Panda Greetings
Custom Greeting Cards
I’ve always been bad at writing messages for birthdays, holidays, etc. I’d either overthink it or end up writing something super generic, so I started building PandaGreetings.com
The idea is simple:
you describe who the card is for + the vibe, and it generates both the message and the image. Then you can edit everything and turn it into a physical card or eCard.
I worked on this off and on for a couple years, but over the last ~6 months I rebuilt it using AI tools.
Current stack/process:
- GPT → helps generate structured prompts + message content
- Image models → generate the card visuals (moved from DALL·E → gpt-image models)
- Claude → used for writing/refactoring code
- Xcode/tools → debugging + fixing issues
What ended up working best:
- breaking features into small tasks instead of huge prompts
- iterating quickly instead of trying to design everything upfront
- focusing on prompt structure to improve output quality
Right now the product flow is:
input → generate message + image → edit → send (physical or digital)
It’s live and I’m still actively improving it.
What I’m trying to figure out next:
- how to make the outputs more consistent stylistically
- which features actually matter vs. what I can cut
- how to get initial users/traction
Would love feedback from anyone here:
- does this feel like a real problem worth solving?
- anything obvious I should change or simplify?
4 Comments
This is such a specific, real pain — the blank card moment is brutal. I built EzWrite for the same root problem but for everyday messages (work, apologies, awkward texts) rather than cards specifically. Did you find people needed just one good version, or did options/tones matter for something as personal as a card message? Curious because that's been the debate on my end too.
It is a common frustration when you find the perfect card but then spend twenty minutes staring at the blank inside because you don't want to sound like a generic store-bought message.
The real opportunity here is using structured prompts to capture inside jokes or specific shared memories because that is the one thing AI can turn into a personalized sentiment that a standard card aisle completely lacks.
Are you looking into a subscription model for "power users" who send cards for every single holiday and birthday throughout the year to help stabilize that monthly revenue?
I actually have added something like that. Signed in users can save contacts with personalized info, and the system will try to incorporate that into the message depending on the context. It’s not perfect though. I’m still experimenting with how structured that should be.
Yeah, I’ve started testing a subscription for heavier users. It lets them generate more, lowers the cost per card, and adds a couple of free ecards each month. Still trying to figure out pricing.
Incorporating saved contact info is a smart move because the jump from a generic wish to a personalized reference makes the difference between a card that gets kept and one that gets tossed.
Testing a subscription for heavy users helps build a predictable revenue stream especially if you include a "reminder service" that suggests cards for upcoming birthdays in their contact list.
I use this same strategy of personalization in high-tier PR and media placement where we tailor every pitch to a specific journalist to build massive brand authority on top news outlets.
Are you finding that users prefer the AI to generate a complete message from scratch or do they mostly want a few starting points they can then edit themselves?