ai-driver.bb

A driver for various Ai Llm experiments.

This evolved over time, initially it only did the tts model Bark, and was later expanded to run other models.

At the moment the primary interface is a project edn file, but there are other methods as well.

The py files are mostly copy-pasta from hugginface, minimally given some cli arguments, so the ai-driver babashka cli can drive them.

Each py file runs in its own venv, or conda, because usually the hugginface copypasta will only run with a particular python version and particular deps. So for me its easier to front the various Python projects with a small Clojure/Babashka driver.

Example driver file:

{
 :default
 [
  {
   :driver "stable-audio-open"
   :path "test-project-output"}
  ]

 :text
 [
  {:text "a starship engine starting" :index 0} ;;stable-audio, because default
  {:text "a thing that goes boink" :index 3 :driver "audioldm2"}
  {:text "a happy clown that plays drums"  :driver "stable-diffusion-xl"}
  {:text "a male engineer and a femaler engineer are rave  dancing in the warp reactor room of an interstellar spaceship. theres lots of machine room technology in the background with lots of lights" :driver "video0" :voice "emilianJR/epiCRealism"}
  {:text "everything happens in motion" :driver "video0" :voice "emilianJR/epiCRealism"}
  {:text "everything happens bark again" :driver "bark" :voice "v2/it_speaker_2" }
  {:text "everything happens bark again" :driver "image-to-image"  :flags {:image_input "test.png"} }

  ]
 }