goggledefogger / talk-of-the-town Goto Github PK

dropdowns or placeholders that give known good scenario prompts:

• X teaches Y a lesson
• X debates Y
• X interviews Y
• X and Y perform a rap (battle) together

• change start/stop button to be clicking on the character image

repro: start a conversation and as the character is speaking, click to stop. then start a new one before the first one is finished, and both the old and the new will end up speaking, but it should stop the old one

maybe can use this: https://github.com/jeysonm82/python-google-speech-scripts/blob/master/stt_google.py

all upper case or mixed case causes the parsing code to error

time to start using json?

so the text-to-speech loading doesn't take too long

gpt-3.5-turbo-0613, gpt-3.5-turbo, gpt-4

choose which characters should be in a scene together

ideas for user input:

• which characters
• background context (movie, play, real life scenario, commercial, home/work situation, making a podcast)
• a goal/prompt for what happens in the scene
• a genre (comedy, action, drama, realistic/documentary/broadcast)

questions:

• 1:1 back and forth dialog? (yes, later maybe prompt gpt to inject more randomness into who talks more)

stretch goal:

• parallelize the text/audio so it loads the next audio and is ready by the time the first is done playing