# Can You Use an AI Voice as Your Self-Tape Reader? > Yes, for many tapes, if it is set up right. What casting asks of a reader, why plain text-to-speech fights your pauses, which kind of AI reader waits for you, how to keep it from showing on tape, and when you still need a live human. - URL: https://www.offbook.com/blog/ai-voice-as-self-tape-reader - Published: 2026-10-09 - Last updated: 2026-10-09 It's 11 p.m., the tape is due at nine, and nobody is answering their phone. You have the sides, a tripod, and a laptop that can turn any text into a voice in seconds. The thought arrives on its own: could the computer just read the other part? Plenty of actors have had the same idea, and plenty have tried it with whatever text-to-speech tool was handy. Some got a usable tape. Others got a robot that talked over their best pause and sent a take that felt like two people in different rooms. The short version: yes, an AI voice can be your self-tape reader for many requests, and the UK's casting guidance already lists a pre-recorded app as an acceptable reader. But the voice matters less than the timing. A reader that waits for you to finish works; a track that plays on a fixed clock usually doesn't. Keep it off camera, quieter than you, and plain. For callbacks and scenes built on reactions, get a live human if you possibly can. ## What casting actually asks of a reader Start with the job, because it's smaller than you think. The UK self-tape guidance from the Casting Directors' Guild and Equity, reported in [Backstage UK](https://www.backstage.com/uk/magazine/article/casting-directors-equity-best-practice-self-tapes-73820/), says a reader can be in person, on a video call, or a pre-recorded app, doesn't need to be an actor, and shouldn't be an absolute requirement. Casting wants to see you in a clear scene. The reader is there to give you something to answer. Three things make a reader good on tape. The first is volume. Casting director Caroline Liem, in her [rules for the perfect self-tape](https://www.backstage.com/magazine/article/self-tape-rules-from-casting-director-65931/), puts the reader about a step from the camera and asks them to read quieter than the actor, because they're closer to the microphone. The second is tone. Acting coach Joseph Pearlman, writing on [why a reader can make or break a self-tape](https://www.backstage.com/magazine/article/reader-can-make-break-self-tape-auditions-333/), warns against two opposite mistakes: the robotic monotone, and the reader who puts on a different voice for each character. Both pull attention away from you. The third is timing. A good reader comes in when you finish and not before, and waits through the pause you chose to take. Hold on to that one. It's where most AI readers succeed or fail. ## Is it allowed? In the UK, the code above names a pre-recorded app as an acceptable reader, which is the closest the guidance comes to an answer. In the US, the SAG-AFTRA self-tape rules don't mention AI readers either way. SAG-AFTRA's 2023 TV/Theatrical agreement, summarized by [SAGindie](https://www.sagindie.org/sagaftra/contract-changes-2023/), says actors can't be required to use any specific software or equipment for a self-tape, but that protects you from being forced into a tool; it isn't a ruling on which readers are fine. In practice, the request from casting is your rulebook. What casting worries about with AI is your face and voice, not the reader's. Backstage's piece on [avoiding self-tapes that look AI-generated](https://www.backstage.com/magazine/article/college-admission-self-tape-advice-78529/) describes the concern: tools that can make realistic faces and voices tempt people to doctor the performance itself. Emerson College's Eric Weiss, [speaking to Backstage](https://today.emerson.edu/2025/04/15/ai-can-help-or-hurt-when-making-self-tapes-says-weiss-backstage/), drew the same line for college auditions: steer clear of tools that alter what you do, like autotune, while tools that help you organize the process are welcome. An off-camera reader sits on the safe side of that line. It doesn't change a frame of your performance. It just says the other character's words. Two exceptions. If the request itself asks for a live reader, or says something specific about how the scene should be read, follow it. And if the sides came with confidentiality terms, check what you're agreeing to before you paste them into any online tool, AI or not. When in doubt, an app that keeps your script on your own device is the cautious choice. ## Why plain text-to-speech fights your pauses Real conversation runs on tight timing. A study of turn-taking across ten languages by [Stivers and colleagues](https://pmc.ncbi.nlm.nih.gov/articles/PMC2705608/) found speakers generally avoid both overlap and silence, with the average gap between turns around 200 milliseconds, about one English syllable. Your ear expects that rhythm, and so does casting's. A plain text-to-speech track can't keep it. You either export the other lines as one audio file with gaps you guessed in advance, or you press play line by line with your free hand. Guess the gaps too short and the reader talks over you. Too long, and every exchange has a dead beat that reads on tape as you waiting. Here's where it bites. Look at this exchange: > **LADY BRACKNELL.** ... How old are you? > > **JACK.** Twenty-nine. > > **LADY BRACKNELL.** A very good age to be married at. I have always been of opinion that a man who desires to get married should know either everything or nothing. Which do you know? > > **JACK.** [After some hesitation.] I know nothing, Lady Bracknell. > > *Oscar Wilde, The Importance of Being Earnest, Act I* Say you're Jack. On your first take, "Twenty-nine" comes out instantly and eagerly. On the second, you try a different choice: a tiny pause before it, as if he's deciding whether to round down. A live reader hears that and waits. A fixed track doesn't know you changed anything, so Lady Bracknell's next line arrives on the clock you set for take one and lands on top of your pause. Reset the gap for take two, and take three, played fast again, has a dead beat in the middle. That's the real limit of a fixed track. You can give each gap its own length, but only in advance. The moment your timing changes between takes, which is the whole point of trying something new, the track is wrong. There's also a quieter cost. On a [Backstage panel of casting directors](https://www.backstage.com/magazine/article/sharon-bialy-julie-harkin-theo-park-sherry-thomas-casting-panel-73488/), Sharon Bialy said she prefers a live reader to pre-recorded lines, because an actor playing against a recording tends to wait for cues instead of listening. A rigid track trains exactly that habit. ## The three kinds of AI reader "AI reader" covers tools that behave very differently on tape. Sort them by one question: who controls the timing? - **A fixed track.** A text-to-speech tool reads the other lines into an audio file, with gaps you set. Free and fast, but the clock is in charge, with all the problems above. - **A fixed gap or tap to continue.** Some apps pause at each of your lines for a set time, or until you tap. Better, but the tap is a physical cue on camera, and a set gap still can't tell a quick reply from a long pause. - **A reader that listens.** These apps follow your line through the microphone and play the next cue when you finish. Offbook works this way. You control the timing, the way you would with a person. Video: [Line detection](https://www.offbook.com/learn/line-detection) (0:57) A general voice assistant, like ChatGPT in voice mode, is a fourth option people try. It's built for conversation rather than a fixed script, so it has to guess when you've finished, can drift from the exact words, and gives every character the same voice. Our post on [using ChatGPT as a scene partner](https://www.offbook.com/blog/can-chatgpt-be-your-scene-partner) covers where it helps and where it doesn't. For a tape, the third kind is the one to aim for. Our roundup of [the best self-tape apps](https://www.offbook.com/blog/best-self-tape-apps) compares the listening readers, the record-it-yourself apps, and live reader services side by side, with prices. To learn the scene before you tape it, an [AI scene partner](https://www.offbook.com/scene-partner-ai) runs it with you as often as you need. ## How to set up an AI reader so it doesn't show on tape Whatever tool you use, the setup is what casting notices. Run through this before your first take: 1. Delete the stage directions and scene headings from the reader's text, so it says only dialogue. A reader announcing "after some hesitation" ruins the take. 2. Pick one plain, neutral voice for the other character. Skip the dramatic accents and the gravelly villain; Pearlman's warning about character voices applies to software too. 3. If the reader feels rushed, slow it slightly with your tool's speed setting. A slightly slower cue gives you room to listen. 4. Put the speaker right beside the lens, at the height of a reader's face, so your eyeline goes where a person would stand. 5. Set the volume so the reader is clear but quieter than you, following Liem's rule. Casting has to hear the cue, so don't put the reader in your earbuds. 6. Record a ten-second test and watch it full screen with sound. If the reader sounds louder, closer, or more interesting than you, fix it now. Then rehearse against the reader before you roll. The first two runs are for finding the timing; the takes you send should feel like a conversation you've already had. ### If you're using a fixed track anyway Sometimes a track is all you have at midnight. Make it as forgiving as you can. Mouth your lines while you set the gaps, so each pause fits how long your line really takes. Give long speeches a little extra room, and mark where your planned pauses go. Then keep the scene a touch faster than you'd like, since a reader that comes in early is easier to hide than one you're visibly waiting for. ## When you still want a live human An AI reader is a fine choice for a lot of tapes: a short scene with a few cue lines, a commercial with one off-camera question, or a first-round tape where casting mostly wants to see you in the role. It's also the easiest way to rehearse before a human arrives. A live reader is still worth chasing when: - It's a callback, or a role you badly want. - The scene turns on the other character's reaction, like a fight, a confession, or a seduction, where you need someone who changes when you do. - The rhythm is fast and overlapping, with interruptions a machine can't time. - Casting has asked for different reads, and you want a partner who adjusts with you on take three. Even then, a friend on a video call counts as live. Our guide to [finding a self-tape reader](https://www.offbook.com/blog/how-to-find-a-self-tape-reader) sorts the options by how much time and money you have. ## Rehearse with a reader that waits for you The biggest gain from an AI reader often comes before you tape, not during. If you've run the scene twenty times against a reader that waits for you, you walk in knowing the words cold and the timing in your body, and any reader, human or not, gets your best take. That's what Offbook is built for. Upload your sides, pick your role, and realistic voices read every other character. It listens through your microphone while you say your line and plays the next cue when you finish. Choose Strict or Lenient line detection, or switch to Manual and press Enter or say "next line", so a long hesitation like Jack's stays yours. You can set the playback speed of the reader's lines and hide your own lines for a final test. Rehearse on your laptop with the full script on a big screen, and listen to your scenes on your phone anywhere. When it's time to record, it reads the other parts while you tape on your phone or camera. Our [self-tape without a reader](https://www.offbook.com/self-tape-without-reader) page shows the setup. And if you want the cue lines on paper too, our free [cue script maker](https://www.offbook.com/tools/cue-script-maker) pulls them out of your sides. ## Questions actors ask ### Will casting know I used an AI reader? They may, and that's usually fine. Casting knows actors tape alone. What hurts a tape is a reader that's louder than you, oddly timed, or performing. A quiet, well-timed synthetic voice draws less attention than a loud, eager friend. ### Is using an AI reader the same as using AI in my audition? No. The worry in casting is AI that changes your performance, like altered faces, voices, or sound. An off-camera reader saying the other character's lines doesn't touch your performance at all. ### Can I use an AI reader on my phone while I film on my phone? Some apps record video and read at the same time on one phone. Otherwise use two devices: one films, the other reads from beside the lens. With Offbook, the reader runs on a laptop or desktop browser while your phone or camera films. ### Should the AI voice match the character's gender or age? It doesn't need to. Casting knows the reader is a stand-in. A clear, plain voice that's easy to understand matters more than a match, and a heavily characterized voice can distract. ### What if the request says the reader must be live? Then use a live reader, even a friend on a video call. Requests that specify how to tape always win over general advice.