TomPal: Recipe Manager App

Save recipes from anywhere - TikTok, Instagram, screenshots, cookbook photos, or text. Get clean recipe cards with ingredients, instructions, and a smart grocery list. Finally cook what you save.

Comment

Inspiration

I’ve always had the same problem: I save recipes everywhere, and then I don’t actually cook them.

A recipe might be bookmarked on TikTok, saved on Instagram, dropped into a chat, pinned on Pinterest, or screenshot into my camera roll. When I finally want to cook, I spend 10–20 minutes searching, rewatching, pausing, and trying to decode tiny captions or messy descriptions. Most of the time, I give up and cook something familiar instead.

Eitan Bernath’s brief described the exact same pain: people are flooded with recipes, but getting from “I saw it” to “it’s in my kitchen” is surprisingly hard. TomPal exists to remove that friction and make cooking feel doable again.

What it does

TomPal turns recipe inspiration into action:

  • Import from social links (TikTok / Instagram): copy a link, paste it, and TomPal generates a clean recipe card.
  • Import from screenshots: upload any recipe screenshot from your camera roll.
  • Import from cookbook photos: take a photo of a cookbook page and import it.
  • Import by pasting text: paste a written recipe from Notes or any digital notebook.

After import, TomPal generates a structured recipe (title, servings, ingredients, step-by-step method), a high-quality dish image, and estimated calories (even when the original post doesn’t include nutrition). The original source link stays attached, so you can always jump back to the creator’s video.

If you don’t want to import anything, Discovery helps you find recipes using filters (e.g., quick, healthy, low-calorie).

Finally, you can choose multiple recipes for the week and add them to Grocery List. TomPal automatically merges ingredients across all selected recipes and sums quantities, so shopping becomes one clear list instead of guesswork.

How I built it

I joined the hackathon late and had about three weeks to deliver a working MVP, so I focused on the shortest path to a real “wow” moment: fast import + clean recipe cards + a grocery list that makes cooking happen.

  • Flutter for a cross-platform codebase (I’m submitting an iOS build for the contest, with Android planned next).
  • RevenueCat for subscriptions and paywall logic (free tier + 7-day trial).
  • A link ingestion pipeline for TikTok/Instagram:
    • I use a scraper/extractor to retrieve the video caption and key metadata (title/author/source link).
    • Then I run recipe extraction with a carefully engineered prompt.
  • Recipe extraction via GPT-5 Nano:
    • The hardest part wasn’t “using AI”. It was designing a prompt that can reliably structure messy social captions, while staying fast and cost-efficient.
    • I tested multiple prompt variants and models and landed on GPT-5 Nano for speed, with strict instructions/guardrails to keep outputs consistent.
  • Design workflow:
    • I used AI (sleek.design) to generate initial UI concepts quickly, then manually refined and corrected what AI got wrong.
    • For MVP ingredient visuals, I intentionally used emoji instead of a full ingredient image database so I could ship faster.

I also used “vibe coding” tooling (Cursor, Chat GPT Codex 5.3)) to iterate quickly while keeping the app stable and demo-ready.

Challenges I ran into

  • Import reliability: Social captions are messy. Ingredients and steps are often incomplete, reordered, or mixed with marketing text and emojis. Getting consistent structured output required repeated prompt testing.
  • Speed vs. quality tradeoffs: Import needed to feel fast in a demo and still produce usable results, which forced tight constraints on prompt size and output structure.
  • Scope discipline: I had more ideas than time. The main challenge was choosing what creates the strongest “from idea to kitchen” outcome and cutting everything else.

Accomplishments that I'm proud of

  • Built a working MVP in ~3 weeks that feels like a real product, not a slideshow.
  • Delivered a clear “wow moment”: paste a social link → get a cookable recipe + grocery list.
  • Implemented multiple import modes (link, screenshot, cookbook photo, pasted text) without bloating the core UX.
  • Shipped weekly planning → merged grocery list with summed quantities, which directly bridges inspiration to real cooking.
  • Integrated RevenueCat with a clean free tier and a trial-based upgrade path.

What I learned

  • The product isn’t the extraction. The product is removing friction between discovering a recipe and cooking it.
  • AI-assisted development works best when you treat AI like a tool, not a magician: clear specs, strict constraints, rapid testing, and manual refinement.

What's next for TomPal: Recipe Manager App

After the hackathon submission, I’m continuing development regardless of the results because this solves a problem I personally face daily.

Next steps:

  • Expand link import beyond TikTok/Instagram (more platforms).
  • Improve onboarding and personalizations.
  • Add weakly meal planning.
  • Replace MVP emoji ingredient visuals with a scalable ingredient media system.
  • Improve nutrition accuracy and add dietary preferences.
  • Increase recipe import speed and overall reliability.
  • Release publicly on the App Store, then ship Android, followed by marketing and creator-focused distribution.

Built With

  • brightdata.com
  • flutter
  • openai
  • sleek.design
  • supabase
Share this project:

Updates