What So Ordered Is, and How It's Made
An introduction to So Ordered: why it exists, how each episode is produced, and an up-front, complete disclosure that every voice on the podcast is AI-generated.
About this trailer
This is the introductory trailer for So Ordered. It explains what the podcast is, how each episode is made, and what to expect before the regular opinion readings begin.
What the podcast does
The Supreme Court releases dozens of written opinions every term, and those opinions shape American law. Oral arguments are recorded and freely available, but the Court does not release audio of the justices reading their final opinions. If you want to take in an opinion, you have to read it.
So Ordered closes that accessibility gap by producing audio readings of each majority opinion across the term. The reading is the Court’s actual published opinion, lightly adapted for spoken delivery (for example, numerical volume-and-page citations are simplified or omitted because they do not translate well to audio). The substance is the Court’s own words, unabridged.
How episodes are made
Each opinion is read in a synthesized voice lightly stylized to resemble the authoring justice. A short clip of that justice’s voice, taken from a publicly available oral argument, is fed into an open-source AI voice model, which then generates the reading. You may hear something that carries a hint of the justice who wrote the opinion — but it is not their voice and not a recording of them.
AI disclosure: Every voice on this podcast, including the introduction, is generated by artificial intelligence. No human narrates the intro and no human reads the opinion. Tone, inflection, pacing, and emphasis are all produced by the model from the written text alone and should not be attributed to any justice. If a passage is ever unclear, consult the original opinion at supremecourt.gov.
Practical notes
This is a personal, free, non-commercial project. There is no advertising, no paywall, no political angle. Episodes are not abridged, so expect some that run over an hour. The model is not perfect; occasional glitches or odd pronunciations may slip through review, and the process will be refined over time.