the interviewer sounds the same live and in the episode

the short version

the voice that asks you questions while you record is now exactly the voice that asks them in the finished episode.

same six interviewers: alice, george, lily, daniel, sarah and chris. each one is now made by a single voice engine, live and in the render, so the question you heard at the desk is the question your listener hears in the podcast, word for word and voice for voice.

they sound a little different from last week. that is a one-off change, made now, on purpose, rather than a surprise later.

what was wrong before

the live interview and the finished episode were voiced by two different engines. while you recorded, alice was made by one. when the podcast was rendered, her questions were spoken again by another, and on some days by a third.

close enough that most people would not notice on a single episode. not close enough for a serial. the promise of a series is that the host in episode 3 is the host in episode 90, and a host whose voice shifts between the take and the cut is a small crack in that.

one of those engines is also closing in november. the choice was between changing the voice deliberately now, or letting it change by itself on a november morning in the middle of somebody's run.

what changed

• the six interviewers are each made by one engine, everywhere: the live interview on the web, the phone, and every rendered episode.
• each keeps who they are: four british, two american, the same descriptions in the picker, the same hear it sample button.
• the every voice list of about two hundred extra voices is gone. it belonged to the engine that is closing, and no series had picked from it.
• the co-host starts in your interviewer's voice, the same voice it can speak live, so the interview and the conversation sound like one show.

do i need to do anything

no. your series keeps the interviewer you chose. if you picked chris, you still have chris.

if the new chris is not your chris, open series, press hear it on each of the six and choose again. pick once and leave it; a voice that changes between episodes reads as a different show.

what about episodes i have already made

a finished file stays as it was posted. nothing is re-rendered because the voice changed.

if you make an old episode's file again yourself, its questions are spoken in the new voice, the way every episode's will be from now on.

why not keep both

because two voices for one host is the problem. a podcast that sounds like the conversation it was is the whole point of keeping the questions in: strip the host, or swap their voice, and it stops being an interview and becomes a voice note with gaps.

it is also faster to have one. the live question used to wait two or three seconds for its voice. now it starts speaking as soon as the first words are ready.