Adding audio to a WordPress multisite network: what is per site, what is shared

WordPress multisite is the setup you end up with when one organisation runs many sites: the local editions of a newspaper, a university's faculties, a school district, a brand's family of topic sites. One installation, shared plugins, shared user accounts, but separate posts, separate settings and often a separate editorial voice on each site.

That makes audio a slightly different problem than on a standalone site. The question is not only "how do I add a listen button", it is "which parts of this are network-wide, and which belong to each site?" Get that wrong and you either end up with one voice reading twenty different publications, or twenty disconnected accounts nobody can oversee.

Here is how the WeDispatch plugin actually behaves on a network, based on what the code does rather than on what would sound good in a brochure.

Each site connects itself

You install the plugin once, as with any plugin on a network, and you can activate it network-wide or site by site. Activation alone connects nothing. The link to a WeDispatch account is made from the WeDispatch screen of each individual site, not from Network Admin.

This is deliberate. The access key is stored in the site's own options table, and Network Admin is not a site, so there would be nowhere sensible to put it. That is also why, after a network activation, the plugin does not redirect you to a welcome screen: there is nothing to configure at network level.

For each site, the steps are:

1. open that site's dashboard and go to the WeDispatch screen;
2. click "Connect"; your browser goes to WeDispatch, which shows the host and the name of the site asking to be connected;
3. authorise it; a key is created for that site in your account's name and delivered straight into WordPress, with nothing to copy and paste.

Only the account owner or a team administrator can approve that step, because a key acts on the whole account. Any key can be revoked later from the client area without affecting the other sites. If you are new to the plugin, the WordPress page covers the single-site installation; everything below builds on it.

What stays per site

Because each site in a network has its own options and its own posts, the plugin keeps these separate:

• the access key, one per site;
• the default voice, and the voice assigned to each category, so the sports desk on one edition can sound different from the culture pages on another;
• the voiced posts, their status and their audio, stored alongside each site's content;
• background jobs such as generating audio or refreshing it after a correction, scheduled per site.

On the WeDispatch side, a connection identifies the site by its address. If your plan includes multi-site, each WordPress site becomes a secondary site in your account, with its own player template, font, speed, logo and voice, inheriting from the account for anything you leave unset. Reconnecting a site that is already known does not create a duplicate: the same address maps back to the same site.

Each site also reads in its own language. The plugin takes the language from that site's WordPress locale, so a network mixing French and English editions does not end up with a French voice reading the English site.

What is shared

Three things are common to the whole network, on purpose.

One account, one quota. There is no separate bill per site. A busy edition will consume more than a quiet one, and when the account's credit runs out, generation stops for every site at once. The other budget safeguards are covered separately.

WordPress users. On multisite, user accounts live in a network-wide table, so a user's WeDispatch preferences, such as the voice picked in their author profile, follow them from one site to the next.

The team. Roles in the WeDispatch client area cover every site in the account. You manage one group, not twenty unrelated logins. Our article on team roles for several writers goes into who can do what.

The limit to know before connecting many sites

An account can hold at most ten active API keys, and each connected WordPress site uses one. A ten-site network fits in one account. The eleventh connection is refused with a message that says so plainly (it currently reads in French: « Maximum 10 clés actives sur ce compte »), and you would need to revoke an unused key first.

We would rather state that here than have you discover it halfway through a rollout. If your network is larger, get in touch before you start, so the account structure is decided up front instead of improvised.

One more detail: if your plan does not include multi-site, sites can still connect with their own keys, but on the WeDispatch side they share the account's main player configuration. The settings made inside WordPress, such as voice per category, remain per site either way.

Uninstalling across a network

Technical teams rightly ask what is left behind. When the plugin is deleted from a network, its uninstall routine walks through every site in turn: it removes the plugin's options, post data, category voices and scheduled jobs on each one. User preferences, which live in the shared users table, are removed once for the whole network.

Your audio files are not deleted from your WeDispatch account by uninstalling the plugin, which matters if you ever move to another tool. We cover that case in switching audio providers without losing your files.

A sensible rollout

On a network, connect one site first, ideally the one that publishes most often. Set the voice, the player and the category voices there, listen to a week of articles, then connect the others one at a time, changing only what genuinely needs to differ. Inheritance from the account handles the rest, and you find any surprise on one site rather than on twenty.

Read next

• Security questionnaire for an audio tool: what to ask, and our answers, including the noes
• Breton, Basque, Corsican, Alsatian: why French text-to-speech trips on regional names
• A woman's or a man's voice for a quote: what French grammar reveals about who is speaking