@gentbajko/slopify 2.5.0 → 3.0.0

package/dist/slices/onboarding/install.js

@@ -0,0 +1,119 @@
+import { transact } from "../../kernel/db/tx.js";
+import { detectSlots } from "../admission/substitute.js";
+import { nameMax } from "../library/model.js";
+import { insertPrompt, promptById, promptByName } from "../library/repo.js";
+import { templateById, templateSummaries } from "../project-templates/repo.js";
+import { createTemplate } from "../project-templates/service.js";
+import { insertVoice, listVoices } from "../settings/repo.js";
+import { packById } from "./packs.js";
+import { readPackRecords, writePackRecord } from "./state.js";
+import { packTemplate } from "./template.js";
+// Installs a pack into the library. Idempotent: an item the pack installed before, and that
+// still exists, is kept as it is, even if it was edited since. A user's own prompt, voice or
+// template is never overwritten: a prompt with the pack's name and the pack's exact text is
+// used as it is, and any other one with that name makes the pack's copy come in as
+// "<name> (2)". `template: false` installs the prompts and the voice only (what the
+// 60-second short needs).
+export function installPack(deps, packId, options = { template: true }) {
+    const pack = packById(packId);
+    if (pack === undefined)
+        return { ok: false, reason: "not-found" };
+    return transact(deps.db, () => {
+        const before = readPackRecords(deps.db)[pack.id];
+        let added = false;
+        const at = deps.clock.now().toISOString();
+        const prompts = {};
+        const names = {};
+        for (const prompt of pack.prompts) {
+            const kept = before?.prompts[prompt.key];
+            const existing = kept === undefined ? undefined : promptById(deps.db, kept);
+            if (existing !== undefined) {
+                prompts[prompt.key] = existing.id;
+                names[prompt.key] = existing.name;
+                continue;
+            }
+            const placed = placePrompt(deps, prompt.kind, prompt.name, prompt.body, at);
+            added ||= placed.added;
+            prompts[prompt.key] = placed.prompt.id;
+            names[prompt.key] = placed.prompt.name;
+        }
+        const voice = placeVoice(deps, pack, before);
+        added ||= voice.added;
+        let template = before?.template;
+        if (options.template && (template === undefined || !templateById(deps.db, template))) {
+            const id = deps.uuid();
+            const created = createTemplate(deps, {
+                id,
+                name: freeTemplateName(deps, `${pack.name} starter`),
+                document: packTemplate(pack, names, voice.voiceId),
+            });
+            if (!created.ok)
+                throw new Error(`Slopify hit an internal error (the ${pack.name} pack's template was refused: ${created.reason}). Try again; if it happens again, use Download diagnostics in Settings and report it.`);
+            template = id;
+            added = true;
+        }
+        const record = {
+            installedAt: before?.installedAt ?? at,
+            prompts,
+            voice: voice.id,
+            ...(template === undefined ? {} : { template }),
+        };
+        writePackRecord(deps.db, pack.id, record);
+        return {
+            ok: true,
+            value: {
+                packId: pack.id,
+                added,
+                prompts: names,
+                templateId: template ?? null,
+                voice: voice.voiceId,
+            },
+        };
+    });
+}
+function placePrompt(deps, kind, name, body, at) {
+    for (let n = 1;; n += 1) {
+        const suffix = n === 1 ? "" : ` (${String(n)})`;
+        const candidate = `${name.slice(0, nameMax - suffix.length).trimEnd()}${suffix}`;
+        const taken = promptByName(deps.db, kind, candidate);
+        if (taken?.body === body)
+            return { prompt: taken, added: false };
+        if (taken !== undefined)
+            continue;
+        const prompt = {
+            id: deps.ids.next(),
+            kind,
+            name: candidate,
+            body,
+            slots: detectSlots(body).names,
+            updatedAt: at,
+        };
+        insertPrompt(deps.db, prompt);
+        return { prompt, added: true };
+    }
+}
+// The suggested voice as a saved voice, so Play's voice list offers it. One the user already
+// saved under the same voice ID is used as it is, name and all.
+function placeVoice(deps, pack, before) {
+    const voices = listVoices(deps.db);
+    const same = voices.find((voice) => (before?.voice !== undefined && voice.id === before.voice) ||
+        (voice.provider === pack.voice.provider && voice.voiceId === pack.voice.voiceId));
+    if (same !== undefined)
+        return { id: same.id, voiceId: same.voiceId, added: false };
+    const id = deps.ids.next();
+    insertVoice(deps.db, {
+        id,
+        provider: pack.voice.provider,
+        name: pack.voice.name,
+        voiceId: pack.voice.voiceId,
+    });
+    return { id, voiceId: pack.voice.voiceId, added: true };
+}
+function freeTemplateName(deps, name) {
+    const taken = new Set(templateSummaries(deps.db).map((row) => row.name.toLowerCase()));
+    for (let n = 1;; n += 1) {
+        const candidate = n === 1 ? name : `${name} (${String(n)})`;
+        if (!taken.has(candidate.toLowerCase()))
+            return candidate;
+    }
+}

package/dist/slices/onboarding/model.js

@@ -0,0 +1,64 @@
+import { z } from "zod";
+// The first five minutes' wire shapes. Browser-safe: the first-run screen, the Library's
+// packs and the sample's page read these too.
+// The bundled samples: "The Library of Alexandria" (narrated video, shorts, article, PDF) and
+// two multi-voice demos, an audiobook and a podcast.
+export const sampleIds = ["library", "audiobook", "podcast"];
+// Each sample's project, or null while it is not in Projects.
+export const sampleProjectsSchema = z.object({
+    library: z.string().nullable(),
+    audiobook: z.string().nullable(),
+    podcast: z.string().nullable(),
+});
+export const firstRunViewSchema = z
+    .object({
+    show: z.boolean(),
+    // The Library of Alexandria's project, as before the demos; `samples` has all three.
+    sampleProjectId: z.string().nullable(),
+    samples: sampleProjectsSchema,
+    clis: z.array(z.object({
+        id: z.enum(["claude-code", "codex", "gemini"]),
+        name: z.string(),
+        installed: z.boolean(),
+        ready: z.boolean(),
+        version: z.string().nullable(),
+        issue: z.string().nullable(),
+        // Codex draws the images too.
+        draws: z.boolean(),
+    })),
+    packs: z.array(z.object({
+        id: z.string(),
+        name: z.string(),
+        summary: z.string(),
+        installed: z.boolean(),
+        templateId: z.string().nullable(),
+    })),
+})
+    .readonly();
+export const quickShortInputSchema = z
+    .object({
+    topic: z.string().trim().min(1).max(200),
+    packId: z.string().max(40).optional(),
+    requestId: z.uuid(),
+})
+    .strict();
+export const quickShortResultSchema = z.object({
+    projectId: z.string(),
+    replayed: z.boolean(),
+});
+export const installedPackSchema = z.object({
+    packId: z.string(),
+    added: z.boolean(),
+    prompts: z.record(z.string(), z.string()),
+    templateId: z.string().nullable(),
+    voice: z.string().nullable(),
+});
+export const sampleCopySchema = z.object({ projectId: z.string() });
+export const sampleStateSchema = z.object({
+    projectId: z.string().nullable(),
+    samples: sampleProjectsSchema,
+});
+// Make my own copy of one sample; without a project, the Library of Alexandria.
+export const sampleCopyInputSchema = z
+    .object({ projectId: z.string().max(60).optional() })
+    .strict();

package/dist/slices/onboarding/packs.js

@@ -0,0 +1,491 @@
+import { defaultVideoEdit } from "../video/edit-settings.js";
+const look = (grade, change = {}) => ({
+    ...defaultVideoEdit,
+    grade,
+    ...change,
+});
+const noText = "No text, letters, captions, logos or watermarks anywhere in the image.";
+const shortsRules = (voice) => [
+    "Pick the moments of this video that work best as vertical shorts.",
+    "",
+    "- Each short must stand on its own for someone who never saw the video.",
+    `- ${voice}`,
+    "- Start on the strongest line of the moment, never on a transition or a recap.",
+    "- End on a finished thought.",
+    "- Title: at most 60 characters, plain words, no clickbait, no hashtags.",
+    "- Description: one line saying what the viewer gets from the clip.",
+    "- Hashtags: 3-5, about the subject of that clip, most specific last.",
+].join("\n");
+const scriptRules = [
+    "Write the narration for a vertical short video of about 60 seconds: 140-160 words.",
+    "",
+    "- Plain prose only: no headings, lists, stage directions, emoji or hashtags. Every word is spoken.",
+    "- The first sentence is the hook. It must make someone stop scrolling on its own.",
+    "- One idea, told as a small story with a turn near the end. No introduction, no recap, no 'in this video'.",
+    "- Short sentences that read well aloud. Spell out numbers the way they are said.",
+    "- End on a line that lands, not on a call to subscribe.",
+    "- Stay accurate. If something is uncertain or disputed, say so in a few words instead of inventing detail.",
+].join("\n");
+export const starterPacks = [
+    {
+        id: "sleep-lore",
+        name: "Sleep lore",
+        summary: "Slow, gentle myths and legends told to fall asleep to.",
+        voice: {
+            provider: "openai-tts",
+            model: "gpt-4o-mini-tts",
+            voiceId: "sage",
+            name: "Sleep lore narrator (calm)",
+        },
+        style: {
+            research: false,
+            expectedWords: 3000,
+            imagesPerVideo: 16,
+            imageSeconds: 24,
+            zoomPercent: 8,
+            motionStyle: "pan",
+            captions: { fontSize: 44, position: "bottom" },
+            videoEdit: look("warm", {
+                transition: "crossfade",
+                transitionSeconds: 2,
+                vignette: "subtle",
+                atmosphere: "fog",
+            }),
+        },
+        prompts: [
+            {
+                key: "article",
+                kind: "article",
+                name: "Sleep lore · Story",
+                body: [
+                    "Write a long, calm bedtime telling of {{topic}} for listeners who want to drift off to sleep.",
+                    "",
+                    "Length: 2800-3200 words.",
+                    "",
+                    "Voice and pace:",
+                    "- Warm, unhurried and low in energy, like someone reading by lamplight. Nothing sudden, loud or frightening; soften any violence in the source into distance and suggestion.",
+                    "- Long, flowing sentences with gentle rhythm. Linger on places, weather, light, textures and small sounds.",
+                    "- Speak to the listener now and then ('you might picture...'), never more than once a section.",
+                    "",
+                    "Shape:",
+                    "- Open by setting a quiet scene, not with the plot.",
+                    "- Follow the story in order, in six to eight parts, each with a short heading.",
+                    "- No cliffhangers and no questions left hanging. Each part settles before the next begins.",
+                    "- Close with a slow, peaceful ending that asks nothing of the listener.",
+                    "",
+                    "Accuracy: tell the story as the traditional sources tell it. Where versions differ, choose one and mention the other in a single soft sentence. Do not invent names, dates or quotations.",
+                    "",
+                    "Return only the story in Markdown: the headings and the prose. No title page, notes or sources.",
+                ].join("\n"),
+            },
+            {
+                key: "shortScript",
+                kind: "article",
+                name: "Sleep lore · 60-second short",
+                body: [
+                    `Subject: {{topic}}, as a quiet moment from myth or legend.`,
+                    "",
+                    scriptRules,
+                    "- Keep the calm, hushed tone of a bedtime story even in the hook.",
+                ].join("\n"),
+            },
+            {
+                key: "image",
+                kind: "image",
+                name: "Sleep lore · Scene",
+                body: [
+                    "A dreamlike painted scene from the legend of {{topic}}.",
+                    "Soft oil-painting style with visible brushwork, muted twilight palette of deep blues, dusky violets and warm lantern gold.",
+                    "Gentle haze, low contrast, wide calm composition with plenty of quiet space; distant figures rather than close faces.",
+                    "Peaceful and still, nothing threatening in the frame.",
+                    noText,
+                ].join(" "),
+            },
+            {
+                key: "thumbnail",
+                kind: "thumbnail",
+                name: "Sleep lore · Thumbnail",
+                body: [
+                    "A serene, luminous painting that sums up the legend of {{topic}} in one image.",
+                    "One clear subject in the centre, soft moonlight or candlelight, deep blue and gold palette, gentle mist.",
+                    "Readable at a small size: simple shapes, strong silhouette, calm mood.",
+                    noText,
+                ].join(" "),
+            },
+            {
+                key: "description",
+                kind: "description",
+                name: "Sleep lore · Description",
+                body: [
+                    "Write a YouTube description for a sleep story about {{topic}}.",
+                    "",
+                    "- Summary: 2-3 soothing sentences that invite the viewer to settle in. Mention that the story is calm and has no sudden sounds.",
+                    "- Chapters: one per part of the story, 3-6 gentle words each.",
+                    "- Hashtags: 3-5, such as sleep story and the legend's subject, most specific last.",
+                    "- Tags: 15-25 search terms for sleep stories, bedtime myths and the legend itself.",
+                ].join("\n"),
+            },
+            {
+                key: "shorts",
+                kind: "shorts",
+                name: "Sleep lore · Shorts",
+                body: shortsRules("Prefer the most beautiful or mysterious images of the story over plot summaries."),
+            },
+        ],
+    },
+    {
+        id: "true-crime",
+        name: "True crime",
+        summary: "Measured, factual retellings of real cases and their investigations.",
+        voice: {
+            provider: "openai-tts",
+            model: "gpt-4o-mini-tts",
+            voiceId: "onyx",
+            name: "True crime narrator (deep)",
+        },
+        style: {
+            research: true,
+            expectedWords: 2400,
+            imagesPerVideo: 18,
+            imageSeconds: 10,
+            zoomPercent: 15,
+            motionStyle: "zoom",
+            captions: { fontSize: 52, position: "bottom" },
+            videoEdit: look("desaturated", {
+                transition: "fadeblack",
+                transitionSeconds: 0.8,
+                vignette: "strong",
+                grain: "subtle",
+            }),
+        },
+        prompts: [
+            {
+                key: "article",
+                kind: "article",
+                name: "True crime · Case",
+                body: [
+                    "Write a documentary narration about the case of {{topic}}.",
+                    "",
+                    "Length: 2200-2600 words.",
+                    "",
+                    "Tone: measured, serious and respectful of the victims and their families. Build tension from facts and timing, never from gore or speculation. No graphic detail of injuries.",
+                    "",
+                    "Shape:",
+                    "- Open with one concrete moment that sets the stakes, then step back to who the people were.",
+                    "- Tell the events in order, then the investigation: what was found, when, and what it meant.",
+                    "- Give the outcome as it stands today: charges, trial, verdict, appeals, or that it is unsolved.",
+                    "- Close by returning to the people at the centre of the case, not to the offender.",
+                    "- Six to eight sections with short, plain headings.",
+                    "",
+                    "Accuracy rules:",
+                    "- Use only facts supported by the research notes or reliable public records. If a detail is disputed or unproven, say so plainly ('police believed', 'the defence argued').",
+                    "- Name a living person who was never convicted only as the sources do, and never imply guilt.",
+                    "- Do not invent dialogue, quotes, times or places.",
+                    "",
+                    "Return only the narration in Markdown: headings and prose. No title page and no source list.",
+                ].join("\n"),
+            },
+            {
+                key: "shortScript",
+                kind: "article",
+                name: "True crime · 60-second short",
+                body: [
+                    "Subject: one striking, well-documented moment from the case of {{topic}}.",
+                    "",
+                    scriptRules,
+                    "- Respectful and factual: no gore, no speculation about guilt, no invented detail.",
+                ].join("\n"),
+            },
+            {
+                key: "image",
+                kind: "image",
+                name: "True crime · Scene",
+                body: [
+                    "A moody documentary-style photograph evoking the setting of {{topic}}: the place, the era, the objects, the weather.",
+                    "Cinematic 35mm look, overcast or night light, desaturated cool palette with deep shadows and a single practical light source.",
+                    "Empty streets, closed doors, rain on glass, paperwork on a desk; no recognisable real people, no bodies, no blood, no weapons pointed at anyone.",
+                    noText,
+                ].join(" "),
+            },
+            {
+                key: "thumbnail",
+                kind: "thumbnail",
+                name: "True crime · Thumbnail",
+                body: [
+                    "A tense cinematic image that sums up the case of {{topic}}: one strong object or place in sharp focus against a dark, blurred background.",
+                    "High contrast, cold blue and amber light, heavy vignette, readable at phone size.",
+                    "No real people's faces, no gore.",
+                    noText,
+                ].join(" "),
+            },
+            {
+                key: "description",
+                kind: "description",
+                name: "True crime · Description",
+                body: [
+                    "Write a YouTube description for a documentary about the case of {{topic}}.",
+                    "",
+                    "- Summary: 2-3 factual sentences on what happened and what the video covers. No sensational words.",
+                    "- Chapters: split where the story moves from events to investigation to outcome; 3-6 plain words each.",
+                    "- Hashtags: 3-5 about the case and true crime, most specific last.",
+                    "- Tags: 15-25 search terms for the case, the place, the people named in the sources and true crime documentaries.",
+                    "- End the summary with one line noting that the video discusses real crimes.",
+                ].join("\n"),
+            },
+            {
+                key: "shorts",
+                kind: "shorts",
+                name: "True crime · Shorts",
+                body: shortsRules("Prefer turning points of the investigation and verified surprising facts; never pick a moment that dwells on violence."),
+            },
+        ],
+    },
+    {
+        id: "history",
+        name: "History",
+        summary: "Narrative history of people, places and events, told like a documentary.",
+        voice: {
+            provider: "openai-tts",
+            model: "gpt-4o-mini-tts",
+            voiceId: "fable",
+            name: "History narrator (storyteller)",
+        },
+        style: {
+            research: true,
+            expectedWords: 2200,
+            imagesPerVideo: 16,
+            imageSeconds: 12,
+            zoomPercent: 20,
+            motionStyle: "mixed",
+            captions: { fontSize: 48, position: "bottom" },
+            videoEdit: look("sepia", {
+                transition: "crossfade",
+                transitionSeconds: 1,
+                vignette: "subtle",
+                grain: "subtle",
+                atmosphere: "dust",
+                chapterCards: true,
+            }),
+        },
+        prompts: [
+            {
+                key: "article",
+                kind: "article",
+                name: "History · Documentary",
+                body: [
+                    "Write a documentary narration about {{topic}}.",
+                    "",
+                    "Length: 2000-2400 words.",
+                    "",
+                    "Tone: vivid and curious, like a good museum guide who loves the subject. Explain terms the first time they appear. Avoid modern slang and avoid judging the past by today's words without saying so.",
+                    "",
+                    "Shape:",
+                    "- Open in a specific scene: a place, a year and a person doing something.",
+                    "- Then the background a newcomer needs, in a few paragraphs.",
+                    "- The main story in order, with the causes and consequences made clear.",
+                    "- A section on how we know: the sources, what archaeology or documents show, and what historians still argue about.",
+                    "- Close with what changed because of it and what survives today.",
+                    "- Six to nine sections with short headings.",
+                    "",
+                    "Accuracy: stay with what the research notes and mainstream scholarship support. Give dates where they matter. Mark legends as legends and estimates as estimates. Never invent quotations.",
+                    "",
+                    "Return only the narration in Markdown: headings and prose. No title page and no source list.",
+                ].join("\n"),
+            },
+            {
+                key: "shortScript",
+                kind: "article",
+                name: "History · 60-second short",
+                body: [
+                    "Subject: one surprising, true story or detail about {{topic}}.",
+                    "",
+                    scriptRules,
+                    "- Put the year and the place in the first two sentences.",
+                ].join("\n"),
+            },
+            {
+                key: "image",
+                kind: "image",
+                name: "History · Scene",
+                body: [
+                    "A historically grounded illustration of {{topic}}: period-accurate clothing, architecture, tools and landscape.",
+                    "Rich painterly realism in the manner of a museum reconstruction, warm golden light, earthy ochre, umber and faded blue palette.",
+                    "Wide establishing views and mid shots of daily life; figures shown at a distance or from behind rather than as portraits of real people.",
+                    noText,
+                ].join(" "),
+            },
+            {
+                key: "thumbnail",
+                kind: "thumbnail",
+                name: "History · Thumbnail",
+                body: [
+                    "A dramatic, detailed painting that sums up {{topic}} in one moment: one clear focal subject, strong light from one side, dust or smoke in the air.",
+                    "Warm ochre and deep blue palette, high contrast, simple composition that reads at phone size.",
+                    noText,
+                ].join(" "),
+            },
+            {
+                key: "description",
+                kind: "description",
+                name: "History · Description",
+                body: [
+                    "Write a YouTube description for a history documentary about {{topic}}.",
+                    "",
+                    "- Summary: 2-3 sentences that name the period and place and say what the viewer will understand by the end.",
+                    "- Chapters: one per section of the narration, 2-6 words each, in title case.",
+                    "- Hashtags: 3-5 about the period and subject, most specific last.",
+                    "- Tags: 15-25 search terms for the event, the people, the place, the century and history documentaries.",
+                ].join("\n"),
+            },
+            {
+                key: "shorts",
+                kind: "shorts",
+                name: "History · Shorts",
+                body: shortsRules("Prefer little-known facts, turning points and vivid scenes over dates and summaries."),
+            },
+        ],
+    },
+    {
+        id: "science",
+        name: "Science explainers",
+        summary: "Clear, friendly explanations of how the world works, one idea at a time.",
+        voice: {
+            provider: "openai-tts",
+            model: "gpt-4o-mini-tts",
+            voiceId: "nova",
+            name: "Science narrator (bright)",
+        },
+        style: {
+            research: true,
+            expectedWords: 1600,
+            imagesPerVideo: 14,
+            imageSeconds: 9,
+            zoomPercent: 15,
+            motionStyle: "zoom",
+            captions: { fontSize: 50, position: "bottom" },
+            videoEdit: look("cold", { transition: "slide", transitionSeconds: 0.6 }),
+        },
+        prompts: [
+            {
+                key: "article",
+                kind: "article",
+                name: "Science · Explainer",
+                body: [
+                    "Write an explainer video narration about {{topic}} for curious viewers with no science background.",
+                    "",
+                    "Length: 1400-1800 words.",
+                    "",
+                    "Tone: friendly, clear and precise. Enthusiastic without hype. Use 'we' for shared discovery.",
+                    "",
+                    "Shape:",
+                    "- Open with an everyday observation or a puzzling question the topic answers.",
+                    "- Build the idea step by step: each section adds one concept and ends where the next begins.",
+                    "- Use one concrete analogy per hard idea, then say where the analogy breaks down.",
+                    "- Give real numbers with units and a comparison that makes them graspable.",
+                    "- A short section on what scientists still don't know or are testing now.",
+                    "- Close by returning to the opening question, answered.",
+                    "- Five to seven sections with short headings.",
+                    "",
+                    "Accuracy: follow the research notes and the current scientific consensus. Separate established results from hypotheses. Do not invent studies, figures or quotations.",
+                    "",
+                    "Return only the narration in Markdown: headings and prose. No title page and no source list.",
+                ].join("\n"),
+            },
+            {
+                key: "shortScript",
+                kind: "article",
+                name: "Science · 60-second short",
+                body: [
+                    "Subject: one mind-bending but true fact about {{topic}}, and why it is so.",
+                    "",
+                    scriptRules,
+                    "- Give one real number with its unit and a comparison that makes it graspable.",
+                ].join("\n"),
+            },
+            {
+                key: "image",
+                kind: "image",
+                name: "Science · Visual",
+                body: [
+                    "A clean, striking visual that explains an aspect of {{topic}}.",
+                    "Modern 3D-rendered or macro-photography look, crisp detail, dark background with vivid cyan, magenta and amber accents, soft rim light.",
+                    "One clear subject, scientifically plausible shapes and scales, generous negative space.",
+                    noText,
+                ].join(" "),
+            },
+            {
+                key: "thumbnail",
+                kind: "thumbnail",
+                name: "Science · Thumbnail",
+                body: [
+                    "A bold, eye-catching image about {{topic}}: one surprising subject, glowing against a deep dark background.",
+                    "Saturated cyan and amber, strong contrast, simple composition that reads at phone size.",
+                    noText,
+                ].join(" "),
+            },
+            {
+                key: "description",
+                kind: "description",
+                name: "Science · Description",
+                body: [
+                    "Write a YouTube description for a science explainer about {{topic}}.",
+                    "",
+                    "- Summary: 2-3 sentences that pose the question and promise a clear answer, without exaggeration.",
+                    "- Chapters: one per step of the explanation, 2-6 words each.",
+                    "- Hashtags: 3-5 about the field and the topic, most specific last.",
+                    "- Tags: 15-25 search terms, from the broad field to the specific phenomenon.",
+                ].join("\n"),
+            },
+            {
+                key: "shorts",
+                kind: "shorts",
+                name: "Science · Shorts",
+                body: shortsRules("Prefer the most surprising facts and the clearest 'aha' explanations."),
+            },
+        ],
+    },
+];
+// What "Make a 60-second short" installs when no pack is picked.
+export const starterSet = {
+    id: "starter",
+    name: "Starter",
+    summary: "What a first short uses when no pack is picked.",
+    voice: {
+        provider: "openai-tts",
+        model: "gpt-4o-mini-tts",
+        voiceId: "alloy",
+        name: "Starter narrator",
+    },
+    style: {
+        research: false,
+        expectedWords: 150,
+        imagesPerVideo: 4,
+        imageSeconds: 15,
+        zoomPercent: 15,
+        motionStyle: "zoom",
+        captions: { fontSize: 48, position: "bottom" },
+        videoEdit: defaultVideoEdit,
+    },
+    prompts: [
+        {
+            key: "shortScript",
+            kind: "article",
+            name: "Starter · 60-second short",
+            body: ["Subject: {{topic}}.", "", scriptRules].join("\n"),
+        },
+        {
+            key: "image",
+            kind: "image",
+            name: "Starter · Vertical scene",
+            body: [
+                "A cinematic vertical image about {{topic}}.",
+                "Rich colour, dramatic light, one clear subject large in the frame so it reads on a phone.",
+                "Keep the lower half calm: captions are drawn over it.",
+                noText,
+            ].join(" "),
+        },
+    ],
+};
+export function packById(id) {
+    return id === starterSet.id ? starterSet : starterPacks.find((pack) => pack.id === id);
+}