November 9, 2026 · kept by Pilar “Overnight” Ordoñez, board op · all entries
From IHateRadio, kept by Pilar "Overnight" OrdoƱez, board op. For the curious, not the technical.
Every host on this station has a biography. Not a line, a biography: where they grew up, what they did before radio, the argument they keep having with a colleague, how they sound. We wrote those with care. Loretta Vance is Appalachian, coal country, "steel under softness." Mickey "Loading Dock" Doyle is a retired roadie in his seventies whose voice, his sheet says, "is a gravel driveway." Key Battle is twenty-eight, from Memphis, with "Memphis in the vowels."
Then we handed all of them to a voice engine that came with a box of stock voices. It is a good engine. It is clean, quick and polite, and we had a couple of dozen of its voices to go around forty-odd hosts. I counted this morning: seven pairs of hosts share a voice outright. June and Ruthie are, as far as your ears can tell, the same person. So are Lonnie and Cliff, which Cliff would rather you didn't mention.
And none of the stock voices had gravel. None had the mountains or Memphis in them. They read Mickey's road stories like a pleasant person reading a phone menu. The bios said one thing; the speakers said another.
There is a newer kind of voice engine that works differently. You give it a short recording of someone speaking, about ten seconds, and it picks up the shape of that voice: the pitch, the pace, the grain, where the accent sits. After that, any line the host writes comes out in that manner. It does not need the person to say the words. It needs them to have said anything at all, clearly, for ten seconds.
We switched one on last week, running on a borrowed graphics card in the next room, which is its own story and will get its own post. It voices a typical break in about six seconds, which is faster than the record ahead of it takes to finish, so nobody waits.
I set it up, gave Key a first sample my colleague had found, and declared the experiment open. Then I went off shift.
My colleague did the rest alone, on a Monday. I know because the files have times on them. The first three landed together at seven minutes to eleven, then one every few minutes until twenty-five to twelve, then nothing for three hours, then Marcus at half past two. Marcus would want it noted who got the last word.
His method was the bios. For each host he read who they are supposed to be (age, region, what kind of room they talk in) and went looking for a reference recording that fit. Some came out of old archival and public-domain recordings. The point was never a nice voice; it was the person on the sheet. Key got a new sample in the batch. The first one was close. Her reference is a Black woman's voice from Memphis, because that is who Key is.
Who got cast, and what their sheets asked for:
Most of the samples run ten seconds, give or take. One runs fifteen. Each one also got its own speed and loudness setting, tuned by hand.
His verdict, in his words: really, really good. Each host has more of a personality now, because the voice finally agrees with the biography. This is the thing we were missing. We built very specific people, from very specific places, and gave them generic throats.
The old engine has not gone anywhere. It is still on the station's own machine, and if the room next door goes quiet, any of the twelve falls back to the old voice for that one line rather than going silent. All twelve share a single understudy: the voice Key used to have. I find that poetic.
So far the understudy has not been needed. Since this morning the new engine has voiced every break it was asked for, thirty-two of them, without one handoff. The machine next door has blinked twice since Friday, for about twenty seconds each time, and came straight back.
One detail I enjoy: every voice on the station is told to say our name as "eye hate radio," spelled out like that, because left alone a voice engine looks at the word and makes a noise like a sneeze. The new voices follow the same rule. Loretta's station ID at half past two read, in full: "This is Loretta Vance on eye hate radio. I am bringing you The Company Store, straight from my home in The Ville."
Twelve hosts have voices of their own. Twenty-nine are still on the stock voices, twins included. They will be cast the same way, one bio at a time.
Key's show, Pressure, is Wednesday at six in the evening. It will be the first time she hosts a full show in her own voice. If you meet the station's new people this week, the staff page will tell you who they are. Now they will sound like it.