Softpedia >Windows >Internet >Chat >Other Chat Tools >AI Sparks Studio >  Changelog

AI Sparks Studio Changelog

What's new in AI Sparks Studio 1.3

Feb 19, 2024
  • Improvements:
  • Added support for the new OpenAI models: gpt-3.5-turbo-0125, gpt-4-0125-preview, and the model name alias gpt-4-turbo-preview.
  • Added support for the Seed parameter in OpenAI text generation.
  • Changes:
  • Changed the gpt-3.5-turbo model name alias to automatically switch to the gpt-3.5-turbo-0125 model pricing on 02/16/2024.
  • Improved the sorting of the list of OpenAI models.
  • Switched the positions of the ‘Generate a new note’ button and the ‘Create a new note’ button to prevent accidental text generation when attempting to use the ‘Browse notes…’ button.
  • Ensured automatic scrolling to the beginning of the text of a note opened from the ‘Browse notes…’ selection screen.
  • Fixes:
  • Prevented the implicit creation and saving of a new named discussion during the removal of a named discussion.
  • Prevented the display of an exception window when a gaming device (a gamepad or joystick) gets disconnected.
  • Prevented text from overlapping their respective window areas (e.g., the Note details window) when the user interface is greatly magnified.
  • Prevented the brief, invalid display of the ‘Play this note’ button immediately after the start of text generation.
  • Prevented the display of the ‘Generate a new note’ button for an Assistant note that is the first in a discussion, aligning with the application’s limitation of not starting text generation in an empty discussion.

New in AI Sparks Studio 1.2 (Nov 24, 2023)

  • Improvements:
  • Added full support for the GPT-4 Turbo 128K (gpt-4-1106-preview) and gpt-3.5-turbo-1106 models, and limited support for the GPT-4V Turbo 128K (gpt-4-vision-preview) model.
  • Implemented a feature allowing the generation of multiple notes with a single API request (Text Generation - Number of Generated Notes), optimizing costs as the input text is processed only once per request.
  • Enhanced the user interface (UI) resizing functionality by adding standard hotkeys for resizing controls, and enabling continuous size adjustment by holding the respective button/hotkey.
  • Changes:
  • Increased the default scale of the UI for displays with resolutions wider than 1920p, improving readability right from the start.
  • Modified text generation request creation to include the max_tokens parameter only if the Text Generation - Maximum Token Count option is set above zero.
  • Fixes:
  • Resolved occasional text generation communication errors with the OpenAI service.

New in AI Sparks Studio 1.1 (Oct 12, 2023)

  • Improvements:
  • Added the ability to edit discussions. To use this feature, click on an existing discussion note and then select the new “Edit this note” option in the Actions floating panel on the right.
  • Enabled the removal of discussions via the Discussion details window. (In keeping with AI Sparks Studio’s commitment to data preservation, files are not deleted but relocated to the “Removed” folder in the application’s working directory).
  • Added the ability to effectively remove discussion notes. This new feature, called “Permanently hide this note”, can be accessed during note editing. While the note will no longer be visible in the application interface, it will be preserved within the discussion file in line with our non-destructive approach to data handling.
  • Improved ElevenLabs audio clips caching. The current voice settings, retrieved from an ElevenLabs account, are now considered when searching for a corresponding cached audio clip. Cached audio clips also now store the Style & SpeakerBoost settings used during their creation.
  • Implemented the storage of text generation input & output token counts in discussion files.
  • Changes:
  • Adjusted the default OpenAI text generation temperature from 0.7 to 0.25 to facilitate higher-quality results in expert discussions.
  • Enabled text selection and context menu for dialog windows and the About section, making it easier to copy or search for any referenced links or information.
  • Altered the “Go to the concept note” action hotkey to Ctrl+O, freeing up Ctrl+E for the newly added “Edit this note” action.
  • Fixes:
  • Removed the display of the newly introduced OpenAI GPT Instruct models in the list of available text generation models, as they are incompatible with chat.
  • Eliminated unintentional activation of auto-scrolling upon the completion of text generation.
  • Rectified issues with auto-scrolling when set to a low speed.
  • Corrected the calculation of chat completion tokens for the new ChatGPT models (0613 and above, including 16k), allowing for slightly more tokens to be used before reaching the context size limit of a model.