# Add and remove

After setup, you edit one file, `sources.yml`. Each entry in it is one episode. Each commit starts a run that makes the episodes match the list: it dictates new entries, removes deleted ones, and dictates again any entry whose MP3 is missing.


# Add something

Edit `sources.yml` in the GitHub web editor (the pencil icon works on a phone too), add a line, and commit:

``` yaml
- https://example.com/an-article-i-meant-to-read
- files/attention-is-all-you-need.pdf
- source: https://example.com/a-long-read
  title: A better title
  voice: bf_emma
```

- A **URL** is fetched and dictated.
- A **document** (PDF, DOCX, PPTX, EPUB, HTML, Markdown) must be in the repo. Upload it into `files/` with **Add file → Upload files**. Then list it by its path, such as `files/paper.pdf`.
- An entry can also be a `source:` with its own `title`, `author`, `date`, `voice`, `speed` or `profile`. See the [`sources.yml` reference](../reference/configuration.md#sources.yml).

A run takes about half as long as the audio it makes, so a 25-minute article takes about 12 minutes. The **Actions** tab shows its progress. When it is done, pull to refresh in your podcast app.


# Remove something

Delete the entry's line in `sources.yml` and commit. The next run removes the episode, its MP3 and its text.

Deleting only the MP3 does not remove an episode. While the entry is still in `sources.yml`, the next run dictates it again.


# Dictate something again

Delete its MP3 in `audio/` and commit. The next run dictates it again with the current settings. Do this after you change the voice, the speed or a `[replace]` word, because a change to a setting only applies to new episodes.

In the web editor, open `audio/`, click the MP3, then **⋯ → Delete file**.


# Change the settings

The first run writes `earmark.toml` into the repo. Edit it in the web editor to change the defaults for new episodes: the voice, the speed, the feed title, or the cover image. [Settings](../settings/index.md) lists every setting you can change, in `earmark.toml` or on one entry in `sources.yml`.


# Limits

- **Everything in a public repo is public.** This includes every file in `files/`. Only upload what you have the right to share. GitHub Pages on a private repo needs a paid plan.
- **One file cannot be larger than 100 MB.** At the default bitrate, that is about 3.4 hours of audio. Split a longer book, or lower `bitrate` in `earmark.toml`.
- **A Pages site stops at about 1 GB**, which is about 30 hours of audio. The run summary warns you after 800 MB. Delete old entries from `sources.yml` to stay below the limit.