What is the theoretical fastest time in an RTA speedrun?

There are a few types of “theoretical fastest times” for an RTA speedrun. This will explain the three main types using Any% as an example, since it has the most documentation and effort put into it.

1. TAS (Tool-Assisted Speedrun)

“What is the fastest time this game can physically be completed, if you could control exactly what inputs go into it down to the millisecond?” This is explored using TAS input: slowing the game down using an emulator and refining the inputs, then playing it back at full speed. The current TAS record is 1:07:52, and you can watch it here.

2. Community Sum of Best Segments (SoB)

This is a compilation of best individual-shine times by top-level runners of the game. Converted to RTA timing, this time is currently (early 2023) 1:10:45, and its improvement is tracked using a spreadsheet, which you can find here. An edited video with a similar goal that gets a time of 1:11:17 is here (The difference between the times is a result of the fact that not every fastest segment time is available on video).
While this run is technically more feasible than the TAS run since all the inputs are being performed in real time, it still is completely unrealistic for a human to achieve. This is because the longer a run is, the more difficult it is to play consistently enough to match your best segment of all time for each shine. RNG adds even more variance, and as a result, top-level runners can only expect to play around 2 minutes slower than their sum of best.

3. Realistic / Achievable Run

“What time is realistic for someone to achieve in the near future?” This is impossible to quantify with a single number, so this is where speculation comes in. Every runner and viewer of the game may have their own opinion, but only the runners with extensive experience playing Any% at a top level have the best idea of what is possible for them. These speculation times will always be changing as the feasibility and consistency of strats changes, and as different runners with different mindsets go for the world record. Currently (early 2023), most people see high 1:12 as possible, and mid 1:12 as a stretch goal.

All of these times are subject to change, as the speedrun evolves. However, the feasibility of each type of time stays the same, meaning TAS will be computer-level precise, community SOB will be technically possible but far out of reach, and the realistic time will be tough but in reach.