>`, a different prefix AND a
961
- * different encoding — the sole cross-repo id divergence. This module removes it
962
- * by emitting the same `<prefix>_<ULID>` bytes.
979
+ * - `media_duration_ms` is the source media's intrinsic full length (a resource
980
+ * fact, written to the part);
981
+ * - `play_in` / `play_out` are the optional trim window into that media; when
982
+ * omitted the whole media is used (`play_in=0`, `play_out=media_duration_ms`).
963
983
  *
964
- * The ULID is generated inline (Crockford Base32, 48-bit time + 80-bit random)
965
- * rather than pulling the `ulid` npm package: the randomness class matches the
966
- * old generator (both `Math.random`-based) and it keeps `@mengine/medeo-client`
967
- * dependency-free for a purely mechanical id string. Part ids only need to be
968
- * unique and lexicographically time-sortable, which this satisfies.
969
- */
970
- /**
971
- * Online part-id semantic prefixes. `clip` (video clip) is the only value the
972
- * engine currently mints (see `addVideoClips`); the rest are declared so the
973
- * type documents the shared vocabulary and guards against reintroducing the old
974
- * `vc`/`sp`/`cp`/`bg` names. Speech/caption/bgm ids arrive pre-minted in op
975
- * payloads, so the engine never generates them itself.
976
- */
977
- type PartIdPrefix = 'clip' | 'spe' | 'cap' | 'bgm' | 'ti';
978
- declare function generatePartId(prefix: PartIdPrefix): string;
979
- /** Injectable part-id mint; defaults to `generatePartId` on `SemanticEditor`. */
980
- type PartIdFactory = (prefix: PartIdPrefix) => string;
984
+ * The clip's effective timeline duration is derived by the projection from the
985
+ * trim window and `speed_shift` — it is never an input here.
986
+ */
987
+ declare const addVideoClipsInputSchema: z.ZodObject<{
988
+ clips: z.ZodArray<z.ZodObject<{
989
+ media_id: z.ZodString;
990
+ start_ms: z.ZodOptional<z.ZodNumber>;
991
+ media_duration_ms: z.ZodNumber;
992
+ play_in: z.ZodOptional<z.ZodNumber>;
993
+ play_out: z.ZodOptional<z.ZodNumber>;
994
+ track_id: z.ZodOptional<z.ZodString>;
995
+ }, z.core.$strip>>;
996
+ before_clip_id: z.ZodOptional<z.ZodString>;
997
+ after_clip_id: z.ZodOptional<z.ZodString>;
998
+ }, z.core.$strip>;
999
+ type AddVideoClipsInput = z.infer<typeof addVideoClipsInputSchema>;
1000
+ //#endregion
1001
+ //#region src/editor/schemas/adjust-bgm-volume.d.ts
1002
+ declare const adjustBgmVolumeInputSchema: z.ZodObject<{
1003
+ bgm: z.ZodArray<z.ZodObject<{
1004
+ bgm_id: z.ZodString;
1005
+ volume: z.ZodNumber;
1006
+ }, z.core.$strip>>;
1007
+ }, z.core.$strip>;
1008
+ type AdjustBgmVolumeInput = z.infer<typeof adjustBgmVolumeInputSchema>;
1009
+ //#endregion
1010
+ //#region src/editor/schemas/adjust-speech-volume.d.ts
1011
+ declare const adjustSpeechVolumeInputSchema: z.ZodObject<{
1012
+ speeches: z.ZodArray<z.ZodObject<{
1013
+ speech_id: z.ZodString;
1014
+ volume: z.ZodNumber;
1015
+ }, z.core.$strip>>;
1016
+ }, z.core.$strip>;
1017
+ type AdjustSpeechVolumeInput = z.infer<typeof adjustSpeechVolumeInputSchema>;
1018
+ //#endregion
1019
+ //#region src/editor/schemas/adjust-video-clip-duration.d.ts
1020
+ /**
1021
+ * Re-trim existing video clips (the user-facing "adjust duration" gesture is a
1022
+ * trim of the source window). The new `play_in` / `play_out` are the facts; the
1023
+ * effective timeline duration is derived from them and the clip's `speed_shift`,
1024
+ * and the change reflows downstream clips, speeches, and the timeline inside the
1025
+ * op's transaction (no caller-materialized cascade).
1026
+ */
1027
+ declare const adjustVideoClipDurationInputSchema: z.ZodObject<{
1028
+ clips: z.ZodArray<z.ZodObject<{
1029
+ clip_id: z.ZodString;
1030
+ play_in: z.ZodNumber;
1031
+ play_out: z.ZodNumber;
1032
+ }, z.core.$strip>>;
1033
+ }, z.core.$strip>;
1034
+ type AdjustVideoClipDurationInput = z.infer<typeof adjustVideoClipDurationInputSchema>;
1035
+ //#endregion
1036
+ //#region src/editor/schemas/adjust-video-clip-volume.d.ts
1037
+ declare const adjustVideoClipVolumeInputSchema: z.ZodObject<{
1038
+ clips: z.ZodArray<z.ZodObject<{
1039
+ clip_id: z.ZodString;
1040
+ volume: z.ZodNumber;
1041
+ }, z.core.$strip>>;
1042
+ }, z.core.$strip>;
1043
+ type AdjustVideoClipVolumeInput = z.infer<typeof adjustVideoClipVolumeInputSchema>;
1044
+ //#endregion
1045
+ //#region src/editor/schemas/change-speech.d.ts
1046
+ /**
1047
+ * Change a speech's script or voice. Both re-run TTS upstream and return the
1048
+ * regenerated speech / caption parts in the same materialized shape as
1049
+ * `AddSpeeches` (`speech-assets.ts`); the op upserts them by id (the speech part
1050
+ * id is preserved across a re-TTS), re-seats at `start_ms`, and reflows. Old
1051
+ * caption parts no longer owned by the speech are removed via `caption_ids`.
1052
+ */
1053
+ declare const changeSpeechScriptInputSchema: z.ZodObject<{
1054
+ speeches: z.ZodArray<z.ZodObject<{
1055
+ speech_id: z.ZodString;
1056
+ anchor_part_id: z.ZodNullable<z.ZodString>;
1057
+ offset_ms: z.ZodNumber;
1058
+ audio_storage_key: z.ZodString;
1059
+ duration_ms: z.ZodNumber;
1060
+ audio_script: z.ZodString;
1061
+ volume: z.ZodNumber;
1062
+ voice: z.ZodObject<{
1063
+ id: z.ZodString;
1064
+ name: z.ZodString;
1065
+ }, z.core.$strip>;
1066
+ origin_speech_id: z.ZodString;
1067
+ caption_ids: z.ZodArray<z.ZodString>;
1068
+ }, z.core.$strip>>;
1069
+ captions: z.ZodArray<z.ZodObject<{
1070
+ caption_id: z.ZodString;
1071
+ speech_part_id: z.ZodString;
1072
+ text: z.ZodString;
1073
+ start_ms: z.ZodNumber;
1074
+ duration_ms: z.ZodNumber;
1075
+ }, z.core.$strip>>;
1076
+ }, z.core.$strip>;
1077
+ declare const changeSpeechVoiceInputSchema: z.ZodObject<{
1078
+ speeches: z.ZodArray<z.ZodObject<{
1079
+ speech_id: z.ZodString;
1080
+ anchor_part_id: z.ZodNullable<z.ZodString>;
1081
+ offset_ms: z.ZodNumber;
1082
+ audio_storage_key: z.ZodString;
1083
+ duration_ms: z.ZodNumber;
1084
+ audio_script: z.ZodString;
1085
+ volume: z.ZodNumber;
1086
+ voice: z.ZodObject<{
1087
+ id: z.ZodString;
1088
+ name: z.ZodString;
1089
+ }, z.core.$strip>;
1090
+ origin_speech_id: z.ZodString;
1091
+ caption_ids: z.ZodArray<z.ZodString>;
1092
+ }, z.core.$strip>>;
1093
+ captions: z.ZodArray<z.ZodObject<{
1094
+ caption_id: z.ZodString;
1095
+ speech_part_id: z.ZodString;
1096
+ text: z.ZodString;
1097
+ start_ms: z.ZodNumber;
1098
+ duration_ms: z.ZodNumber;
1099
+ }, z.core.$strip>>;
1100
+ }, z.core.$strip>;
1101
+ type ChangeSpeechScriptInput = z.infer<typeof changeSpeechScriptInputSchema>;
1102
+ type ChangeSpeechVoiceInput = z.infer<typeof changeSpeechVoiceInputSchema>;
1103
+ //#endregion
1104
+ //#region src/editor/schemas/delete-bgm.d.ts
1105
+ /**
1106
+ * Remove the document BGM. Pure document edit: clears the bgm lane and removes
1107
+ * the bgm part. Takes no input (a document holds at most one bgm); an empty
1108
+ * object keeps the op signature uniform with the rest.
1109
+ */
1110
+ declare const deleteBgmInputSchema: z.ZodObject<{}, z.core.$strip>;
1111
+ type DeleteBgmInput = z.infer<typeof deleteBgmInputSchema>;
1112
+ //#endregion
1113
+ //#region src/editor/schemas/delete-speeches.d.ts
1114
+ /**
1115
+ * Delete speeches with their captions. Pure document edit (no side effect): the
1116
+ * op removes each speech part, cascade-deletes the captions it owns (via
1117
+ * `caption_ids` / `speech_part_id`), drops their track items, and reflows.
1118
+ */
1119
+ declare const deleteSpeechesInputSchema: z.ZodObject<{
1120
+ speech_ids: z.ZodArray<z.ZodString>;
1121
+ }, z.core.$strip>;
1122
+ type DeleteSpeechesInput = z.infer<typeof deleteSpeechesInputSchema>;
1123
+ //#endregion
1124
+ //#region src/editor/schemas/delete-video-clips.d.ts
1125
+ /**
1126
+ * How a delete handles the anchored subtree (speeches anchored to a deleted clip,
1127
+ * and their captions) — a delete-op policy, not a data-model field (reference/17
1128
+ * §6). `cascade` (default) removes the subtree; `detach` keeps the direct
1129
+ * anchored children, re-pinning them to `absolute` so they stay on the timeline.
1130
+ */
1131
+ declare const anchoredDeletePolicySchema: z.ZodEnum<{
1132
+ cascade: "cascade";
1133
+ detach: "detach";
1134
+ }>;
1135
+ type AnchoredDeletePolicy = z.infer<typeof anchoredDeletePolicySchema>;
1136
+ declare const deleteVideoClipsInputSchema: z.ZodObject<{
1137
+ clip_ids: z.ZodArray<z.ZodString>;
1138
+ on_anchored: z.ZodOptional<z.ZodEnum<{
1139
+ cascade: "cascade";
1140
+ detach: "detach";
1141
+ }>>;
1142
+ }, z.core.$strip>;
1143
+ type DeleteVideoClipsInput = z.infer<typeof deleteVideoClipsInputSchema>;
1144
+ //#endregion
1145
+ //#region src/editor/schemas/move-speeches.d.ts
1146
+ /**
1147
+ * Move speeches in time. Pure document edit: the op re-seats each speech at its
1148
+ * new absolute `start_ms`; the cascade reassigns it to the host video clip,
1149
+ * resolves overlaps, and reflows. Captions follow their speech.
1150
+ */
1151
+ declare const moveSpeechesInputSchema: z.ZodObject<{
1152
+ speeches: z.ZodArray<z.ZodObject<{
1153
+ speech_id: z.ZodString;
1154
+ new_start_ms: z.ZodNumber;
1155
+ }, z.core.$strip>>;
1156
+ }, z.core.$strip>;
1157
+ type MoveSpeechesInput = z.infer<typeof moveSpeechesInputSchema>;
1158
+ //#endregion
1159
+ //#region src/editor/schemas/move-bgm.d.ts
1160
+ declare const moveBgmInputSchema: z.ZodObject<{
1161
+ new_start_ms: z.ZodNumber;
1162
+ }, z.core.$strip>;
1163
+ type MoveBgmInput = z.infer<typeof moveBgmInputSchema>;
1164
+ //#endregion
1165
+ //#region src/editor/schemas/move-video-clips-by-anchor.d.ts
1166
+ /**
1167
+ * Where the moved block lands on the main track.
1168
+ *
1169
+ * A discriminated union rather than two optional `before_clip_id` /
1170
+ * `after_clip_id` fields (the shape `addVideoClips` had to use, because there the
1171
+ * two modes share a whole clip description): here the alternatives carry nothing
1172
+ * in common, so making them mutually exclusive *by type* removes three runtime
1173
+ * `superRefine` checks that would otherwise have to be written and tested.
1174
+ *
1175
+ * `track_start` is an explicit member, not the absence of an anchor. The
1176
+ * agent-harness mutation this maps from treats "neither anchor given" as
1177
+ * "move to the front" (`applyBatchMoveVideoClips` falls back to `insertIndex = 0`),
1178
+ * which is a default buried in a tool description. Requiring the caller to name
1179
+ * that intent keeps a forgotten field from silently reordering the timeline.
1180
+ */
1181
+ declare const moveAnchorSchema: z.ZodDiscriminatedUnion<[z.ZodObject<{
1182
+ position: z.ZodLiteral<"before">;
1183
+ clip_id: z.ZodString;
1184
+ }, z.core.$strip>, z.ZodObject<{
1185
+ position: z.ZodLiteral<"after">;
1186
+ clip_id: z.ZodString;
1187
+ }, z.core.$strip>, z.ZodObject<{
1188
+ position: z.ZodLiteral<"track_start">;
1189
+ }, z.core.$strip>], "position">;
1190
+ /**
1191
+ * What happens to the speeches anchored to the clips being moved.
1192
+ *
1193
+ * - `follow` keeps each speech anchored where it is, so it travels with its clip
1194
+ * to the new position. In the anchored model this is the *no-op* branch: a
1195
+ * speech's authoritative fact is `{ anchorPartId, offsetMs }` and its absolute
1196
+ * time is derived on read from the host's position, so moving the host moves
1197
+ * the speech with no write to the speech at all.
1198
+ * - `keep_absolute` preserves each speech's current absolute landing instead, then
1199
+ * re-anchors it to whichever clip now covers that time (RFC 02 §9.1/§11.1). This
1200
+ * is the branch that costs an extra pass, and the one the FE timeline uses.
1201
+ *
1202
+ * **Required, with no default**, matching `deleteVideoClips` and
1203
+ * `replaceVideoClipSequence`. The two branches decide which picture the user's
1204
+ * narration ends up over, which is too consequential to infer from a missing
1205
+ * field — and neither branch is "safe enough" to be the implicit one.
1206
+ */
1207
+ declare const movedClipAnchoredPolicySchema: z.ZodEnum<{
1208
+ follow: "follow";
1209
+ keep_absolute: "keep_absolute";
1210
+ }>;
1211
+ /**
1212
+ * Reorder a set of main-track clips relative to a reference clip.
1213
+ *
1214
+ * Distinct from `moveVideoClips`, which positions clips by absolute time
1215
+ * (`new_start_ms`) and is what the FE timeline dispatches after a drag. Main-track
1216
+ * clips are `sequential`-positioned, so absolute time is not authoritative state
1217
+ * there (RFC 02 §4/§7): `moveVideoClips` has to *guess* an index back out of the
1218
+ * time it was handed, whereas an anchor already is the ordinal fact being changed.
1219
+ * Keeping them separate also isolates blast radius — this method can diverge on
1220
+ * speech policy without touching the FE path.
1221
+ *
1222
+ * `clip_ids` need not be contiguous. They move as one block, keeping their
1223
+ * relative order, which is the agent-harness `batch_move_video_clips` contract.
1224
+ */
1225
+ declare const moveVideoClipsByAnchorInputSchema: z.ZodObject<{
1226
+ clip_ids: z.ZodArray<z.ZodString>;
1227
+ anchor: z.ZodDiscriminatedUnion<[z.ZodObject<{
1228
+ position: z.ZodLiteral<"before">;
1229
+ clip_id: z.ZodString;
1230
+ }, z.core.$strip>, z.ZodObject<{
1231
+ position: z.ZodLiteral<"after">;
1232
+ clip_id: z.ZodString;
1233
+ }, z.core.$strip>, z.ZodObject<{
1234
+ position: z.ZodLiteral<"track_start">;
1235
+ }, z.core.$strip>], "position">;
1236
+ on_anchored: z.ZodEnum<{
1237
+ follow: "follow";
1238
+ keep_absolute: "keep_absolute";
1239
+ }>;
1240
+ }, z.core.$strip>;
1241
+ type MoveAnchor = z.infer<typeof moveAnchorSchema>;
1242
+ type MovedClipAnchoredPolicy = z.infer<typeof movedClipAnchoredPolicySchema>;
1243
+ type MoveVideoClipsByAnchorInput = z.infer<typeof moveVideoClipsByAnchorInputSchema>;
1244
+ //#endregion
1245
+ //#region src/editor/schemas/move-video-clips.d.ts
1246
+ declare const moveVideoClipsInputSchema: z.ZodObject<{
1247
+ clips: z.ZodArray<z.ZodObject<{
1248
+ clip_id: z.ZodString;
1249
+ new_start_ms: z.ZodNumber;
1250
+ new_track_id: z.ZodOptional<z.ZodString>;
1251
+ }, z.core.$strip>>;
1252
+ }, z.core.$strip>;
1253
+ type MoveVideoClipsInput = z.infer<typeof moveVideoClipsInputSchema>;
1254
+ //#endregion
1255
+ //#region src/editor/schemas/replace-video-clip-content.d.ts
1256
+ /**
1257
+ * Replace the media backing existing video clips. The media import runs upstream
1258
+ * (Director); its stable result — the new media id, intrinsic length, and the
1259
+ * reset trim window — arrives materialized (see
1260
+ * `results/phase-4-side-effect-payload-contract.md` §4). Director resets
1261
+ * `play_in=0` / `play_out=media_duration_ms` and clears `speed_shift` on
1262
+ * replacement. The clip `part_id`s (hence their track items) are unchanged; the
1263
+ * editor reflows the main track from the new effective durations.
1264
+ */
1265
+ declare const replaceVideoClipContentInputSchema: z.ZodObject<{
1266
+ clips: z.ZodArray<z.ZodObject<{
1267
+ clip_id: z.ZodString;
1268
+ origin_media_id: z.ZodString;
1269
+ media_duration_ms: z.ZodNumber;
1270
+ play_in: z.ZodNumber;
1271
+ play_out: z.ZodNumber;
1272
+ volume: z.ZodNumber;
1273
+ }, z.core.$strip>>;
1274
+ }, z.core.$strip>;
1275
+ type ReplaceVideoClipContentInput = z.infer<typeof replaceVideoClipContentInputSchema>;
1276
+ //#endregion
1277
+ //#region src/editor/schemas/replace-video-clip-sequence.d.ts
1278
+ /**
1279
+ * What happens to the speeches anchored to the clips being replaced.
1280
+ *
1281
+ * - `remap` re-anchors each surviving speech to the new clip in the SAME POSITION
1282
+ * of the sequence, keeping its offset — old[i]'s children become new[i]'s
1283
+ * children. An old clip with no counterpart (fewer new clips than old) has its
1284
+ * subtree deleted, because there is nothing left to anchor to.
1285
+ * - `cascade` deletes every anchored speech (and its captions) outright, like
1286
+ * `deleteVideoClips`.
1287
+ *
1288
+ * **Required, with no default.** The two branches differ in whether the user's
1289
+ * narration survives, and the agent tool that drives this op makes its
1290
+ * `preserve_speeches` flag required for that reason. A default here would let a
1291
+ * caller that forgot the field silently delete speech.
1292
+ */
1293
+ declare const anchoredReplacePolicySchema: z.ZodEnum<{
1294
+ cascade: "cascade";
1295
+ remap: "remap";
1296
+ }>;
1297
+ type AnchoredReplacePolicy = z.infer<typeof anchoredReplacePolicySchema>;
1298
+ /**
1299
+ * Replace a contiguous run of main-track clips with a new run.
1300
+ *
1301
+ * A composite of delete + insert that cannot be expressed as the two ops in
1302
+ * sequence, because the anchored speeches have to survive *across* the swap: with
1303
+ * `remap` they are re-anchored positionally, which needs both the old and the new
1304
+ * ids in the same transaction (ADR 0009 — the cascade stays in the editor, callers
1305
+ * never re-wire anchors themselves).
1306
+ *
1307
+ * Duration facts follow `addVideoClips`: `media_duration_ms` is the source's
1308
+ * intrinsic length and the trim window defaults to the whole media. The effective
1309
+ * timeline duration is derived by the projection, never an input.
1310
+ *
1311
+ * `media_id` is optional: omitting it creates a **deliberate empty placeholder
1312
+ * clip** (`origin_media_id: ''`) — structure with no picture. This is the only op
1313
+ * that can produce one, and it is authoritative state, unlike the gap fillers the
1314
+ * read-side solve mints (which never enter the document).
1315
+ */
1316
+ declare const replaceVideoClipSequenceInputSchema: z.ZodObject<{
1317
+ old_clip_ids: z.ZodArray<z.ZodString>;
1318
+ new_clips: z.ZodArray<z.ZodObject<{
1319
+ media_id: z.ZodOptional<z.ZodString>;
1320
+ media_duration_ms: z.ZodNumber;
1321
+ play_in: z.ZodOptional<z.ZodNumber>;
1322
+ play_out: z.ZodOptional<z.ZodNumber>;
1323
+ }, z.core.$strip>>;
1324
+ on_anchored: z.ZodEnum<{
1325
+ cascade: "cascade";
1326
+ remap: "remap";
1327
+ }>;
1328
+ }, z.core.$strip>;
1329
+ type ReplaceVideoClipSequenceInput = z.infer<typeof replaceVideoClipSequenceInputSchema>;
1330
+ //#endregion
1331
+ //#region src/editor/schemas/resize-caption.d.ts
1332
+ /** Resize one caption placement without changing its cue/source duration. */
1333
+ declare const resizeCaptionInputSchema: z.ZodObject<{
1334
+ caption_id: z.ZodString;
1335
+ duration_ms: z.ZodNumber;
1336
+ }, z.core.$strip>;
1337
+ type ResizeCaptionInput = z.infer<typeof resizeCaptionInputSchema>;
1338
+ //#endregion
1339
+ //#region src/editor/schemas/set-bgm.d.ts
1340
+ /**
1341
+ * Set the document BGM. The media's stable result (storage key) arrives
1342
+ * materialized from upstream (see
1343
+ * `results/phase-4-side-effect-payload-contract.md` §3). The op upserts the bgm
1344
+ * part and seats it on the bgm lane; its effective length is always the whole
1345
+ * timeline, derived by the projection on read — so there is no `duration_ms`
1346
+ * input or fact (RFC 02 / `reference/16` §0b). A `bgm_id` lets the op replace an
1347
+ * existing bgm part by id.
1348
+ */
1349
+ declare const setBgmInputSchema: z.ZodObject<{
1350
+ bgm_id: z.ZodString;
1351
+ audio_storage_key: z.ZodString;
1352
+ origin_media_id: z.ZodString;
1353
+ volume: z.ZodNumber;
1354
+ }, z.core.$strip>;
1355
+ type SetBgmInput = z.infer<typeof setBgmInputSchema>;
1356
+ //#endregion
1357
+ //#region src/editor/schemas/set-caption-style.d.ts
1358
+ /**
1359
+ * Set the caption visual style. GLOBAL by design: the style applies to every
1360
+ * caption part in the document — it carries NO `caption_id`. This mirrors the FE,
1361
+ * whose caption-style store (`caption-style.ts:persistCaptionStylePatch`) iterates
1362
+ * ALL captions and writes the same normalized style to each; the product has a
1363
+ * single document-wide caption style, not per-caption styling.
1364
+ *
1365
+ * Every field is optional and maps to a `CaptionStyle` attribute (snake_case
1366
+ * IDL). A field present in the input is written to every caption; a field ABSENT
1367
+ * from the input is left untouched on each caption (the editor merges the patch
1368
+ * onto each caption's existing style — this is a value edit, not a full-style
1369
+ * replace, so a partial patch such as "recolor only" does not wipe font size).
1370
+ *
1371
+ * Pure document edit, no cascade — captions keep their positions; only the style
1372
+ * sub-map of each caption part changes.
1373
+ */
1374
+ declare const setCaptionStyleInputSchema: z.ZodObject<{
1375
+ font_id: z.ZodOptional<z.ZodString>;
1376
+ font_size: z.ZodOptional<z.ZodNumber>;
1377
+ font_color: z.ZodOptional<z.ZodString>;
1378
+ font_weight: z.ZodOptional<z.ZodNumber>;
1379
+ entrance_animation: z.ZodOptional<z.ZodString>;
1380
+ entrance_animation_duration_ms: z.ZodOptional<z.ZodNumber>;
1381
+ stroke_color: z.ZodOptional<z.ZodString>;
1382
+ stroke_width: z.ZodOptional<z.ZodNumber>;
1383
+ position_x: z.ZodOptional<z.ZodNumber>;
1384
+ position_y: z.ZodOptional<z.ZodNumber>;
1385
+ }, z.core.$strip>;
1386
+ type SetCaptionStyleInput = z.infer<typeof setCaptionStyleInputSchema>;
1387
+ //#endregion
1388
+ //#region src/editor/schemas/set-caption-visibility.d.ts
1389
+ /**
1390
+ * Toggle caption visibility (the caption track's `is_hidden` flag). Pure
1391
+ * document edit, no cascade — captions keep their positions; only the lane's
1392
+ * hidden flag changes.
1393
+ */
1394
+ declare const setCaptionVisibilityInputSchema: z.ZodObject<{
1395
+ is_hidden: z.ZodBoolean;
1396
+ }, z.core.$strip>;
1397
+ type SetCaptionVisibilityInput = z.infer<typeof setCaptionVisibilityInputSchema>;
1398
+ //#endregion
1399
+ //#region src/editor/schemas/set-video-clip-speed-shift.d.ts
1400
+ /**
1401
+ * Set the playback speed of existing video clips. Per the speed-shift decision
1402
+ * (`reference/16` §0): the op writes only the `speed_shift` fact — it does NOT
1403
+ * store an effective `duration_ms` (projection derives it from the trim window /
1404
+ * speed) and does NOT scale anchored speeches' relative offsets (offsets stay
1405
+ * put; the cascade reflows absolute positions). A `null` speed_shift clears the
1406
+ * speed back to original (1×).
1407
+ */
1408
+ declare const setVideoClipSpeedShiftInputSchema: z.ZodObject<{
1409
+ clips: z.ZodArray<z.ZodObject<{
1410
+ clip_id: z.ZodString;
1411
+ speed_shift: z.ZodNullable<z.ZodObject<{
1412
+ category: z.ZodEnum<{
1413
+ curve: "curve";
1414
+ linear: "linear";
1415
+ }>;
1416
+ mode: z.ZodString;
1417
+ config: z.ZodUnion<readonly [z.ZodObject<{
1418
+ linear: z.ZodObject<{
1419
+ speed: z.ZodNumber;
1420
+ }, z.core.$strip>;
1421
+ }, z.core.$strip>, z.ZodObject<{
1422
+ curve: z.ZodObject<{
1423
+ keyframes: z.ZodArray<z.ZodObject<{
1424
+ position: z.ZodNumber;
1425
+ rate: z.ZodNumber;
1426
+ in_tangent: z.ZodOptional<z.ZodObject<{
1427
+ x: z.ZodNumber;
1428
+ y: z.ZodNumber;
1429
+ }, z.core.$strip>>;
1430
+ out_tangent: z.ZodOptional<z.ZodObject<{
1431
+ x: z.ZodNumber;
1432
+ y: z.ZodNumber;
1433
+ }, z.core.$strip>>;
1434
+ }, z.core.$strip>>;
1435
+ }, z.core.$strip>;
1436
+ }, z.core.$strip>]>;
1437
+ }, z.core.$strip>>;
1438
+ }, z.core.$strip>>;
1439
+ }, z.core.$strip>;
1440
+ type SetVideoClipSpeedShiftInput = z.infer<typeof setVideoClipSpeedShiftInputSchema>;
1441
+ //#endregion
1442
+ //#region src/editor/schemas/shared.d.ts
1443
+ declare const clipIdSchema: z.ZodString;
1444
+ declare const clipIdsSchema: z.ZodArray<z.ZodString>;
1445
+ declare const mediaIdSchema: z.ZodString;
1446
+ declare const speechIdSchema: z.ZodString;
1447
+ declare const timelineMsSchema: z.ZodNumber;
1448
+ declare const positiveMsSchema: z.ZodNumber;
1449
+ declare const volumeSchema: z.ZodNumber;
1450
+ declare const speechIdsSchema: z.ZodArray<z.ZodString>;
1451
+ /**
1452
+ * A clip's playback-speed fact, the only thing `SetVideoClipSpeedShift` writes.
1453
+ * Mirrors the IDL `SpeedShift`: `category` is `linear` | `curve`, and `config`
1454
+ * is a discriminated union — `{ linear: { speed } }` for a constant multiplier
1455
+ * (the multiplier projection reads at `config.linear.speed`) or `{ curve: {
1456
+ * keyframes } }` for a Bezier-controlled variable speed (RFC 02 / `reference/16`
1457
+ * §0). Exactly one of `linear` / `curve` is present.
1458
+ */
1459
+ declare const speedShiftSchema: z.ZodObject<{
1460
+ category: z.ZodEnum<{
1461
+ curve: "curve";
1462
+ linear: "linear";
1463
+ }>;
1464
+ mode: z.ZodString;
1465
+ config: z.ZodUnion<readonly [z.ZodObject<{
1466
+ linear: z.ZodObject<{
1467
+ speed: z.ZodNumber;
1468
+ }, z.core.$strip>;
1469
+ }, z.core.$strip>, z.ZodObject<{
1470
+ curve: z.ZodObject<{
1471
+ keyframes: z.ZodArray<z.ZodObject<{
1472
+ position: z.ZodNumber;
1473
+ rate: z.ZodNumber;
1474
+ in_tangent: z.ZodOptional<z.ZodObject<{
1475
+ x: z.ZodNumber;
1476
+ y: z.ZodNumber;
1477
+ }, z.core.$strip>>;
1478
+ out_tangent: z.ZodOptional<z.ZodObject<{
1479
+ x: z.ZodNumber;
1480
+ y: z.ZodNumber;
1481
+ }, z.core.$strip>>;
1482
+ }, z.core.$strip>>;
1483
+ }, z.core.$strip>;
1484
+ }, z.core.$strip>]>;
1485
+ }, z.core.$strip>;
1486
+ declare const voiceSchema: z.ZodObject<{
1487
+ id: z.ZodString;
1488
+ name: z.ZodString;
1489
+ }, z.core.$strip>;
1490
+ //#endregion
1491
+ //#region src/editor/schemas/speech-assets.d.ts
1492
+ /**
1493
+ * The materialized TTS result shared by `AddSpeeches` / `ChangeSpeechScript` /
1494
+ * `ChangeSpeechVoice` (see `results/phase-4-side-effect-payload-contract.md`
1495
+ * §1/§2). The side effect (TTS/ASR + billing) runs upstream; the op receives the
1496
+ * stable speech + caption parts and writes them as authoritative facts. No
1497
+ * cascade runs on write — the projection derives absolute positions on read.
1498
+ *
1499
+ * `anchor_part_id` states the root relationship explicitly: a video/BGM id
1500
+ * creates an anchored speech; `null` creates an absolute audio root.
1501
+ * `offset_ms` is relative to the anchor or absolute for a root. Captions always
1502
+ * anchor to their speech via `start_ms`.
1503
+ */
1504
+ declare const speechAssetSchema: z.ZodObject<{
1505
+ speech_id: z.ZodString;
1506
+ anchor_part_id: z.ZodNullable<z.ZodString>;
1507
+ offset_ms: z.ZodNumber;
1508
+ audio_storage_key: z.ZodString;
1509
+ duration_ms: z.ZodNumber;
1510
+ audio_script: z.ZodString;
1511
+ volume: z.ZodNumber;
1512
+ voice: z.ZodObject<{
1513
+ id: z.ZodString;
1514
+ name: z.ZodString;
1515
+ }, z.core.$strip>;
1516
+ origin_speech_id: z.ZodString;
1517
+ caption_ids: z.ZodArray<z.ZodString>;
1518
+ }, z.core.$strip>;
1519
+ declare const captionAssetSchema: z.ZodObject<{
1520
+ caption_id: z.ZodString;
1521
+ speech_part_id: z.ZodString;
1522
+ text: z.ZodString;
1523
+ start_ms: z.ZodNumber;
1524
+ duration_ms: z.ZodNumber;
1525
+ }, z.core.$strip>;
1526
+ /** A materialized speech-subtree write (speeches + their captions). */
1527
+ declare const speechAssetsSchema: z.ZodObject<{
1528
+ speeches: z.ZodArray<z.ZodObject<{
1529
+ speech_id: z.ZodString;
1530
+ anchor_part_id: z.ZodNullable<z.ZodString>;
1531
+ offset_ms: z.ZodNumber;
1532
+ audio_storage_key: z.ZodString;
1533
+ duration_ms: z.ZodNumber;
1534
+ audio_script: z.ZodString;
1535
+ volume: z.ZodNumber;
1536
+ voice: z.ZodObject<{
1537
+ id: z.ZodString;
1538
+ name: z.ZodString;
1539
+ }, z.core.$strip>;
1540
+ origin_speech_id: z.ZodString;
1541
+ caption_ids: z.ZodArray<z.ZodString>;
1542
+ }, z.core.$strip>>;
1543
+ captions: z.ZodArray<z.ZodObject<{
1544
+ caption_id: z.ZodString;
1545
+ speech_part_id: z.ZodString;
1546
+ text: z.ZodString;
1547
+ start_ms: z.ZodNumber;
1548
+ duration_ms: z.ZodNumber;
1549
+ }, z.core.$strip>>;
1550
+ }, z.core.$strip>;
1551
+ type SpeechAsset = z.infer<typeof speechAssetSchema>;
1552
+ type CaptionAsset = z.infer<typeof captionAssetSchema>;
1553
+ type SpeechAssets = z.infer<typeof speechAssetsSchema>;
1554
+ declare namespace index_d_exports {
1555
+ export { AddSpeechesInput, AddVideoClipsInput, AdjustBgmVolumeInput, AdjustSpeechVolumeInput, AdjustVideoClipDurationInput, AdjustVideoClipVolumeInput, AnchoredDeletePolicy, AnchoredReplacePolicy, CaptionAsset, ChangeSpeechScriptInput, ChangeSpeechVoiceInput, DeleteBgmInput, DeleteSpeechesInput, DeleteVideoClipsInput, MoveAnchor, MoveBgmInput, MoveSpeechesInput, MoveVideoClipsByAnchorInput, MoveVideoClipsInput, MovedClipAnchoredPolicy, ReplaceVideoClipContentInput, ReplaceVideoClipSequenceInput, ResizeCaptionInput, SetBgmInput, SetCaptionStyleInput, SetCaptionVisibilityInput, SetVideoClipSpeedShiftInput, SpeechAsset, SpeechAssets, addSpeechesInputSchema, addVideoClipsInputSchema, adjustBgmVolumeInputSchema, adjustSpeechVolumeInputSchema, adjustVideoClipDurationInputSchema, adjustVideoClipVolumeInputSchema, anchoredDeletePolicySchema, anchoredReplacePolicySchema, changeSpeechScriptInputSchema, changeSpeechVoiceInputSchema, clipIdSchema, clipIdsSchema, deleteBgmInputSchema, deleteSpeechesInputSchema, deleteVideoClipsInputSchema, mediaIdSchema, moveAnchorSchema, moveBgmInputSchema, moveSpeechesInputSchema, moveVideoClipsByAnchorInputSchema, moveVideoClipsInputSchema, movedClipAnchoredPolicySchema, positiveMsSchema, replaceVideoClipContentInputSchema, replaceVideoClipSequenceInputSchema, resizeCaptionInputSchema, setBgmInputSchema, setCaptionStyleInputSchema, setCaptionVisibilityInputSchema, setVideoClipSpeedShiftInputSchema, speechAssetsSchema, speechIdSchema, speechIdsSchema, speedShiftSchema, timelineMsSchema, voiceSchema, volumeSchema };
1556
+ }
981
1557
  //#endregion
982
1558
  //#region src/editor/schema-validator.d.ts
983
1559
  interface SnapshotReadable {
@@ -1025,9 +1601,11 @@ declare class SchemaValidator {
1025
1601
  validateChangeSpeechVoice(input: ChangeSpeechVoiceInput, doc: SnapshotReadable): void;
1026
1602
  validateDeleteSpeeches(input: DeleteSpeechesInput, doc: SnapshotReadable): void;
1027
1603
  validateMoveSpeeches(input: MoveSpeechesInput, doc: SnapshotReadable): void;
1604
+ validateMoveBgm(input: MoveBgmInput, doc: SnapshotReadable): void;
1028
1605
  validateSetBgm(input: SetBgmInput, _doc: SnapshotReadable): void;
1029
1606
  validateDeleteBgm(input: DeleteBgmInput, _doc: SnapshotReadable): void;
1030
1607
  validateSetCaptionVisibility(input: SetCaptionVisibilityInput, _doc: SnapshotReadable): void;
1608
+ validateResizeCaption(input: ResizeCaptionInput, doc: SnapshotReadable): void;
1031
1609
  validateSetCaptionStyle(input: SetCaptionStyleInput, _doc: SnapshotReadable): void;
1032
1610
  private parse;
1033
1611
  private mainTrackIds;
@@ -1036,12 +1614,8 @@ declare class SchemaValidator {
1036
1614
  /** Each regenerated speech (re-TTS) must already exist in the document. */
1037
1615
  private assertSpeechesExist;
1038
1616
  /**
1039
- * Each speech's `anchor_part_id` must point at a video clip that already
1040
- * exists. A relative speech anchored to a missing clip cannot be positioned
1041
- * by the projection and has no host to recover to (RFC 02 §4/§11.1), so the
1042
- * write must be rejected rather than landing a dangling reference. (Restores
1043
- * the dangling-reference guard the old `validateMaterializedPatch` carried;
1044
- * ADR 0009.)
1617
+ * Anchored speeches may follow a video or BGM root. A null anchor is an
1618
+ * explicit absolute speech root. Missing and wrong-kind parents are rejected.
1045
1619
  */
1046
1620
  private assertAnchorsExist;
1047
1621
  private assertPartKind;
@@ -1076,7 +1650,7 @@ declare class SchemaValidator {
1076
1650
  * side-effect payload contract in
1077
1651
  * `docs/projects/medeo-integration/results/phase-4-side-effect-payload-contract.md`.
1078
1652
  */
1079
- type ImplementedSemanticOpKind = 'MoveVideoClips' | 'MoveVideoClipsByAnchor' | 'DeleteVideoClips' | 'AddVideoClips' | 'AdjustVideoClipVolume' | 'SetVideoClipSpeedShift' | 'ReplaceVideoClipContent' | 'ReplaceVideoClipSequence' | 'AdjustVideoClipDuration' | 'AddSpeeches' | 'DeleteSpeeches' | 'MoveSpeeches' | 'ChangeSpeechScript' | 'ChangeSpeechVoice' | 'AdjustSpeechVolume' | 'SetCaptionVisibility' | 'SetCaptionStyle' | 'SetBgm' | 'DeleteBgm' | 'AdjustBgmVolume';
1653
+ type ImplementedSemanticOpKind = 'MoveVideoClips' | 'MoveVideoClipsByAnchor' | 'DeleteVideoClips' | 'AddVideoClips' | 'AdjustVideoClipVolume' | 'SetVideoClipSpeedShift' | 'ReplaceVideoClipContent' | 'ReplaceVideoClipSequence' | 'AdjustVideoClipDuration' | 'AddSpeeches' | 'DeleteSpeeches' | 'MoveSpeeches' | 'MoveBgm' | 'ChangeSpeechScript' | 'ChangeSpeechVoice' | 'AdjustSpeechVolume' | 'ResizeCaption' | 'SetCaptionVisibility' | 'SetCaptionStyle' | 'SetBgm' | 'DeleteBgm' | 'AdjustBgmVolume';
1080
1654
  /**
1081
1655
  * Operations on the roadmap but NOT yet implemented by `SemanticEditor`. Empty:
1082
1656
  * Phase 4 covers every op with a real entry point. Kept as a named type so the
@@ -1131,19 +1705,14 @@ interface OpActor {
1131
1705
  role: 'user' | 'agent';
1132
1706
  }
1133
1707
  interface CommitOptions {
1134
- /**
1135
- * Caller-supplied intent attached to the op's audit message.
1136
- *
1137
- * Typed as `unknown` to match `TransactAudit.intent`. The wire (real
1138
- * mengine-server and the in-memory test double) only surfaces **string**
1139
- * intents onto the audit log — non-string values stay in the Loro commit
1140
- * message but parse to `null`. Agent callers should pass a string.
1141
- */
1142
- intent?: unknown;
1708
+ intent?: {
1709
+ kind: string;
1710
+ payload: unknown;
1711
+ };
1143
1712
  /** Author of this op; omitted for writes with no acting user (bootstrap, repair). */
1144
1713
  actor?: OpActor;
1145
1714
  }
1146
- type SemanticOpInput = MoveVideoClipsInput | MoveVideoClipsByAnchorInput | DeleteVideoClipsInput | AddVideoClipsInput | AdjustVideoClipVolumeInput | AdjustSpeechVolumeInput | AdjustBgmVolumeInput | SetVideoClipSpeedShiftInput | ReplaceVideoClipContentInput | ReplaceVideoClipSequenceInput | AdjustVideoClipDurationInput | AddSpeechesInput | DeleteSpeechesInput | MoveSpeechesInput | SetCaptionVisibilityInput | SetCaptionStyleInput | SetBgmInput | DeleteBgmInput;
1715
+ type SemanticOpInput = MoveVideoClipsInput | MoveVideoClipsByAnchorInput | DeleteVideoClipsInput | AddVideoClipsInput | AdjustVideoClipVolumeInput | AdjustSpeechVolumeInput | AdjustBgmVolumeInput | SetVideoClipSpeedShiftInput | ReplaceVideoClipContentInput | ReplaceVideoClipSequenceInput | AdjustVideoClipDurationInput | AddSpeechesInput | DeleteSpeechesInput | MoveSpeechesInput | MoveBgmInput | ResizeCaptionInput | SetCaptionVisibilityInput | SetCaptionStyleInput | SetBgmInput | DeleteBgmInput;
1147
1716
  type SemanticOpName = ImplementedSemanticOpKind;
1148
1717
  /** Audit metadata committed alongside an op's writes. */
1149
1718
  interface TransactAudit {
@@ -1184,8 +1753,7 @@ interface SemanticDocumentAdapter extends SnapshotReadable {
1184
1753
  declare class SemanticEditor {
1185
1754
  private readonly doc;
1186
1755
  private readonly validator;
1187
- private readonly idFactory;
1188
- constructor(doc: SemanticDocumentAdapter, validator?: SchemaValidator, idFactory?: PartIdFactory);
1756
+ constructor(doc: SemanticDocumentAdapter, validator?: SchemaValidator);
1189
1757
  moveVideoClips(input: MoveVideoClipsInput, options?: CommitOptions): Promise<void>;
1190
1758
  /**
1191
1759
  * Reorder main-track clips relative to an anchor clip, moving them as one block.
@@ -1260,6 +1828,8 @@ declare class SemanticEditor {
1260
1828
  * derives absolute positions on read.
1261
1829
  */
1262
1830
  moveSpeeches(input: MoveSpeechesInput, options?: CommitOptions): Promise<void>;
1831
+ /** Move the BGM root; any speech/caption descendants follow through their anchors. */
1832
+ moveBgm(input: MoveBgmInput, options?: CommitOptions): Promise<void>;
1263
1833
  /** Change a speech's script. Re-TTS runs upstream; the regenerated parts arrive materialized (no cascade). */
1264
1834
  changeSpeechScript(input: ChangeSpeechScriptInput, options?: CommitOptions): Promise<void>;
1265
1835
  /** Change a speech's voice. Re-TTS runs upstream; the regenerated parts arrive materialized (no cascade). */
@@ -1290,6 +1860,8 @@ declare class SemanticEditor {
1290
1860
  deleteBgm(input: DeleteBgmInput, options?: CommitOptions): Promise<void>;
1291
1861
  /** Toggle caption visibility (caption track `is_hidden`). */
1292
1862
  setCaptionVisibility(input: SetCaptionVisibilityInput, options?: CommitOptions): Promise<void>;
1863
+ /** Resize the caption placement window; cue/source duration stays immutable. */
1864
+ resizeCaption(input: ResizeCaptionInput, options?: CommitOptions): Promise<void>;
1293
1865
  /**
1294
1866
  * Set the document-wide caption style. GLOBAL (no `caption_id`): the patch is
1295
1867
  * merged onto EVERY caption part's `style`, mirroring the FE, which applies a
@@ -1356,57 +1928,6 @@ declare class MirrorVideoDocumentAdapter implements SemanticDocumentAdapter {
1356
1928
  declare function createMirrorVideoDocument(document: VideoDocument, options?: MirrorVideoDocumentOptions): LoroDoc;
1357
1929
  /** Build a `MirrorVideoDocumentAdapter` over a fresh doc seeded with `document`. */
1358
1930
  declare function createMirrorVideoDocumentAdapter(document: VideoDocument, options?: MirrorVideoDocumentOptions): MirrorVideoDocumentAdapter;
1359
- /** Write a whole `VideoDocument` into a draft (seed a fresh doc / plain-memory state). */
1360
- declare function writeVideoDocumentToDraft(draft: VideoDocumentDraft, document: VideoDocument): void;
1361
- //#endregion
1362
- //#region src/document/plain-memory-adapter.d.ts
1363
- /**
1364
- * A `TransactAudit` widened with the ordered ids minted during that transact.
1365
- * `generated_ids` is empty when the op never called the id factory.
1366
- */
1367
- interface JournalEntry extends TransactAudit {
1368
- /** Ids produced by the adapter's id factory during this transact, in mint order. */
1369
- generated_ids: string[];
1370
- }
1371
- interface PlainMemoryAdapterOptions {
1372
- /** Underlying id mint; wrapped so each call inside a transact is journaled. */
1373
- idFactory?: PartIdFactory;
1374
- }
1375
- /**
1376
- * Pure in-memory `SemanticDocumentAdapter` — no Loro/WASM. Holds a
1377
- * `VideoDocumentDraft` object and applies each `transact` via immer `produce`,
1378
- * journaling every audit that actually mutated state (plus any ids minted
1379
- * during that transact).
1380
- *
1381
- * Known benign difference vs `MirrorVideoDocumentAdapter`: the editor's
1382
- * `setPart` assigns a fresh part object on every call, so a same-value rewrite
1383
- * produces a new immer state and IS journaled here, while the mirror's deep
1384
- * diff emits no commit. Terminal `snapshot()` stays equal; replay is idempotent.
1385
- */
1386
- declare class PlainMemoryAdapter implements SemanticDocumentAdapter {
1387
- private state;
1388
- private readonly _journal;
1389
- private readonly baseIdFactory;
1390
- /** Non-null only while a `transact` edit callback is running. */
1391
- private pendingIds;
1392
- /**
1393
- * Recording wrapper around the underlying factory. Callers (sandbox editor)
1394
- * use this so every minted id is appended to the current transact's list.
1395
- */
1396
- readonly idFactory: PartIdFactory;
1397
- constructor(document: VideoDocument, options?: PlainMemoryAdapterOptions);
1398
- get journal(): readonly JournalEntry[];
1399
- hasContent(): boolean;
1400
- snapshot(): VideoDocument;
1401
- /**
1402
- * Apply one op via immer. A throw in `edit` discards the draft (state and
1403
- * journal unchanged). When `produce` returns the same reference, there was
1404
- * no structural change — skip journal, matching mirror "no change, no commit".
1405
- */
1406
- transact(edit: (draft: VideoDocumentDraft) => void, audit: TransactAudit): void;
1407
- }
1408
- /** Build a `PlainMemoryAdapter` seeded with `document`. */
1409
- declare function createPlainMemoryAdapter(document: VideoDocument, options?: PlainMemoryAdapterOptions): PlainMemoryAdapter;
1410
1931
  //#endregion
1411
1932
  //#region src/document/mirror-read.d.ts
1412
1933
  /**
@@ -1525,6 +2046,7 @@ declare const videoDocumentSchema: z.ZodObject<{
1525
2046
  offsetMs: z.ZodNumber;
1526
2047
  }, z.core.$loose>], "mode">;
1527
2048
  fallback_abs_ms: z.ZodOptional<z.ZodNumber>;
2049
+ duration_override_ms: z.ZodOptional<z.ZodNumber>;
1528
2050
  }, z.core.$loose>>>;
1529
2051
  }, z.core.$loose>>>;
1530
2052
  part_library: z.ZodOptional<z.ZodRecord<z.ZodString, z.ZodUnion<readonly [z.ZodObject<{
@@ -1591,4 +2113,4 @@ declare const videoDocumentSchema: z.ZodObject<{
1591
2113
  }, z.core.$loose>]>>>;
1592
2114
  }, z.core.$loose>;
1593
2115
  //#endregion
1594
- export { Track as $, TrackDraft as A, DerivedItemPosition as B, SemanticOpKind as C, ValidationError as D, SnapshotReadable as E, VideoDocumentValidationError as F, toVideoDocument as G, buildSpeechHostMap as H, assertValidVideoDocument as I, DEFAULT_UNIT_TIME_MS as J, BgmPart as K, validateVideoDocument as L, VideoDocumentDraft as M, VideoDocumentMirrorSchema as N, PartIdFactory as O, videoDocumentMirrorSchema as P, Timeline as Q, InitialDocumentFacts as R, PlannedSemanticOpKind as S, SchemaValidator as T, derivePositionFromAbs as U, SpeechHostMap as V, fromVideoDocument as W, PartUnion as X, PartKind as Y, SpeechPart as Z, SemanticOpInput as _, Timeline$1 as _t, PlainMemoryAdapter as a, VideoDocumentSchemaVersion as at, IMPLEMENTED_SEMANTIC_OP_KINDS as b, MirrorVideoDocumentAdapter as c, VideoDraft as ct, createMirrorVideoDocumentAdapter as d, speedOf as dt, TrackItem as et, writeVideoDocumentToDraft as f, Attachment as ft, SemanticEditor as g, SpeedShift as gt, SemanticDocumentAdapter as h, PartAggregation as ht, JournalEntry as i, VideoDocument as it, TrackItemDraft as j, generatePartId as k, MirrorVideoDocumentOptions as l, VideoDraftPartUnion as lt, OpActor as m, CaptionStyle as mt, videoDocumentSchema as n, VIDEO_DOCUMENT_SCHEMA_VERSION as nt, PlainMemoryAdapterOptions as o, VideoDocumentValidationIssue as ot, CommitOptions as p, CaptionPart$1 as pt, CaptionPart as q, readVideoDocumentFromDraft as r, VideoClipPart as rt, createPlainMemoryAdapter as s, VideoDocumentValidationIssueCode as st, partUnionSchema as t, TrackItemTimePosition as tt, createMirrorVideoDocument as u, effectiveVideoClipDurationMs as ut, SemanticOpName as v, Track$1 as vt, isImplementedSemanticOpKind as w, ImplementedSemanticOpKind as x, TransactAudit as y, TrackItem$1 as yt, buildInitialVideoDocument as z };
2116
+ export { VideoDocumentSchemaVersion as $, assertValidVideoDocument as A, BgmPart as B, index_d_exports as C, VideoDocumentMirrorSchema as D, VideoDocumentDraft as E, SpeechHostMap as F, SpeechPart as G, DEFAULT_UNIT_TIME_MS as H, buildSpeechHostMap as I, TrackItem as J, Timeline as K, derivePositionFromAbs as L, InitialDocumentFacts as M, buildInitialVideoDocument as N, videoDocumentMirrorSchema as O, DerivedItemPosition as P, VideoDocument as Q, fromVideoDocument as R, ValidationError as S, TrackItemDraft as T, PartKind as U, CaptionPart as V, PartUnion as W, VIDEO_DOCUMENT_SCHEMA_VERSION as X, TrackItemTimePosition as Y, VideoClipPart as Z, PlannedSemanticOpKind as _, MirrorVideoDocumentOptions as a, CaptionPart$1 as at, SchemaValidator as b, CommitOptions as c, SpeedShift as ct, SemanticEditor as d, TrackItem$1 as dt, VideoDocumentValidationIssue as et, SemanticOpInput as f, ImplementedSemanticOpKind as g, IMPLEMENTED_SEMANTIC_OP_KINDS as h, MirrorVideoDocumentAdapter as i, Attachment as it, validateVideoDocument as j, VideoDocumentValidationError as k, OpActor as l, Timeline$1 as lt, TransactAudit as m, videoDocumentSchema as n, VideoDraft as nt, createMirrorVideoDocument as o, CaptionStyle as ot, SemanticOpName as p, Track as q, readVideoDocumentFromDraft as r, VideoDraftPartUnion as rt, createMirrorVideoDocumentAdapter as s, PartAggregation as st, partUnionSchema as t, VideoDocumentValidationIssueCode as tt, SemanticDocumentAdapter as u, Track$1 as ut, SemanticOpKind as v, TrackDraft as w, SnapshotReadable as x, isImplementedSemanticOpKind as y, toVideoDocument as z };