You just heard a line land wrong. The narrator rushed into the next sentence, a pause you wanted is gone, and one character's voice is a shade too flat for the scene. You know exactly what you want. The problem has always been getting from knowing to changed — scrubbing a timeline, finding the right segment, opening the right panel, nudging the right value.
The agentic director closes that gap. It's a conversational editor built into the Midsummerr chapter workspace, and as of today it's on by default for every producer. You describe the edit in plain language. It makes it, shows you the line it touched, and lets you take it back.
From hunting to telling
Here's the old loop. You're reviewing a chapter, you catch a problem, and you go looking for it. You expand the timeline table, scan for the line, match the wording, open the pacing control or the voice panel, guess at a value, apply, and play it back to check. Repeat for the next note. Every fix is a small hunt before it's an edit.
The new loop is one step. You tell the director:
"Add a half-second beat before Eleanor says 'I never asked for this.'"
It finds the line, applies the pause, jumps the editor to that exact segment, and rings it with a teal highlight so you can see and hear precisely what moved. The same holds for the notes you actually give during production — recast a line read, swap a character's voice, adjust or move a sound effect, fix a pronunciation. You say what you want; it does the finding.
Ready to try it yourself?
Create your first audiobook free →It shows its work
An agent that edits your chapter has to be accountable, or it's just guesswork you can't audit. So every action the director takes lands in a readable log — not raw tool output, but a plain-English record that names the speaker and quotes the line it changed. You can see the sequence of edits it made and trace each one back to the moment in the chapter it affected.
And because production is iterative, any edit can be undone conversationally. If a change isn't what you meant, you say so and it reverses it. Nothing is committed past your ability to walk it back.
Why this is different
Plenty of tools generate audio from text. The harder problem is revising a finished, multi-voice production — a chapter where speech, music, and sound effects are already timed against each other — without breaking the mix. That's what the director operates on: the real timeline of a produced chapter, with the full context of who's speaking and what's cued where. It's the same Midsummerr editor where script review, casting, music, and sound design already live in one workspace — now with a way to direct it by talking.
The result is that the distance between a production note and a finished change collapses to a sentence. You stay in the scene, listening and directing, instead of dropping into a spreadsheet to hunt.
Try it on a real chapter
The fastest way to understand it is to hear what it's editing. Listen to a full-cast production like Frankenstein or Jane Eyre — every pause, voice, and sound cue in those chapters is the kind of thing you can now adjust by asking.
Then open a chapter of your own in the editor. The director is already there, waiting for your first note. Tell it what you'd change, and watch it change.
Ready to direct your own? Start a project and bring a chapter into the editor.
