Working With the Who's On First Script
The Who's On First Script is basically a digital version of the Abbott and Costello routine. You feed it a series of yes/no prompts and character names, and it outputs the confused dialogue. It's been around in various forms since at least the early 2000s. Most versions you'll find are Python-based, and they range from bare-bones to somewhat polished. The core mechanic is a lookup table mapped to response chains. When the program receives an input token like "yes," it checks the current character context and returns the appropriate confused reply. The script doesn't use any kind of natural language processing. It's a deterministic state machine. Input goes in, predefined output comes out based on position in the routine and which character is "speaking." I spent some time working with a custom implementation a few years back for a small interactive comedy demo. The main issue I ran into was timing. The routine relies on rapid-fire back-and-forth, and most naive implementations don't handle delays well. The default version I was using just dumped all output at once. My workaround was adding a sleep variable between each exchange and then parameterizing it so users could speed it up or slow it down depending on whether they were performing live or just reading through it. That cut my troubleshooting time from about two hours down to twenty minutes.
The script expects input in a specific format. Some versions take command line arguments, others read from a JSON file. The JSON approach is cleaner if you're doing anything beyond a quick test. I usually format my runs with a simple array of turns and a speaker tag for each line. One thing beginners miss is that the script breaks if you give it unexpected input. Feeding it "maybe" or "no" when the state machine only handles "yes" will cause it to fall through to a default response or crash entirely. I've seen people spend ten minutes debugging before realizing the input didn't match the expected tokens. The fix is just wrapping your input in a try/except block and catching the KeyError or TypeError that gets thrown, then logging a clear error instead of letting the script die silently.
Setting It Up and Running It
Most versions are on GitHub. You'll want to clone the repo, check the requirements.txt, and install dependencies. Python 3.7 or later is standard. After that, you can run it straight from the terminal. If you're looking for the Who's On 1st Script download, search GitHub for "whos on first python" or "abbott costello script." The most commonly starred repos are from around 2018 to 2022. Read the README carefully. Some are educational demos, some are production-ready enough for a live show. The difference usually comes down to whether they handle edge cases and whether the output formatting is clean enough for stdout display. A few repos include web interfaces. Those are fine if you want a browser-based version, but they add dependencies you probably don't need. A terminal-based run is faster to set up and easier to modify if you want to customize the dialogue.
Known Issues and Where It Falls Apart
These scripts aren't bulletproof. Here's what tends to go wrong: Longer routines hit buffer limits in older implementations. If you're running the full Abbott-Costello extended version, expect crashes around turn forty-five in some builds. The fix is usually updating the buffer size or upgrading to a newer fork. Encoding issues show up on Windows systems. I've had problems with non-ASCII characters breaking the output on Python 3.8 when the console wasn't set to UTF-8. The workaround is setting the environment variable PYTHONIOENCODING to utf-8 before running the script.
Maintaining custom versions is a pain. Most repos haven't been updated in three to five years. If you fork one and start making changes, you're on your own for keeping it compatible with newer Python versions. I ended up forking one repo and stripping out the optional web interface just to get it stable on Python 3.11. Took about an hour of work. If you need something more flexible than the standard script, writing your own state machine from scratch might be faster than maintaining a stale fork. The logic isn't complex. It's just a bunch of if/else branches or a dictionary mapping. Two hundred lines of code and you own it. There's no single canonical version of the Who's On 1st Script. The one you pick depends on whether you want quick setup or something you can actually modify without fighting outdated dependencies. Start with a minimal terminal version, see if it does what you need, and only add complexity if the basic routine isn't enough for your use case.
Get the Full Details
