Fixes to some frequently encountered issues
[h2]More logical TTS detection[/h2]
Previously, all quotes were assumed to be spoken by the current NPC. Now, extra logic exists to detect whether it was spoken by the player, and also take the player's gender into account when determining the TTS voice. Note: All TTS extraction logic for dialogue currently only works if the story is in English. The TTS voice is not (yet) determined by AI, because of the extra token cost that it would incur every turn. However, one fun fact is that the quotation logic was coded almost entirely by ChatGPT 4o via a detailed prompt, and worked out of the box with minimal tweaking. Here is the link to the conversation in case anyone is interested: https://chatgpt.com/share/67344c9c-6364-8012-8b18-d24ac5e9e299
[h2]Improvement to trivial actions[/h2]
Added an option in Options for trivial actions: "No roll check". Select this option if you didn't like rolling on trivial actions.
For now, the default at first launch of the game remains "minor failure" because there's still logic in the game to give +1 to the roll result unless it would make it a critical success (meaning the only way to get a minor failure in this mode is to roll a critical failure).
[h2]Player action instruction suffix[/h2]
Added an option in Text Generation Options for adding some instruction to the LLM at the end of each player action. You may experiment with this to describe how you want it to generate the story (e.g. not take extra actions for you). Since it's later in the prompt, it should have greater weight than a preamble, but results aren't guaranteed to work as expected.
[h2]Player-confirmed enemy defeat and charm[/h2]
This was slightly more work than other buttons but should be working now. If the enemy would've been defeated during a turn you made, and player-confirmed defeat or charm is turned on, you will be presented with a choice for whether the enemy should be defeated or take their turn.
Previously, all quotes were assumed to be spoken by the current NPC. Now, extra logic exists to detect whether it was spoken by the player, and also take the player's gender into account when determining the TTS voice. Note: All TTS extraction logic for dialogue currently only works if the story is in English. The TTS voice is not (yet) determined by AI, because of the extra token cost that it would incur every turn. However, one fun fact is that the quotation logic was coded almost entirely by ChatGPT 4o via a detailed prompt, and worked out of the box with minimal tweaking. Here is the link to the conversation in case anyone is interested: https://chatgpt.com/share/67344c9c-6364-8012-8b18-d24ac5e9e299
[h2]Improvement to trivial actions[/h2]
Added an option in Options for trivial actions: "No roll check". Select this option if you didn't like rolling on trivial actions.
For now, the default at first launch of the game remains "minor failure" because there's still logic in the game to give +1 to the roll result unless it would make it a critical success (meaning the only way to get a minor failure in this mode is to roll a critical failure).
[h2]Player action instruction suffix[/h2]
Added an option in Text Generation Options for adding some instruction to the LLM at the end of each player action. You may experiment with this to describe how you want it to generate the story (e.g. not take extra actions for you). Since it's later in the prompt, it should have greater weight than a preamble, but results aren't guaranteed to work as expected.
[h2]Player-confirmed enemy defeat and charm[/h2]
This was slightly more work than other buttons but should be working now. If the enemy would've been defeated during a turn you made, and player-confirmed defeat or charm is turned on, you will be presented with a choice for whether the enemy should be defeated or take their turn.