Building Mochify: Re-imagining prompt-based image editing workflows
Hey IH,
I’ve always found standard image editing and processing workflows incredibly tedious. Jumping into heavy software or writing custom terminal scripts just to resize a batch of assets feels outdated.
I wanted a workflow where I could just drop an image and talk to it like a human. So, I built and launched Mochify.app.
It lets you upload image assets and type exactly what you need in plain English. For example:
"Give me this in AVIF, WebP, and JPG at 500x500, 1000x1000, and 1200x1200px."
How it works under the hood
I wanted this to be lightning-fast and entirely private, so I avoided the typical bloated tech stacks:
The LLM Layer: The frontend takes your natural language prompt and passes it through Mistral Small to instantly parse the intent.
The JSON Bridge: Mistral outputs a structured JSON object containing all the specific formatting and sizing flags.
The High-Performance Backend: That JSON is sent to a custom backend powered by Drogon (C++ HTTP framework) and libvips for blazing-fast image processing.
Extreme Privacy by Design
Because image assets often contain sensitive or proprietary data, Mochify is built with a zero-retention architecture.
The backend runs entirely inside read-only containers using in-RAM processing only. The moment your processed images are delivered, they vanish. There is zero disk persistence and zero data retention.
Check it out!
Mochify.app is live right now. I’d love for you to take it for a spin and throw your worst, most complex formatting prompts at it.
I'm completely open to feedback: How is the processing speed? Did the LLM correctly interpret your prompt? What edge-case image workflows should I teach Mistral to handle next?
