RecipeMemoir Product Spec
Private product specification. Enter hub password to continue.
That's not right. Try again.
RecipeMemoir Product Spec
The recipe manager that makes your collection beautiful, useful, and permanently yours.
1. Vision & Positioning
The wedgeRecipeMemoir turns a private recipe collection into something worth keeping forever and beautiful enough to share. Import everything you've ever clipped or typed, organize it effortlessly, cook from it daily, and, with one click, publish it as a gorgeous personal recipe website that's fully yours. Grandma's recipes become a family cookbook site, not a ZIP file.
Two convictions anchor the product:
- Anti-lock-in: your data, always yours. Paprika's import is effortless and its export is a clunky proprietary file. RecipeMemoir takes the opposite stance: clean import from everywhere (Paprika first), clean export to everywhere (PDF, JSON-LD, Markdown, HTML), and the publishing wedge itself is the ultimate portability statement.
- The memoir / family-heritage emotional core. A recipe collection is a life's record. No mainstream recipe manager publishes a beautiful public personal website from your collection (the research confirms this row is empty across all 16 profiled apps). That whitespace is the wedge.
Against Paprika specifically: Paprika is a beloved, offline-first filing cabinet in maintenance mode: best-in-class clipper, no web app, no AI, no collaboration, a planner and grocery list that don't talk to each other, painful export, dated UI, single-developer pace. RecipeMemoir matches the filing cabinet (table stakes) and then does the things Paprika structurally can't: web-first, beautiful publishing, AI capture, connected planning, family cookbooks.
2. Target Users
4 personas- Primary: "The Family Archivist" (the wedge buyer). 45 to 75, mostly women, a large collection accumulated across Paprika, recipe boxes, and bookmarks. Motivated by legacy and sharing, not meal-prep optimization. Pays for beauty and permanence. Annette is the reference user: 1,741 real recipes, dogfooding from day one.
- Primary-adjacent: "The Frustrated Paprika Power User." 500+ recipes in Paprika, loves the clipper, resents the planner/grocery disconnect, the missing web app, the per-platform repurchase, and the export lock-in. Wins on painless one-click import, a real web app, planner-to-list flow, and the export story.
- Secondary: "The Everyday Home Cook / Meal Planner." 30 to 55, plans the week, shops from a list, cooks with a phone on the counter. Needs the table-stakes set to be excellent.
- Secondary: "The Viral-Recipe Saver." Saves from TikTok, Reels, YouTube. The most crowded frontier; we fast-follow and win on extraction accuracy.
Explicit non-target for v1: monetizing food creators who want paywalls and subscriber management. Our wedge is home cooks, not creator commerce.
3. Competitive Frame
Table stakes · frontier · whitespaceTable stakes (must have to be credible): URL clipping, cross-platform incl. a real web app, meal-planning calendar, auto shopping list, recipe scaling, cook mode, search and organization, sync.
Frontier (only 1 or 2 apps do each well): social/video AI import, photo/handwriting OCR, AI meal planning, pantry/store intelligence, voice cook mode, family sharing with annotations, rich portable export.
Whitespace (nobody does it): publishing a gorgeous public personal recipe website from your own collection. Recipe.Site is the only adjacency and it's pointed at creator monetization. This is RecipeMemoir's locked wedge, genuinely uncontested as of mid-2026.
| Competitive lesson | Implication for RecipeMemoir |
|---|---|
| Paprika's #1 complaint: planner and grocery list don't talk | Planner-to-list integration is table stakes, day one of the planner |
| Best-designed apps (Mela, Crouton, Pestle) are Apple-only with no web | A genuinely gorgeous web app is itself a differentiator |
| Social/video import is the crowded frontier | Fast-follow; win on accuracy, never lead the roadmap with it |
| Umami/RecipeSage own portability today | Match their export breadth, beat them with the published-website output |
| Pricing ranges $1/mo to $59.99/yr | Room for a fair freemium ladder |
4. Product Pillars
6 pillars- Beautiful by default. Photo-forward, warm, cheerful, bright (Saveur / ATK / Sally's Baking warmth, not cool-editorial minimalism). Every surface themeable via tokens.
- Yours forever. One-click import, one-click export, one-click publish. No lock-in, ever.
- The memoir. Recipes carry stories: annotations, family sharing, version history ("Mom's tweak"), inheritance. The emotional moat. Its strongest expression is a real book in the family's hands: the print-ready cookbook export (KDP) extends "beautiful by default" and "yours forever" from the web all the way to a printed, sellable volume.
- Actually useful in the kitchen. Planner that fills the grocery list, scaling that works, cook mode that survives a glance at a text, offline when the wifi drops.
- AI where it serves, invisible where it doesn't. Capture from photo/video/URL, photo generation, smart planning. AI is a means, never the headline.
- Honest publishing. The copyright-safe pipeline means users can publish clipped collections without legal exposure, something no competitor has thought about.
The memoir chain (the moat sentence). The whitespace features below aren't a grab-bag; they chain into one story. Family contribution links let relatives drop in recipes with no account, feeding handwriting scans and recorded voice stories into the collection, which then flow onto the family website and into the printed KDP book, where a QR code beside each recipe plays the voice story aloud. A photographed index card becomes a living page becomes a sellable book that talks back in Grandma's voice. No competitor has any single link in that chain, let alone the whole chain.
5. Feature Spec
VersionedVersion key: v1 personal web product + publishing wedge · v2 multi-user product + mobile · v3 later.
Table stakes
| Feature | What "good" means | Version |
|---|---|---|
| Paprika import | One-click upload → parsed, categorized, enriched. Built and proven on 1,741 real recipes | v1 exists |
| Web clipper | Parse any recipe page (JSON-LD first, LLM fallback) at Paprika-or-better quality | v2 |
| Browser extension | Chrome/Firefox/Safari one-click save | v2 |
| Organization | 6-facet tagging, ratings, favorites, notes. Exists in v0 | v1 exists |
| Search | Full-text (FTS5) + faceted filtering | v1 bones |
| Recipe scaling | ×½ ×1 ×2 ×3 with fraction/unicode/decimal parsing | v1 |
| Cook mode | Full-screen step-at-a-time, wake lock, in-step timers, state survives navigation | v1/v2 |
| Meal planner | Drag-recipes-onto-days calendar, reusable menus | v2 |
| Planner → grocery list | Planned meals flow ingredients into the list automatically. Paprika's #1 complaint | v2 |
| Grocery lists | Aisle grouping, consolidation, multiple lists, household sharing | v2 |
| Sync | Cloud-backed by default (D1 source of truth), conflict-aware, visible backup/restore | v2 |
| Offline | Read and cook offline on mobile; web gets cached-collection PWA support | v2/v3 |
| Print / PDF | Beautiful print stylesheet (exists) + PDF export | v1 |
Differentiators
| Feature | Spec | Version |
|---|---|---|
| Website publishing (THE wedge) | One click: collection → gorgeous public site at you.recipememoir.com (custom domain on a paid tier). Theme picker via the upload-to-unlock style gallery. Static-fast, SEO-clean, fully exportable | v1→v2 |
| Portability / export | PDF, JSON-LD, Markdown, HTML, full-collection archive. Import FROM other apps as fast-follows | v1/v2 |
| Print-ready cookbook export (KDP) | One click: any collection → a professionally typeset, print-ready cookbook. Interior PDF at standard trim sizes with margins/bleed, automatic TOC and index, photo layouts, plus a cover generator with spine-width calculation from page count. Targets Amazon KDP print-on-demand first; the same PDFs work for Lulu/local printing. Book templates share the site's token-based theme system, so the book matches the family's website. No mainstream competitor does collection-to-printed-book; it widens the publishing whitespace. Strict copyright: a for-sale KDP book is commercial public distribution, so any recipe with a source URL must run the full rewrite-spec pipeline first | v2 (v1 seed: Annette's family cookbook, dogfood pilot) |
| AI photo engine | Exists & approved: flux-1-schnell + reference-grounded plating prompts + Real-ESRGAN upscale → R2 | v1 exists |
| AI capture: photo | Snap a recipe card / cookbook page / handwritten card → structured recipe | v2 |
| AI capture: URL+ | Anything the clipper can't parse goes to an LLM extraction fallback so import never dead-ends | v2 |
| AI capture: video | TikTok / Reels / YouTube → structured recipe. Ship when we win on accuracy | v3 |
| Family cookbooks | Shared collections with annotations, version history, permissions, invitations | v3 |
| Voice cook mode | Hands-free "what's next?" navigation + timers | v3 |
| AI meal planning | "Plan my week from what I have / my goals" → plan + consolidated list | v3 |
| Nutrition | Per-serving nutrition already enriched in v0. Macro filtering later | v1/v3 |
| Themeable everything | User-selectable color themes AND app icon from day one. Token-based, zero baked-in colors | v1 |
Explicitly later (don't build before the wedge validates): full pantry inventory, grocery-store integrations, appliance integrations, community discovery feed, creator monetization, native desktop apps, dietary-restriction-first marketed mode, and cook-along across kitchens (v3, ambitious tier: a synced live cooking session between remote family members with shared step position, shared timers, and optional video).
Memoir & Publishing Whitespace
The uncontested laneThese features all live in the same uncontested territory as the website-publishing wedge and the KDP book: they treat a recipe collection as a living family record, not a database. None is well served by any mainstream competitor, and several chain together into the moat above.
| Feature | Spec | Version | Why it's whitespace |
|---|---|---|---|
| Handwritten recipe rescue | Photograph handwritten recipe cards; AI transcribes to a structured recipe, but the original scan of the handwriting stays displayed beside the typed version on the site and in the KDP book | v2 | Apps OCR the card and discard the image; nobody keeps the handwriting as the keepsake |
| Voice stories on recipes | Record an audio story per recipe; it plays on the recipe's web page, and the printed book gets a QR code beside that recipe linking to the audio | v2 | A print-to-digital bridge for family audio that no recipe app has |
| Holiday mode | Select N recipes for an event → a merged prep timeline working backward from serve time, oven/burner conflict scheduling, a consolidated shopping list, and a day-of checklist | v2 | Zero multi-recipe orchestration in the market; every app plans one dish at a time |
| Family contribution links | No-account links relatives open to submit a recipe (typed, a photo of a card, or a voice note) into the family collection, pending the owner's approval | v2 | Collaborative collection-building without forcing relatives to install an app |
| Adapt-without-losing-the-dish | AI dietary adaptation that preserves the dish's identity, using on-theme substitutions only (never bolt a foreign protein onto a dish to hit a macro) | v2/v3 | Everyone filters by diet; nobody respectfully adapts your recipe while keeping it the dish you love |
| Cook from what I have (MY collection) | Fridge/pantry input matches against the user's own collection, not generic internet recipes | v2 | Personal-collection grounding, versus Samsung Food and others suggesting generic web recipes |
| Real measurement intelligence | Cups↔grams with per-ingredient density, metric/imperial, and pan-size conversion math | v1/v2 | Broadly useful and cheap; chronically half-done everywhere |
| Gift formats | Printed 4×6 recipe-card decks, family recipe calendars, single-recipe art prints (handwriting included). Extends the KDP pipeline; same rights/rewrite rules for sourced recipes | v3 | Natural premium revenue; no competitor turns a collection into giftable physical objects |
| Collection health | Duplicate/variant detection on import and on demand, with a side-by-side merge that preserves provenance | v1.5/v2 | Import-adjacent; grounded in a real bug we hit on Annette's own import today (slug dedupe collisions). Every Paprika migrant arrives with duplicates |
Also far-tier: cook-along across kitchens (v3) lives in the explicitly-later list above with the rest of the ambitious work.
6. Platform Plan
Web → iOS → AndroidLocked priority: Web first, iOS second, Android third. Design decisions favor web-first workflows; the mobile apps follow the web model.
Web (v1): the existing Cloudflare stack carries forward (Pages + Functions, D1, R2, Workers AI). The /site/ bones become Annette's personal site, then the template for every published site. Published sites stay static-fast vanilla; the multi-user app shell likely adopts a small framework (Svelte or Preact-class) in v2.
Mobile (v2): the tech choice is a decided item (see Decisions below). Capacitor-wrapped web app first with a native-feel quality bar, stepping up to React Native via Expo if it misses, native Swift/Kotlin only if it earns it at scale.
7. Architecture Direction
Continuation, not a rewriteEverything grows from the running v0 stack:
- Hosting/compute: Cloudflare Pages + Functions, Workers for background jobs (photo queues, import processing, publish builds), Queues/scheduled Workers for long imports.
- Data: D1 (SQLite) remains system of record. R2 for photos (bucket exists) and export archives.
- AI: free-tier Workers AI (flux-1-schnell for photos) + cheap OpenRouter routing (DeepSeek for rewrites/extraction, Llama 3.3 70B for bulk classify and the verify gate). Measured unit economics: ~$0.03 per 344-recipe LLM pass, ~$1 per 1,700 photos. A structural cost advantage; no premium models in batch paths.
Multi-tenant evolution: add users and collections tables; every recipe gets a collection_id (Annette's 1,741 become collection #1). All queries become collection-scoped. Published sites get a sites table and a publish pipeline that snapshots a collection through the copyright gate into static output. Keep a clean seam for later per-tenant or sharded D1.
8. Content & Copyright Pipeline
A quiet moatThe pipeline applies on publish to any recipe with a source URL (clipped): titles kept verbatim (with a site-wide non-affiliation disclaimer and a no-logos rule), directions rewritten by DeepSeek and checked by a cross-model verify gate (Llama judge, Grok escalation, deterministic number-check, human spot-check), headnotes dropped, domain-only attribution, ingredients/nutrition kept as facts, and photos replaced via the words-not-pixels engine. User-authored recipes pass through untouched. Annette's private v1 site is exempt (verbatim is fine for personal use). A short IP-lawyer sanity check rides along pre-commercial-launch.
The same gate governs the print-ready cookbook export (KDP), and more strictly. Selling a book on Amazon is commercial public distribution, a higher-exposure act than a free web page, so any recipe with a source URL must run the full pipeline (rewritten directions + verify gate + AI-generated or owned photos + logged rights attestation) before it can be exported for sale. A personal, not-for-sale print can be looser; the for-sale KDP path is not.
This pipeline is quietly a moat: it's what makes "publish your clipped collection" (as a website or a sellable book) legally shippable at all, and no competitor has it.
9. Monetization
Income projectMonetization is committed (income project, 2026-06-11). The pricing model is now decided: freemium anchored on publishing.
- Free: unlimited private recipes (recipe caps are the most-hated pattern in the category), full import/export (portability is never a paywall), core organizing/search/cook mode, a published site on a
*.recipememoir.comsubdomain with a starter theme set. - Premium (~$29 to $39/yr, with a lifetime option as a Paprika-refugee bridge): custom domain, full premium theme gallery, family cookbook seats, generous AI quotas, planner+ features.
- Print-ready cookbook export (KDP) is a natural premium feature: a typeset, sellable book is exactly the high-value "publish" moment payment attaches to. Pricing shape is a sub-option under the decided freemium model (not a new open decision): bundle it into Premium, or charge a per-book export fee so non-subscribers can buy a single book. A launch-time test decides which (likely both: included for subscribers, per-book for everyone else). The strict KDP copyright pipeline runs before any for-sale export.
- Principle: never paywall getting your data out. Plain export (PDF/JSON-LD/Markdown/HTML) stays free forever; it's the brand. The premium charge is for the typeset book, not for portability.
Decided 2026-06-11
Annette took the recommendations- ✓ 1. Mobile tech stack Decided
Capacitor-wrapped web app first with a native-feel quality bar (cold start, offline cooking, share-sheet import), stepping up to React Native via Expo if it misses the bar. Native Swift/Kotlin only if the app earns it at scale. - ✓ 2. Auth provider Decided
Hosted auth (Clerk-class) for v2 speed, abstracted behind one swappable module. - ✓ 3. Mobile offline/sync engine Decided
Roll-own SQLite oplog sync in its simplest honest form (last-write-wins per field, item-level merge for grocery lists), re-checked against Replicache/PowerSync at v2 build time. - ✓ 4. Pricing model Decided
Freemium with premium publishing tiers at ~$29 to $39/yr plus a lifetime option. Getting your data out is never paywalled (export stays free forever). - ✓ 5. Launch posture Decided
Dogfood (Annette's own published site is the first output) → waitlist with founding members (lifetime pricing) → public launch aimed at Paprika migrators, seeded by her newsletters. - ✓ 6. Success scorecard Decided
100 published sites + 25 paying users in the first 90 days of multi-user launch; revenue targets get set once conversion data exists.
Still Open
1 decision for Annette- ? 7. Your personal site's style
Still open by your explicit call ("don't assume it's editorial"); the bones stay token-based until you pick from the gallery/mockups.
Product spec drafted 2026-06-11 · Decisions 1-6 recorded 2026-06-11 (only personal-site style remains open) · Private · ← Hub