cyle / stills

Utils for stills bots

What is this?

A lil' util to help you create a stills generator.

Setup

This project is sort of a weird command-line library.

Create your own project and make this a dependency, i.e. npm init && npm install https://github.com/shahkashani/stills.git

For inspiration and example setup, check out the fieriframes bot.

In your own project, you'll need three folders:

• videos/ - where the source material will live. Can basically be any video format.
• stills/ - where all the stills will end up
• gifs/ - where all the gifs will end up

Below is a list of commands. You'll need ./node_modules/.bin in your path, or just prepend that to every command.

Generating stills? Here's all you need

I usually run these wherever I keep the videos, i.e. my personal computer.

stills

Create stills from videos folder into stills folder. Run stills --help for more info.

Detecting and training faces

If you ran stills --help, you'll note there's a --faces options. This will, out of the box, work for faces! If all you care about is any face, then just stop here.

However, if you're looking for specific faces, you're gonna have to train it to detect them. Thankfully, that's super easy.

1. In your project, create a faces folder.
2. Make a subfolder and name it to match the person you're training, e.g. faces/fieri. Put a bunch of images in here of that person. It does not have to be close-ups, but make sure your person is the only one in the image.
3. Run train --name=fieri. This will create a file called faces/fieri.json.
4. That's it. Next time you run stills --faces=10, it'll make sure 10% of images contain this fieri person.

Adding captions

1. Put a bunch of SRT files into a folder called captions
2. Run stills --captions=10 or whatever percentage you want
3. For 10% of images, this will pick a random caption from one of your SRT files. Bonkers.

gifs

Same as stills, but creates GIFs in a gifs folder. Run gifs --help for more info

upload

Upload whatever is in ./gifs and ./stills to Dropbox.

Posting stills? You'll wanna read this

Setup

If you're posting stills, you'll need the following things in either an .env file or as environment variables:

DROPBOX_ACCESS_TOKEN=Access token for the Dropbox account the stills live in
DROPBOX_FOLDER=Name of the folder the stills live in, e.g. /stills
TUMBLR_CONSUMER_KEY=Tungle app consumer key
TUMBLR_CONSUMER_SECRET=Tungle app consumer secret
TUMBLR_ACCESS_TOKEN_KEY=Tungle user access token, create this from the API console
TUMBLR_ACCESS_TOKEN_SECRET=Tungle user access token
TUMBLR_BLOG_NAME=Tungle blog name, don't include .tumblr.com
TWITTER_CONSUMER_KEY=Twitter app consumer key
TWITTER_CONSUMER_SECRET=Twitter app consumer secret
TWITTER_ACCESS_TOKEN_KEY=Twitter access token key for the account that'll be doing the tweeting
TWITTER_ACCESS_TOKEN_SECRET=Twitter access token secret for ditto

(You don't have to provide both Tumblr and Twitter keys, skip whichever ones you don't need.)

You're also gonna need to do the following in Dropbox as a one-time step:

1. A folder where the stills live, e.g. /stills (specify the name in the DROPBOX_FOLDER field in .env)
2. An empty JSON file called <name>.json, e.g. /stills.json, where all the posted images will be registered. It'll live at the same level as the folder. Start by just uploading a file with the contents [].

OK, now onto the commands. I run these on a cron on Heroku.

connect

Check the connection with Twitter and Dropbox. Good for validating the .env file.

post

Connects to Dropbox and posts a still, and updates the JSON file with the image name and the live URL(s).