@gentbajko/slopify 3.25.9 → 4.1.0

package/dist/slices/keywords/values.js

@@ -0,0 +1,74 @@
+// The facts a run copies when it starts, for the keywords Slopify fills from them, and how each
+// becomes a keyword's text. Browser-safe.
+import { previousVideoLink } from "../youtube/placeholders.js";
+// The description's own link placeholder for the next video, filled when it is shown like
+// "Previous video".
+export const nextVideoLink = "Next video";
+// Every keyword the facts fill, for one place. A Description prompt keeps the links as the
+// placeholders the description page fills when it shows the description, so they stay live.
+export function factValues(facts, title, place) {
+    const described = place === "description";
+    const links = facts?.links ?? [];
+    const previous = facts?.previousVideo;
+    return {
+        Slopify_Title: title,
+        Slopify_Channel: facts?.channel ?? "",
+        Slopify_Series: facts?.series?.trim() ?? "",
+        Slopify_Links: links
+            .map((link) => `${link.name}: ${described ? `{{${link.name}}}` : link.url}`)
+            .join("\n"),
+        Slopify_Previous_Video: described
+            ? `{{${previousVideoLink}}}`
+            : previous === undefined
+                ? ""
+                : previous.url === undefined || previous.url === ""
+                    ? previous.title
+                    : `${previous.title} – ${previous.url}`,
+        Slopify_Next_Video: described ? `{{${nextVideoLink}}}` : (facts?.nextVideo?.title ?? ""),
+        Slopify_Cast: (facts?.cast ?? [])
+            .map((member) => member.description.trim() === "" ? member.name : `${member.name}: ${member.description}`)
+            .join("\n"),
+        Slopify_Covered_Before: (facts?.coveredBefore ?? []).join("\n"),
+        Slopify_Episode: facts?.episode === undefined ? "" : String(facts.episode),
+        Slopify_Brand: brandText(facts?.brand),
+    };
+}
+function brandText(brand) {
+    if (brand === undefined)
+        return "";
+    const parts = [
+        ...(brand.colours.length === 0 ? [] : [`Colors: ${brand.colours.join(", ")}.`]),
+        ...(brand.fonts.length === 0 ? [] : [`Fonts: ${brand.fonts.join(", ")}.`]),
+    ];
+    return parts.join(" ");
+}
+// The finished length as the description and post prompts say it: "42 min 10 s".
+export function durationText(seconds) {
+    const total = Math.max(0, Math.round(seconds));
+    const hours = Math.floor(total / 3600);
+    const minutes = Math.floor((total % 3600) / 60);
+    const rest = total % 60;
+    return [
+        ...(hours > 0 ? [`${String(hours)} h`] : []),
+        ...(hours > 0 || minutes > 0 ? [`${String(minutes)} min`] : []),
+        `${String(rest)} s`,
+    ].join(" ");
+}
+// The release date as a prompt says it: "Thursday 22 October 2026", in the computer's time zone
+// like every other release time Slopify shows. Blank when the video has no release time.
+export function releaseDateText(at, timeZone) {
+    if (at === undefined || at.trim() === "")
+        return "";
+    const date = new Date(at);
+    if (Number.isNaN(date.getTime()))
+        return "";
+    return new Intl.DateTimeFormat("en-GB", {
+        weekday: "long",
+        day: "numeric",
+        month: "long",
+        year: "numeric",
+        ...(timeZone === undefined ? {} : { timeZone }),
+    })
+        .format(date)
+        .replace(",", "");
+}

package/dist/slices/library/lint.js

@@ -2,14 +2,35 @@
 // restated here - it is admission/substitute.ts's, imported so the editor and the run see
 // the same parse of the same body.
 import { detectSlots } from "../admission/substitute.js";
+import { keywordProblems, positionOf, textEntryProblems } from "../keywords/rules.js";
 import { bodyMax, nameMax } from "./model.js";
-export function lintPrompt(draft) {
-    return lint(draft.name, draft.body);
+// `grandfathered` is the slot names the stored template already has: an unknown `Slopify_`
+// name saved before the prefix was reserved does not stop it saving again.
+export function lintPrompt(draft, grandfathered = []) {
+    return lint(draft.name, draft.body, draft.kind, grandfathered);
 }
-export function lintEntry(draft) {
-    return lint(draft.name, draft.body);
+export function lintEntry(draft, grandfathered = []) {
+    return lint(draft.name, draft.body, draft.category, grandfathered);
 }
-function lint(name, body) {
+// What the editor warns about without refusing the save: a keyword where it can't be filled
+// (Play refuses that run) and an old keyword name.
+export function keywordWarnings(draft) {
+    return [
+        ...keywordProblems(draft.body, placeOf(draft))
+            .filter((problem) => problem.kind !== "unknown")
+            .map((problem) => ({ field: "body", message: problem.message })),
+        ...("category" in draft && draft.mode === "text"
+            ? textEntryProblems(draft.body, draft.category).map((problem) => ({
+                field: "body",
+                message: problem.message,
+            }))
+            : []),
+    ];
+}
+function placeOf(draft) {
+    return "kind" in draft ? draft.kind : draft.category;
+}
+function lint(name, body, place, grandfathered) {
     const fields = [];
     const trimmed = name.trim();
     if (trimmed === "") {
@@ -35,6 +56,9 @@ function lint(name, body) {
     for (const error of detectSlots(body).errors) {
         fields.push({ field: "body", message: describe(error, body) });
     }
+    for (const problem of keywordProblems(body, place, grandfathered))
+        if (problem.kind === "unknown")
+            fields.push({ field: "body", message: problem.message });
     return fields;
 }
 // The parser reports a character offset; a person editing a textarea counts lines and
@@ -50,10 +74,3 @@ function describe(error, body) {
             return `The keyword at ${where} contains a brace. Keywords cannot be placed inside other keywords.`;
     }
 }
-function positionOf(body, at) {
-    const before = body.slice(0, at);
-    const lastBreak = before.lastIndexOf("\n");
-    const line = before.split("\n").length;
-    const column = at - lastBreak;
-    return `line ${String(line)}, column ${String(column)}`;
-}

package/dist/slices/library/model.js

@@ -12,6 +12,8 @@ export const promptKinds = [
     "review",
     // Speaker turns for a multi-voice run (`slices/voices/script.ts`), in place of an article.
     "script",
+    // A post for another site: X, LinkedIn, Instagram, TikTok, Reddit or Bluesky (`slices/social`).
+    "post",
 ];
 export const entryCategories = ["intro", "outro"];
 // The name is what the Play pickers show, so it shares admission's title bound.

package/dist/slices/library/photorealistic.js

@@ -1,4 +1,4 @@
-import { readSetting, writeSetting } from "../settings/repo.js";
+import { deleteSetting, readJsonSetting, stringListSchema, writeSetting, } from "../settings/repo.js";
 import { promptById, promptByName } from "./repo.js";
 // An Image prompt's "Draws photorealistic pictures" tick (Library → Prompts → the prompt).
 // Pictures an image model draws in such a style can pass for real scenes, which is YouTube's
@@ -9,16 +9,8 @@ import { promptById, promptByName } from "./repo.js";
 // for when it is restored.
 export const photorealisticPromptsKey = "library.photorealisticPrompts";
 export function photorealisticPromptIds(db) {
-    const stored = readSetting(db, photorealisticPromptsKey);
-    if (stored === undefined)
-        return new Set();
-    try {
-        const value = JSON.parse(stored);
-        return new Set(Array.isArray(value) ? value.filter((one) => typeof one === "string") : []);
-    }
-    catch {
-        return new Set();
-    }
+    const value = readJsonSetting(db, photorealisticPromptsKey, stringListSchema, []);
+    return new Set(value);
 }
 // The prompts as the Library lists them, the ticked ones marked.
 export function withPhotorealistic(db, prompts) {
@@ -34,7 +26,7 @@ export function setPromptPhotorealistic(db, id, on) {
     const others = [...photorealisticPromptIds(db)].filter((one) => one !== id);
     const next = on ? [...others, id].sort() : others;
     if (next.length === 0)
-        db.prepare("DELETE FROM settings WHERE key = ?").run(photorealisticPromptsKey);
+        deleteSetting(db, photorealisticPromptsKey);
     else
         writeSetting(db, photorealisticPromptsKey, JSON.stringify(next));
     return { ok: true };

package/dist/slices/library/rename.js

@@ -1,8 +1,12 @@
 // Renaming a Library prompt or intro/outro. Everything names the Library by name, so a rename
-// rewrites the name where a future run reads it: each live template's current version (its
-// form and the copy it saved of the Library) and every Play draft still being edited.
-// Schedules run their template as it is now, so they follow. A project revision is the record
-// of what its run used and keeps the old name; "Used by" finds it through `formerNames`.
+// rewrites the name where a future run reads it: every version of every template (its form and
+// the copy it saved of the Library) and every Play draft still being edited. Every version, not
+// only the newest: a gated template's schedules run its newest ready version (`template-studio/
+// gate.ts`), Export pack carries that one, a version's samples stay current only while the
+// versions since differ in what they changed, and a trashed template or an old version can be
+// put back. The name is a pointer, so rewriting it changes nothing a version means. A project
+// revision is the record of what its run used and keeps the old name; "Used by" finds it
+// through `formerNames`.
 import { z } from "zod";
 import { renamedIn } from "./used-by.js";
 const documentRow = z.object({ key: z.string(), version: z.number(), document_json: z.string() });
@@ -56,9 +60,7 @@ export function renameReferences(db, ref, id, to) {
     if (ref.name === to)
         return;
     const templates = db
-        .prepare(`SELECT t.id AS key,r.version,r.document_json FROM project_templates t
-       JOIN project_template_revisions r ON r.template_id=t.id AND r.version=t.head_version
-       WHERE t.deleted_at IS NULL`)
+        .prepare("SELECT template_id AS key,version,document_json FROM project_template_revisions")
         .all()
         .map((row) => documentRow.parse(row));
     const updateTemplate = db.prepare("UPDATE project_template_revisions SET document_json=? WHERE template_id=? AND version=?");

package/dist/slices/library/save.js

@@ -8,8 +8,10 @@ import { lintEntry, lintPrompt } from "./lint.js";
 import { entryCategories, promptKinds } from "./model.js";
 import { renameReferences } from "./rename.js";
 import { entryById, insertEntry, insertPrompt, promptById, replaceEntry, replacePrompt, trashEntry, trashPrompt, } from "./repo.js";
-export function createPrompt(deps, draft) {
-    const fields = lintPrompt(draft);
+// Duplicate names the prompt it copies (`copyOf`): a copy of one saved before the `Slopify_`
+// prefix was reserved keeps the names that one already has, as an edit of it would.
+export function createPrompt(deps, draft, copyOf) {
+    const fields = lintPrompt(draft, copyOf === undefined ? [] : (promptById(deps.db, copyOf)?.slots ?? []));
     if (fields.length > 0) {
         return { ok: false, reason: "invalid", fields };
     }
@@ -21,8 +23,10 @@ export function createPrompt(deps, draft) {
     }, prompt);
 }
 // A save replaces the row whole and keeps what was saved as a new version (`history.ts`).
-export function updatePrompt(deps, id, draft, restoredFrom) {
-    const fields = lintPrompt(draft);
+export function updatePrompt(deps, id, draft, restoredFrom, 
+// Slot names a restored version had, which it may keep like the stored prompt's.
+restoredSlots = []) {
+    const fields = lintPrompt(draft, [...(promptById(deps.db, id)?.slots ?? []), ...restoredSlots]);
     if (fields.length > 0) {
         return { ok: false, reason: "invalid", fields };
     }
@@ -47,7 +51,7 @@ export function restorePrompt(deps, id, version) {
     const kind = promptKinds.find((one) => one === saved?.kind);
     if (saved === undefined || kind === undefined || promptById(deps.db, id) === undefined)
         return { ok: false, reason: "not-found" };
-    return updatePrompt(deps, id, { kind, name: saved.name, body: saved.body }, version);
+    return updatePrompt(deps, id, { kind, name: saved.name, body: saved.body }, version, detectSlots(saved.body).names);
 }
 // Delete moves the prompt to the trash with its history; Settings → Trash puts it back or
 // removes it for good (`slices/trash`).
@@ -56,8 +60,8 @@ export function removePrompt(deps, id) {
         ? { ok: true, value: null }
         : { ok: false, reason: "not-found" };
 }
-export function createEntry(deps, draft) {
-    const fields = lintEntry(draft);
+export function createEntry(deps, draft, copyOf) {
+    const fields = lintEntry(draft, copyOf === undefined ? [] : (entryById(deps.db, copyOf)?.slots ?? []));
     if (fields.length > 0) {
         return { ok: false, reason: "invalid", fields };
     }
@@ -68,8 +72,8 @@ export function createEntry(deps, draft) {
         return true;
     }, entry);
 }
-export function updateEntry(deps, id, draft, restoredFrom) {
-    const fields = lintEntry(draft);
+export function updateEntry(deps, id, draft, restoredFrom, restoredSlots = []) {
+    const fields = lintEntry(draft, [...(entryById(deps.db, id)?.slots ?? []), ...restoredSlots]);
     if (fields.length > 0) {
         return { ok: false, reason: "invalid", fields };
     }
@@ -95,7 +99,7 @@ export function restoreEntry(deps, id, version) {
         saved.mode === null ||
         entryById(deps.db, id) === undefined)
         return { ok: false, reason: "not-found" };
-    return updateEntry(deps, id, { category, mode: saved.mode, name: saved.name, body: saved.body }, version);
+    return updateEntry(deps, id, { category, mode: saved.mode, name: saved.name, body: saved.body }, version, detectSlots(saved.body).names);
 }
 export function removeEntry(deps, id) {
     return trashEntry(deps.db, id, deps.clock.now().toISOString())

package/dist/slices/library/slots.js

@@ -4,8 +4,10 @@
 // is the one that runs.
 import { normaliseDraft, usesNarrationPreparation, usesReference, usesShorts, usesYoutubeDescription, } from "../admission/rules.js";
 import { collectFields, render } from "../admission/substitute.js";
+import { keywordUsesOf, placedProblems, textEntryProblems, } from "../keywords/rules.js";
 import { reviewPromptKey } from "../reviews/model.js";
 import { activeReviewStages } from "../reviews/rules.js";
+import { socialPlatforms } from "../social/model.js";
 import { usesScriptPrompt } from "../voices/model.js";
 import { snapshotEntry, snapshotPrompt } from "./snapshot.js";
 export function pickTemplates(db, input, snapshot) {
@@ -16,40 +18,63 @@ export function pickTemplates(db, input, snapshot) {
     // entries. A prompt left selected on a stage set to Provide asks for no field.
     const text = [];
     const image = [];
+    // Each picked body with the field that picked it and where it runs, for the keyword checks.
+    const placed = [];
+    const body = (kind, name, field, into, key) => {
+        const at = into.length;
+        pickBody(db, kind, name, field, missing, into, key, snapshot);
+        const picked = into[at];
+        if (picked !== undefined)
+            placed.push({ field, place: kind, body: picked.body });
+    };
     if (sources.article === "generate") {
-        body(db, usesScriptPrompt(draft) ? "script" : "article", draft.articlePrompt, "articlePrompt", missing, text, "article", snapshot);
+        body(usesScriptPrompt(draft) ? "script" : "article", draft.articlePrompt, "articlePrompt", text, "article");
     }
     const intro = pickEntry(db, "intro", draft.intro, missing, snapshot);
     if (usesNarrationPreparation(draft))
-        body(db, "narration", draft.narrationPrompt, "narrationPrompt", missing, text, "narration", snapshot);
+        body("narration", draft.narrationPrompt, "narrationPrompt", text, "narration");
     const outro = pickEntry(db, "outro", draft.outro, missing, snapshot);
     push(text, "intro", intro);
     push(text, "outro", outro);
+    if (intro !== undefined)
+        placed.push({ field: "intro", place: "intro", body: intro.body });
+    if (outro !== undefined)
+        placed.push({ field: "outro", place: "outro", body: outro.body });
+    for (const entry of [intro, outro])
+        if (entry?.mode === "text")
+            missing.push(...textEntryProblems(entry.body, entry.category));
     // None picked is the built-in prompt, which asks for no keyword.
     if (usesYoutubeDescription(draft))
-        body(db, "description", draft.descriptionPrompt, "descriptionPrompt", missing, text, "description", snapshot);
+        body("description", draft.descriptionPrompt, "descriptionPrompt", text, "description");
     // The Shorts step's two prompts; none picked is the built-in one, which asks for no keyword.
     if (usesShorts(draft)) {
-        body(db, "shorts", draft.shorts?.prompt, "shorts.prompt", missing, text, "shorts", snapshot);
-        body(db, "image", draft.shorts?.imagePrompt, "shorts.imagePrompt", missing, image, "shortsImage", snapshot);
+        body("shorts", draft.shorts?.prompt, "shorts.prompt", text, "shorts");
+        body("image", draft.shorts?.imagePrompt, "shorts.imagePrompt", image, "shortsImage");
     }
     // Each reviewed stage's Review prompt; none picked is the stage's built-in one.
     for (const stage of activeReviewStages(draft))
-        body(db, "review", draft.reviews?.stages[stage]?.prompt, `reviews.stages.${stage}.prompt`, missing, text, reviewPromptKey(stage), snapshot);
+        body("review", draft.reviews?.stages[stage]?.prompt, `reviews.stages.${stage}.prompt`, text, reviewPromptKey(stage));
+    // Each site's post prompt, for the sites whose post is written by one; none picked is the
+    // built-in one, which asks for no keyword.
+    for (const platform of socialPlatforms) {
+        const choice = draft.social?.[platform];
+        if (choice?.text === "prompt")
+            body("post", choice.prompt, `social.${platform}.prompt`, text, `social.${platform}`);
+    }
     // The establishing image's prompt: an image prompt like the others, its keywords filled the
     // same way.
     if (usesReference(draft) && draft.reference?.source === "prompt")
-        body(db, "image", draft.reference.prompt, "reference.prompt", missing, image, "referencePrompt", snapshot);
+        body("image", draft.reference.prompt, "reference.prompt", image, "referencePrompt");
     if (sources.images === "generate") {
         for (const [index, picked] of draft.imagePrompts.entries()) {
             const field = `imagePrompts.${String(index)}.name`;
-            body(db, "image", picked.name, field, missing, image, `imagePrompts.${String(index)}`, snapshot);
+            body("image", picked.name, field, image, `imagePrompts.${String(index)}`);
         }
     }
     // Both Generate modes read the same thumbnail template; Prompt by LLM sends it as the
     // instruction rather than to the image provider.
     if (sources.thumbnail === "from_prompt" || sources.thumbnail === "prompt_by_llm") {
-        body(db, "thumbnail", draft.thumbnailPrompt, "thumbnailPrompt", missing, image, "thumbnailPrompt", snapshot);
+        body("thumbnail", draft.thumbnailPrompt, "thumbnailPrompt", image, "thumbnailPrompt");
     }
     const fields = collectFields(
     // The title's keywords are asked for like a prompt's: it is filled when the run starts.
@@ -58,7 +83,8 @@ export function pickTemplates(db, input, snapshot) {
         draft: withSavedModes(draft, intro, outro),
         requiredSlots: fields.map((field) => field.name),
         bodies: [...text, ...image],
-        missing,
+        missing: [...missing, ...placedProblems(placed, draft.values)],
+        keywordUses: keywordUsesOf(placed),
     };
 }
 // Every picked body rendered once, with the trimmed values
@@ -70,7 +96,7 @@ export function renderPicked(picked, values) {
     }
     return rendered;
 }
-function body(db, kind, name, field, missing, into, key, snapshot) {
+function pickBody(db, kind, name, field, missing, into, key, snapshot) {
     // An unpicked prompt is admission's rule to state ("Pick an article prompt."), so this
     // says nothing about it and the form marks the field once.
     if (name === undefined || name.trim() === "") {

package/dist/slices/library/transfer.js

@@ -1,4 +1,5 @@
 import { z } from "zod";
+import { errorText } from "../../kernel/errors.js";
 import { documentThemeNameMax } from "../document/model.js";
 import { documentThemeSchema } from "../document/theme-schema.js";
 import { narrationAliasSchema } from "../narration/aliases-schema.js";
@@ -16,7 +17,7 @@ const itemsMax = 10_000;
 const sectionWords = {
     prompts: "prompts",
     entries: "intros or outros",
-    documentThemes: "PDF themes",
+    documentThemes: "PDF looks",
     aliases: "narration aliases",
 };
 // Only what the editor saves travels: ids, keywords and dates are the install's own.
@@ -92,7 +93,7 @@ export function readLibraryFile(text, section) {
     if (!top.success)
         return {
             ok: false,
-            message: `it isn't a Slopify library file. Pick a .json file made by Export on a Library tab.`,
+            message: `it isn't a Slopify library file. Pick a .json file made by Export on Prompts, the PDF looks or Pronunciation.`,
         };
     if (top.data.version > libraryFileVersion)
         return {
@@ -103,7 +104,7 @@ export function readLibraryFile(text, section) {
     if (!Array.isArray(list) || list.length === 0)
         return {
             ok: false,
-            message: `it holds no ${what}. Import it on the Library tab it was exported from.`,
+            message: `it holds no ${what}. Import it on the page it was exported from.`,
         };
     if (list.length > itemsMax)
         return {
@@ -202,7 +203,7 @@ export async function importEach(items, options) {
                     break;
             }
             catch (error) {
-                reason = error instanceof Error ? error.message : String(error);
+                reason = errorText(error);
                 break;
             }
         }

package/dist/slices/library/used-by.js

@@ -5,6 +5,8 @@
 // `snapshot.ts` use.
 import { z } from "zod";
 import { liveProject } from "../admission/repo.js";
+import { runnableVersion } from "../template-studio/gate.js";
+import { gatesOf, notTemplateSample } from "../template-studio/repo.js";
 function record(value) {
     return typeof value === "object" && value !== null && !Array.isArray(value)
         ? value
@@ -40,6 +42,8 @@ function slotsIn(value, ref) {
             return field(shorts, "prompt");
         case "review":
             return Object.values(record(record(value.reviews)?.stages) ?? {}).flatMap((stage) => field(record(stage), "prompt"));
+        case "post":
+            return Object.values(record(value.social) ?? {}).flatMap((one) => field(record(one), "prompt"));
         case "image":
             return [
                 ...(Array.isArray(value.imagePrompts) ? value.imagePrompts : []).flatMap((one) => field(record(one), "name")),
@@ -106,18 +110,28 @@ export function usedBy(db, ref, formerNames = []) {
         const parsed = templateRow.parse(row);
         return matches(formOf(parsed.document_json)) ? [{ id: parsed.id, name: parsed.name }] : [];
     });
-    // A schedule runs its template as it is now (`schedules/scheduler.ts`), so the template's
-    // current version is what it uses.
+    // A schedule runs the version of its template the studio gate lets it run
+    // (`template-studio/schedule-gate.ts`), which can be older than the newest one.
+    const runsOf = new Map();
+    const formRun = (templateId, head) => {
+        if (!runsOf.has(templateId)) {
+            const version = runnableVersion(gatesOf(db, templateId), head) ?? head;
+            const row = db
+                .prepare("SELECT document_json FROM project_template_revisions WHERE template_id=? AND version=?")
+                .get(templateId, version);
+            runsOf.set(templateId, row === undefined ? undefined : formOf(String(row.document_json)));
+        }
+        return runsOf.get(templateId);
+    };
     const schedules = db
-        .prepare(`SELECT s.id,s.name,s.status,r.document_json FROM schedules s
+        .prepare(`SELECT s.id,s.name,s.status,s.template_id,t.head_version FROM schedules s
        JOIN project_templates t ON t.id=s.template_id
-       JOIN project_template_revisions r ON r.template_id=t.id AND r.version=t.head_version
-       WHERE s.deleted_at IS NULL AND s.status IN ('active','paused') AND ${mentions("r.document_json")}
+       WHERE s.deleted_at IS NULL AND s.status IN ('active','paused') AND s.template_kind='project'
        ORDER BY lower(s.name),s.id`)
-        .all(...plain)
+        .all()
         .flatMap((row) => {
         const parsed = scheduleRow.parse(row);
-        return matches(formOf(parsed.document_json))
+        return matches(formRun(parsed.template_id, parsed.head_version))
             ? [{ id: parsed.id, name: parsed.name, status: parsed.status }]
             : [];
     });
@@ -126,7 +140,7 @@ export function usedBy(db, ref, formerNames = []) {
         (SELECT count(*) FROM project_revisions x WHERE x.project_id=r.project_id) AS total
        FROM project_revisions r JOIN projects p ON p.id=r.project_id
        LEFT JOIN project_heads h ON h.project_id=r.project_id
-       WHERE ${liveProject("p")} AND ${mentions("r.config")}
+       WHERE ${liveProject("p")} AND ${notTemplateSample("p")} AND ${mentions("r.config")}
        ORDER BY p.created_at DESC,p.id,r.created_at`)
         .all(...plain)
         .map((row) => revisionRow.parse(row));
@@ -146,7 +160,7 @@ export function usedBy(db, ref, formerNames = []) {
     // A project from before revisions existed has none yet; its own config is what it ran with.
     for (const row of db
         .prepare(`SELECT p.id,p.title,p.config FROM projects p
-       WHERE ${liveProject("p")}
+       WHERE ${liveProject("p")} AND ${notTemplateSample("p")}
        AND NOT EXISTS (SELECT 1 FROM project_revisions r WHERE r.project_id=p.id)
        AND ${mentions("p.config")} ORDER BY p.created_at DESC,p.id`)
         .all(...plain)
@@ -167,7 +181,8 @@ const scheduleRow = z.object({
     id: z.string(),
     name: z.string(),
     status: z.string(),
-    document_json: z.string(),
+    template_id: z.string(),
+    head_version: z.number(),
 });
 const revisionRow = z.object({
     project_id: z.string(),

package/dist/slices/loudness/loudnorm.js

@@ -1,5 +1,6 @@
-import { renameSync } from "node:fs";
+import { renameSync, rmSync } from "node:fs";
 import { z } from "zod";
+import { editSettings } from "../../kernel/ports/settings-places.js";
 import { runFfmpeg } from "../video/ffmpeg.js";
 // ceiling: below -70 LUFS (the filter's absolute gate) or under 0.4 s (one measuring block)
 // there is nothing to measure: a piece that quiet or short is left as it is.
@@ -89,6 +90,7 @@ export async function measure(run, input, goal, chain = "") {
         args: measureArgs(input, goal, chain),
         signal: run.signal,
         log: run.log,
+        priority: run.priority,
         onProgress: () => { },
         onStderr: (text) => {
             stderr += text;
@@ -96,7 +98,7 @@ export async function measure(run, input, goal, chain = "") {
     });
     const measured = parseMeasured(stderr);
     if (measured === undefined)
-        throw new Error("Slopify couldn't measure the loudness of the narration (ffmpeg printed no measurement). Try again; if it happens again, turn off Level the volume in Edit project → Pauses and volume, then use Download diagnostics in Settings and report it.");
+        throw new Error(`Slopify couldn't measure the loudness of the narration (ffmpeg printed no measurement). Try again; if it happens again, turn off Level the volume in ${editSettings.sound}, then use Download diagnostics in Settings and report it.`);
     return measured;
 }
 export async function measureFile(run, path, goal) {
@@ -109,8 +111,15 @@ export async function normalizeFile(run, input, output, goal, format) {
     // The channels are set before anything is measured: a stereo file mixed down to mono is
     // louder per channel (the mix keeps its loudness, so each sample is about 3 dB higher), and
     // the gain and the peak ceiling have to hold for what is written.
-    const channels = `aformat=channel_layouts=${format.channels === 1 ? "mono" : "stereo"}`;
-    const measured = await measure(run, ["-i", input], goal, channels);
+    const measured = await measure(run, ["-i", input], goal, channelsOf(format));
+    await writeLevelled(run, input, output, goal, measured, format);
+    return measured;
+}
+function channelsOf(format) {
+    return `aformat=channel_layouts=${format.channels === 1 ? "mono" : "stereo"}`;
+}
+// `input` written to `output` with the gain that takes `measured` to `goal` (`gainFilter`).
+async function writeLevelled(run, input, output, goal, measured, format) {
     await runFfmpeg({
         bin: run.bin,
         args: [
@@ -125,7 +134,7 @@ export async function normalizeFile(run, input, output, goal, format) {
             "-map",
             "0:a:0",
             "-af",
-            `${channels},${gainFilter(goal, measured, format.sampleRate)}`,
+            `${channelsOf(format)},${gainFilter(goal, measured, format.sampleRate)}`,
             "-c:a",
             "pcm_f32le",
             "-f",
@@ -134,23 +143,55 @@ export async function normalizeFile(run, input, output, goal, format) {
         ],
         signal: run.signal,
         log: run.log,
+        priority: run.priority,
         onProgress: () => { },
     });
-    return measured;
 }
 // ceiling: within a fifth of a LU of the level no one hears the difference.
 const correctAbove = 0.2;
-// One file brought to `goal` (`normalizeFile`), measured again, and corrected once when the
-// limiter left it more than a fifth of a LU under the level. Returns the first measurement and
-// the last.
+// ceiling: two corrections land a voice whose peaks stand 17 dB over its loudness (raw TTS)
+// within a few tenths of a LU; a third would only chase the last tenth.
+const corrections = 2;
+// ceiling: the most gain a correction adds over the plain gain. Past it the limiter would
+// flatten the voice; the file lands short instead, and its report says how far.
+const extraGainMax = 6;
+// One file brought to `goal` (`normalizeFile`), measured again, and corrected when the limiter
+// left it more than a fifth of a LU under the level. Each correction writes the input again
+// with more gain (so the limiter holds the peaks once, never twice), the gain found from the
+// levels the earlier gains gave. Returns the first measurement and the last.
 export async function levelFile(run, input, output, goal, format) {
     const before = await normalizeFile(run, input, output, goal, format);
     let after = await measureFile(run, output, goal);
-    if (measurable(after) && Math.abs(after.integrated - goal.lufs) > correctAbove) {
+    if (!measurable(before))
+        return { before, after };
+    const plain = goal.lufs - before.integrated;
+    const tried = [
+        { gain: plain, lufs: after.integrated },
+    ];
+    for (let pass = 0; pass < corrections; pass += 1) {
+        if (!measurable(after) || Math.abs(after.integrated - goal.lufs) <= correctAbove)
+            break;
+        const last = tried[tried.length - 1];
+        const prior = tried[tried.length - 2];
+        if (last === undefined)
+            break;
+        const slope = prior === undefined || last.lufs === prior.lufs
+            ? 1
+            : (last.lufs - prior.lufs) / (last.gain - prior.gain);
+        const gain = Math.min(plain + extraGainMax, last.gain + (goal.lufs - last.lufs) / (slope > 0.05 ? slope : 1));
+        if (Math.abs(gain - last.gain) < 0.05)
+            break;
         const corrected = `${output}.corrected.wav`;
-        await normalizeFile(run, output, corrected, goal, format);
+        await writeLevelled(run, input, corrected, { ...goal, lufs: before.integrated + gain }, before, format);
+        const next = await measureFile(run, corrected, goal);
+        // A correction that landed further off than the file it would replace is dropped.
+        if (Math.abs(next.integrated - goal.lufs) >= Math.abs(after.integrated - goal.lufs)) {
+            rmSync(corrected, { force: true });
+            break;
+        }
         renameSync(corrected, output);
-        after = await measureFile(run, output, goal);
+        after = next;
+        tried.push({ gain, lufs: next.integrated });
     }
     return { before, after };
 }

package/dist/slices/loudness/model.js

@@ -9,10 +9,11 @@
 // The ambient bed and the shorts' music are mixed before the master and ducked by the levelled
 // voice, so they keep their place under it.
 //
-// Browser-safe: Play, Edit project and Settings convert the same numbers.
+// Browser-safe: Create, Edit settings and Settings convert the same numbers.
 //
 // A run without the setting has no `loudness` at all, which is what every project saved before
 // it was, and every recipe leaves it out of its fingerprint then.
+import { createRows, editSettings } from "../../kernel/ports/settings-places.js";
 // YouTube and Spotify play everything back at about -14 LUFS, so a video mastered there is
 // neither turned down nor left quiet. Audiobook shops ask for about -18 to -20 LUFS with more
 // headroom, so the listening files sit lower.
@@ -87,7 +88,7 @@ export function loudnessProblem(lufs, target) {
         db >= loudnessDbMin - 1e-9 &&
         db <= loudnessDbMax + 1e-9
         ? undefined
-        : `The volume must be between ${String(loudnessDbMin)} dB and +${String(loudnessDbMax)} dB of the recommended level, in steps of ${String(loudnessDbStep)} dB. Change it under Level the volume (Play → Video and style, Edit project → Pauses and volume, or Settings → General).`;
+        : `The volume must be between ${String(loudnessDbMin)} dB and +${String(loudnessDbMax)} dB of the recommended level, in steps of ${String(loudnessDbStep)} dB. Change it under Level the volume (${createRows.video}, ${editSettings.sound}, or Settings → Production defaults).`;
 }
 export function loudnessFields(settings) {
     if (settings === undefined)