AI scene partners for self-tapes. Upload your sides, pick your role, and rehearse with natural-sounding readers who wait for your cue.
INT. MY APARTMENT — THE NIGHT BEFORE A SELF-TAPE
I was auditioning for Dan Murray in Best Medicine. My sides had two scenes and four other speaking parts: Amy, Martin, Mark and Geneva.
A self-tape needs someone feeding you lines on time. I didn't have two actors free to read with me, and booking a reader costs money and a calendar slot. Reading the other parts off my phone killed the rhythm, and the comedy in these scenes lives in the rhythm.
I wanted to rehearse and record the way I would in a room: hear the other characters, react, and come in on my cue.
CONTINUOUS — WHAT I BUILT
I wrote a Claude skill that turns a script into a rehearsal room. It runs in four steps, in this order.
Attach the sides PDF. Claude finds every speaking part and asks which character you're playing.
Each of your lines gets its cue, the private thought behind it, how to deliver it, and what to do with your scene partners. The whole performance hangs on one anchor.
Each reader gets an ElevenLabs voice picked for the character, and every line is recorded with its own direction.
Say “Action” and the prompter counts down, plays each reader, then holds on your line until you've said it. Say “Action” again to cut. Start over, mirror mode and pace controls handle the rest.
INSERT — THE DEMO
54 sec · Amy and Martin are the actual ElevenLabs recordings from my rehearsal. A stand-in voice reads Dan's lines for the demo.
CUT TO — WHAT CHANGED ALONG THE WAY
The first build fell back to the computer's built-in voice, which is no scene partner. I recorded every reader line in ElevenLabs and packed the audio inside the prompter, so playback never depends on an outside link.
Directions written as phrases like “warmly, by habit” risked being read aloud. I re-recorded those lines with one-word directions, so the readers say the dialogue and nothing else.
On camera you can't walk over and press Start. “Action” became the only voice command. It starts the take and cuts it, and no line in the script can trigger it by accident.
I played the readers through my wireless speakers. The voices filled the room instead of coming out of the laptop, and it sounded like other actors were there with me.
FADE IN — THE RESULT
“Once I connected the audio to my wireless speakers, it sounded like other people were in the room.”
EXT. THE SKILL DIRECTORY — DAY
No reader? No problem. Drop in your sides, pick your role, and rehearse with natural-sounding voices that read every other part and wait for your cue.