AI Rehearsal Assistant

Inspiration

I am an actress. Same as many other industries, much of the performance arts work has moved online. Self tapes, online auditions, virtual rehearsals, even virtual live performances... Remote work means we have to rely on ourselves to work on memorizing lines and rehearsals. Instead of dragging my partner to run lines with me at 1 a.m., we thought of the idea of building an AI-assistant who can be the scene partner who not only is available at anytime, but can become any character you need it to be.

What it does

The AI Rehearsal Assistant allows you to upload scripts or finds published scripts for you online. It can provide synopsis of the scripts and summaries of the characters to help you break down long scripts quickly. When you are ready to rehearse, you can choose a character to play and a scene to start with, and the assistant will play the rest of the characters in the scene. The assistant can generate a voice that best represents the characters' profile based on the scripts; you can also customize those voices based on age, gender and voice type. When you start the rehearsal, the assistant reads lines with you, and you can say your lines to the application, which will be evaluated against the original script to tell you where you said wrong. When you get it right, the assistant goes to the next line. After you have got more familiar with the lines, the assistant can also turn into a hands-free scene partner to run lines with you without interrupting until you finish the scene, and then give you a breakdown of your performance.

How we built it

As non-tech background enthusiast, we received great help from Kindo and the Continue Plugin.

Upon settling on the core function we expect, we quickly drafted out a product structure, and used it as our guidance. We then carefully translated our ideas into a step-by-step building workflow and instructed Continue to finish the coding for us.

The product, in the end, is a combination of text parsing, text generation, voice-recognition, all integrated into a powerful agent that serves as a coach.

Challenges we ran into

Our first big challenge was with coding. But thanks to Continue, we were able to "speak" code now.

The voice recognition was also a major issue. We explored deploying Whisper with OpenAI API (GPT-4o). But in the end a complete run on API is a better choice.

Accomplishments that we're proud of

As two people with no coding backgrounds, we were able to build a basic prototype of our idea within 24 hours.

What we learned

We learned that

What's next for AI Rehearsal Assistant

Beyond finishing building basic function of the assistant, there are three scalable aspects that we would like to get funded to go forward.

1. Expansion of target market. Not only actors needs to rehearse - every one needs to memorize something correctly and quickly. Students to memorize the Gettysburg Address, influencers to memorize the sponsors' plug, entrepreneurs to memorize their big investor pitch, interviewees to memorize answers to get the dream job... By leveraging the AI agent built specifically for a smooth memorization experience used by the professional actors who specialize in memorizing lines, we believe that the application can be scaled to meet a much larger market's core needs.
2. Voice contributor economy. Not only users can use the assistant, they should be able to contribute to training the voices and gain a fair compensation. To open the collaboration end of the application to allow people to record voices for the platform to help make the tens of thousands of characters' voices more real, we will be able to create a much larger user base.
3. B2B subscriptions. Businesses and corporations have regular training needs for their employees. By building customized assistant solutions for corporations to help employees ramp up their professional skills and expertise knowledge, we will be able to take the product to the next level.

Built With

• continue
• openai
• python