Generate fine waifu art
Go to file
2024-04-24 19:10:38 +02:00
runpod Better service/credential validation 2024-04-24 19:10:38 +02:00
s3 Better service/credential validation 2024-04-24 19:10:38 +02:00
.gitignore Working image generation, small refactor 2024-04-18 00:14:28 +02:00
config.go Add Katakana and improve bannedword matching speed 2024-04-23 17:17:27 +02:00
display.html Add Katakana and improve bannedword matching speed 2024-04-23 17:17:27 +02:00
fooocus.go Fix social previews and remove default style 2024-04-20 01:58:39 +02:00
gallery.go Add Katakana and improve bannedword matching speed 2024-04-23 17:17:27 +02:00
go.mod Add Katakana and improve bannedword matching speed 2024-04-23 17:17:27 +02:00
go.sum Add Katakana and improve bannedword matching speed 2024-04-23 17:17:27 +02:00
http.go Better service/credential validation 2024-04-24 19:10:38 +02:00
loading.html Separate generation endpoint from gallery, interactive loading 2024-04-21 17:46:19 +02:00
main.go Better service/credential validation 2024-04-24 19:10:38 +02:00
README.md Separate generation endpoint from gallery, interactive loading 2024-04-21 17:46:19 +02:00
submit.go Separate generation endpoint from gallery, interactive loading 2024-04-21 17:46:19 +02:00

Waifu Gallery

Uses Fooocus API running on Runpod serverless to generate '''Beautiful''' artwork with the base constraints of your choosing and display it in a virtual gallery in an ironic twist.

In practice, going to painting_a_futurist_atwork.yourdomain.com will generate an image with {{ random Prompt }},{{ base Prompt }}, (painting a futurist artwork:UserBias) if it hasn't already been created and save it to s3 to speed up further viewings.

Installation

$ go get
$ go build

Usage

Add a config.json configuration file in the your current directory and then execute the program.

Example:

{
}

TODO

  • make imggen api request
  • upload result to S3
  • Poll request
  • Guard from redundant API calls
  • Guard from excessive number of img generation jobs
  • Separate img generation endpoint and call it from HTML
  • Make user wait for result interactively and reload
  • Convert results to WebP
  • Better querying via LLMs
  • Better front-end