#ghjssks

[srsly-messed-up-fruitloop]

Generating Danny Phantom Quotes from AI

Temperature: 0.2

What are you can be all the moves and the come on the student to the Ghost Zone, I was the way to the book.

Oh showing the ghost way to the confident catches and this is the fire to the school and this is a chain of the parents and the way to the parents and the Ghost Zone and the Ghost Zone.

What are you going to be the stand to the packet to the student of the works to the catch and the way to the power that we go to the student of the catches and the Ghost Zone is a free.

Temperature: 0.5

Working. You can't get on the Ghost Zone

Way to be that folks at the student of the screen. What are you still liked to super the way to close.

Sam, I can't do anything to him...

Temperature: 1.0

Lond. Uh? (Landsco Bedderly) Turted to my pads, help me thing? Behind. Whatever ke she'd i got a gian? Ugh, Runnys in where I think you’re ride? Dad's down.

Sam?

Which yee?

I literally saw this prompt last night. And now here I am coding a program to automatically generate quotes based on Danny’s lines in the actual show. I’m dumping these quotes on @dannyphantombot​ cause why the heck not. See the process of setting this up below the cut.

Let me know if I should make more blogs like this for other characters! 

Part 1: Getting the Datasets

I got most of the episode transcripts from here. Since I’m too lazy to teach myself webscraping, I just copied and pasted em into an Excel spreadsheet and filtered for Danny’s lines.

(I’m using LibreCalc here)

Then I copied and pasted the entire Line column to a text file.

Part 2: Playing with textgenrnn

This may get confusing for those who aren’t familiar with programming.

I’m using this library to generate quotes. I have PyCharm and used pip to install textgenrnn (PM me if you try this and run into any trouble, textgenrnn is VERY iffy when installing).

I ran a command to train a model on the text file. My cheap laptop then proceeded to work overtime.

After 7 minutes, I got those lovely quotes you saw above. The higher the temperature, the more random the quotes become.

>>> textgen.generate(temperature=0.1) What are you can be and I can be the beam to the confident walkie and the control there.

>>> textgen.generate(temperature=2) Ukms! *Nice)breat-uhdwbngoog!"" Maymetch,buster.ńalonaUd gotta add-us foortst? Stಠ­long.) Impon,- fun?

See?

Also, going through the original prompt from @reallydumbdannyphantomaus​, there’s a similar function for predicative text in this library.

Unfortunately it goes letter by letter instead of word by word, but still (I’m currently geeking out over how it picked up Ghost Zone though :0 ).

Part 3: Writing a Python Script to Automatically Post These

I won’t go into too much detail here since you could find the code on my Github. Basically, I use the Tumblr API to queue these to @dannyphantombot​.

Going forward, I’ll try to pop this on my Raspberry Pi and schedule the script to run regularly.