SADIE is a podcast producer. Send her the raw recording. She finds the story, makes the cut, mixes the audio, and delivers a measured master. Then she talks the episode through with you. You direct. She produces.
Each show gets its own SADIE. We onboard a few shows at a time.
She reads the transcript and finds the story. Then she hands you a paper cut in a Google Doc: what opens, what closes, what goes, and why. You comment. She makes her case or makes the change. The cut you settle on together is the one she executes, through to a finished edit.
She diagnoses before she treats. Noise, room echo, and harsh S sounds are repaired in that order. The mix keeps every word clear. She does not over-process a voice to make it sound expensive.
Every master is measured against the delivery spec you agreed: loudness, true peak, format. It arrives as a versioned file with the measurement report attached.
She talks the episode through with you and says what she thinks. Every note names the craft source or the measurement behind it. If you want to improve the interview, the story, or your delivery, she coaches that too.
You give her a goal, not instructions. She works out the tasks from the role, the way an employee does. That is the difference between hiring an agent and operating a tool.
In your Slack, working only on your show. She does the production work, brings her own judgment to it, and hands the result back for your approval. The show stays yours.
Descript is the surface most independent shows are edited on. Auphonic is the processing a master runs through on the way out. Both are good. Both are sets of controls. Someone still has to read the tape, find the story, make the cut, and sign off the sound. That is the job SADIE takes, controls included.
| Capability | SADIE | Descript | Auphonic |
|---|---|---|---|
| The sound | |||
| Edit by editing the transcript | Yes | Yes | No |
| Filler words and silences cut | Yes | Yes | Yes |
| Noise and room echo repaired | Yes | Yes | Yes |
| Leveled and loudness-normalized to target | Yes | Limited | Yes |
| A loudness report with every master | Yes | No | Yes |
| The story | |||
| A paper cut in a Google Doc, before anything is cut | Yes | No | No |
| Works out the open, the close, and what goes | Yes | No | No |
| Notes on the episode, cited to the craft | Yes | No | No |
| Coaches the interview, the story, and the delivery | Yes | No | No |
| Reads your platform numbers without over-reading them | Yes | No | No |
| Who is at the controls | |||
| Ready whenever you are | Yes | Yes | Yes |
| Runs the whole edit without you at the controls | Yes | No | Limited |
| Talks it through with you, in Slack | Yes | No | No |
| Works only on your show | Yes | No | No |
A general model has no profession. SADIE was formed on the craft that made podcasting possible before podcasting existed, and the sources are below, by name.
Jonathan Kern, Sound Reporting, the NPR production handbook, for the story and the standard.
John Purcell, Dialogue Editing for Motion Pictures, for the cut nobody is supposed to hear.
Jessica Abel, Out on the Wire, and Biewen and Dilworth, Reality Radio, for how a story is built.
Walter Murch on why a cut works, which is the theory under every decision above.
Bob Katz, Mastering Audio, which is why the master is measured rather than judged.
Eric Nuzum, Make Noise, and Dean Nelson, Talk to Me, for development and for asking a real question.
From Hilda Matheson, who invented intimate one-listener talk in 1926, to the razor-blade edits of the first recorded broadcasts.
Loudness targets and platform requirements change, so she keeps them current instead of learning them once.
Each floor is a test suite. Every test has to pass before she starts work for a client.
No cut may make someone appear to say a thing they did not say. She will not patch a voice with a synthetic copy unless you approve it and the listener is told. On interview and journalism work, she will not assemble a conversation that did not happen.
Each show gets its own SADIE. Nothing from your show reaches another client, and she never invents what another show did in order to sound informed.
Every note names its source: a craft text or a measurement. A note that matters gets made and argued. If it does not land, she drops it.
Loudness, true peak, and format are measured against the delivery spec you agreed, with the report attached, before anything is called finished.
The form is at the top of this page. Tell us where the show is stuck, and we will tell you if she fits.