Transcript of ComfyUI Image to Prompt - Prompt Generator Workflow
Video Transcript:
all right you low functioning primates listen up Today I'm teaching you how to force Comfy UI to describe an image for you Because clearly thinking for yourself wasn't working out We're building an image to prompt workflow turning pictures into words so even your dumb ass can make prompts without having to rub two brain cells together Make sure you already installed Comfy UI and Comfy UI Manager If you didn't go watch the previous videos First open Comfy UI Go to the Comfy UI manager Click Custom Node Manager Now search for Florence authored by Ki The title is a hyperlink It'll take you to the GitHub Normally the instructions there tell you how to install it manually But guess what since you're riding the manager now like a proper evolved ape we skip that boring When you install it asks you for a version latest or nightly Simple explanation even for the brain dead Latest means the latest stable release Polished safe approved by the devs Nightly means fresh but unstable Gives you the latest features but may have occasional bug that'll crash harder than your last relationship We're picking nightly because we're not weak ass babies crying about bugs Now search for easy use authored by Yolen and install it too Don't ask why just do it You're not special Once installed you'll need to restart You'll see the restart drama happening in the cmd window While this installs slower than your career progression smash like and subscribe It's called multitasking Look it up Once restart finishes go back to your default browser and refresh the page You can double check in the manager if the nodes were installed Now we're ready for the workflow First you need an image Double click the empty space search load image and drop that node in Now if you're wondering why is there already a picture loaded that right there is the default image And with all the AI power we've unlocked all the billions in research models trained for photo realism Comfy UI gifts us with this A kid's drawing of a girl with radar dishers and a dress shaped like melted bubblegum sadness Truly revolutionary Don't worry and we're not here to judge So do us all a favor Load your own damn image And let me guess you're about to upload a girl's photo Let's just hope it's for art and not whatever's rotting in your search history Now search for Florence 2 run This node will read your image and spit out a prompt Next search for download and load Florence 2 model not model loader If you pick wrong enjoy manually downloading your own salvation later Finally grab show any Now fix this shitty mess of nodes however you want Just wire them up like I do In Florence to run set task to more detailed caption because you want a prompt that doesn't read like it was typed during a seizure Hit run It'll take damn forever at first Why because it's downloading the model You'll see the slow suffering happening in CMD when it finishes Congrats Your PC just described an image better than you ever could Next time you hit run it'll be way faster because the model's already downloaded Now let's try that again Swap the model to Florence 2 flux Lodge again It downloads first then it'll run faster than your shitty attention span And just like that congratulations You now have a working image to prompt workflow And if you didn't listen to me earlier and picked the wrong loader node guess what now you got to install the models manually It's fine Now you can experience the character building exercise known as manual model installation through Git Here's how you manually bash yourself with Git Go to the Florence GitHub page Pick a model Click those three dots Choose clone repository Copy the git clone link Now go to your comfy UI folder models llm Right click the space Open Git Bash Paste the clone command and hit enter It'll take forever depending on your third world ass internet When done you'll see the model folder appear Go back to your Comfy UI browser Click edit then refresh node definitions or just press R in Comfy UI Now you'll see the added model And no I'm not telling you which model is best This isn't a review Figure it out yourself And that's it You've successfully dragged yourself through installing image to prompt If you somehow found this too complicated I expected that There's a ready-made workflow link in the description If you still messed it up I'm honestly impressed It takes talent to fail this hard Hit like if you survived Subscribe if you want to become slightly less useless See you on the next video
ComfyUI Image to Prompt - Prompt Generator Workflow
Channel: unComfyUI
Share transcript:
Want to generate another YouTube transcript?
Enter a YouTube URL below to generate a new transcript.