6.8.1

Improvements to Editor usage of Ollama.

6.8.0

All new ChatGPT and Gemini models have been added to Chat with AI.

6.7.5

Improvements to Editor usage of Ollama and other AI backends.

6.7.0

New! We've added Ollama support! Now you can run popular LLMs like Llama, Gemma and Qwen locally for free!

6.6.0

New! We've added Runtime DALL·E 3 usage. Now you can generate textures/images with DALL·E 3 in your shipped games.
New! Added o1-preview and o1-mini model support to ChatGPT (both Editor and Runtime usage). Note: Only Usage Tier 5 API accounts have access to o1-preview and o1-mini API models, which are customers that have 30+ days of payment history and previously spent $1000 on the API. More info: https://openai.com/index/introducing-openai-o1-preview/

6.5.0

Added ChatGPT model gpt-4o-mini for Runtime and Editor use.

6.4.1

Added ChatGPT model gpt-4o for Runtime use, additionally to the existing Editor usage.

6.4.1

Fixed ChatGPT model gpt-4o timing out in the Editor.

6.4.0

We’ve added Gemini runtime usage! Now you can use Google Gemini models in your shipped games. Check out the Gemini Runtime page for more details.
Added support for the new ChatGPT 4o model.
Added support for the new Gemini 1.5 Pro and Gemini 1.5 Flash models.
Fixed an issue where some assets could not be exported in Unity 6 Preview.
Lots of UI improvements.

6.3.0

Added Unity 6 Preview support.
Fixed layout of input fields in the Chat with AI window.

6.2.5

General UI improvements.

6.2.0

Added gpt-4-turbo-preview model to the Editor and Runtime components.
Included package manifest, now the dependencies should be imported automatically.
Improved the UI of Editor messages.

6.1.0

Fixed a problem where script generation expects an API key that has been removed.
Improved performance of syntax highlighing.
Fixed and error showing up when RemoteConfig is imported.

6.0.0

New! Text-to-speech (TTS) and Speech-to-text have been added to AI Toolbox as well as respective demo scenes for both of these services.
New! We added support for Moderation — the AI-driven profanity filter for runtime applications. Please find the included demo scene to get familiar with this feature.

5.1.0

New! We have added the latest Google AI model - Gemini Ultra. It can now be used alongside Gemini Pro, as well as the models from OpenAI and Microsoft Azure.
We have greatly simplified the set up of Google Gemini in AI Toolbox. Now it is as simple as generating and pasting an API key.

5.0.0

New! We have added Microsoft Azure as an optional backend to ChatGPT. Azure is now fully integrated with AI Toolbox and available in code generation and freeform chat. See the set up instructions in our documentation.
AI backend configuration UI impromevents.

4.8.0

Added UI to export the current conversation with AI to a text file.
General usability improvements.

4.7.0

New! We have added Google's latest generative model - Gemini. It is now fully integrated with AI Toolbox and available in code generation and freeform chat.

4.6.5

Fixed "Generate Component" button not appearing after domain reload.
Unity UI integrated buttons now reflect currently selected chatbot: ChatGPT or Bard.

4.6.0

Improved integration with Google Cloud SDK for setting up Bard.

4.5.0

Added support for the latest ChatGPT models: GPT-4 Turbo (1106-preview) and GPT-3.5 Turbo. This update greatly reduces the price of using GPT-4 and increases the size of scripts GPT can understand and generate. Read full details on the OpenAI Blog.

4.3.0

Improved integration with custom proxy ChatGPT servers.

4.2.0

Exposed ChatGPT Service methods. Now you can use ChatGPT in your own Editor scripts (additionally to the existing Runtime API).
Fixed URL suffix when using a custom endpoint.

4.1.5

Improved handling of leading and trailing text when parsing code.

4.0.3

Added controls to easily switch between ChatGPT and Google Bard independent of their configuration state.

4.0.2

Fixed a rare bug that caused the last received token to not be included in the produced script. Most often this was a closing bracket of a script or a namespace.

4.0.0

Switched the free Proxy Server API to the latest version.

3.9.5

Improved Google Bard configuration for script generation and freeform chat.

3.9.0

Switched the default ChatGPT model versions to the latest for GPT-3.5 and GPT-4.
Freeform chat UI improvements and more options in the chat settings.
Improved suggested action buttons on script generation errors.

3.8.0

Fixed and issue where DALL·E image generation window was sometimes greyed out.

3.7.5

Improved OpenAI API error handling.
Added suggested actions for the most frequent errors returned by ChatGPT API.

3.7.0

Improved Guiding Prompts for C# code generation.
General availability of gpt-4 model.
General UI improvements.

3.6.0

Updated Guiding Prompts for C# code generation, improving the quality of code generated with the latest ChatGPT model versions.

3.5.5

General availability of gpt-4 model.
General UI improvements.

3.5.0

Improved code generation syntax highlighting.
Added suggestions for best settings for editing large files.

3.4.0

We've added options to select sub-verisons of ChatGPT's GPT-3.5 and GPT-4 models. Also, we've changed the default version used in code generation to the versions that work best for Unity code generation.
General UI improvements.

3.3.0

Added an option to use the model gpt-3.5-turbo-16k that can work with much longer code. It allows 4 times more context, handling up to 20 pages of text/code in a single request.

3.2.5

Optimized the number of tokens used in chat with AI.
Freeform chat UI improvements.

3.2.1

Added the "Generate Component" button to prefabs in the Project view.

3.2.0

Added an ability to submit code files when chatting with AI. Now you can talk about any code in your project, asking for possible improvements, bug fixes, code explanations or anything else. To get you started, we've added convenient buttons for the most common use cases.

3.1.0

Added code syntax highlighting with dark and light color themes.

3.0.0

We've added experimental support for Google Bard. Now for any question or script you can use both ChatGPT and Bard! Check out Bard Setup Guide for more details.
Improved error reporting with more actionable steps.

2.9.0

Added a "Follow-up" button that allows to conveniently iterate on code with ChatGPT. You can now ask for changes to code in the same window where it is generated.
Lots of new features in the "Chat with ChatGPT" window: formatting, ability to copy or export code, default role of the AI as a Unity developer and more.
Improved shader generation: there's now a preview of the generated shader, error detection, as well as follow-up functionality.
Lots of various UI improvements.
Improvments to the Guiding Prompts.

2.8.0

Added Runtime API. This allows using ChatGPT in shipped games, not just in the Editor. You can use it for things like NPC dialogue, random character names, etc.
Added an early version of general chat with ChatGPT. Check it out in Windows > AI Toolbox > Chat with ChatGPT.

2.7.0

NEW: Create seamlessly tileable textures with DALL·E. Just toggle "Tileable" when generating a texture and you'll get an image without visible seams. This is great for textured objects (for example a wooden door), terrains (for example grass) or backgrounds (e.g. for a game menu) etc. Check out how it looks.
NEW: Edit shaders! Now you can edit any code-based shader using ChatGPT. Simply right-click a shader file and select Edit with ChatGPT > Edit Shader, describe the modifications and ChatGPT will do the rest.
Fixed DALL·E textures sometimes giving errors when saving.
Lots of various UI improvements.

2.6.0

Now you can use all OpenAI features for free even after the trial has ended! Check out the API Key settings in Project Settings > AI Toolbox.
Generate textures using Dall-E: to create a new texture go to menu "Create > Texture by DALL·E", input prompt, get image!
Use ChatGPT to create UI Documents! Go to "Create > UI Toolkit > UI Document by ChatGPT".
Create shaders using ChatGPT: Go to "Create > Shader > ChatGPT". Write prompt, get shader.

2.5.0

We're now detecting if the code generated by ChatGPT does not compile. If the code is broken, you can ask ChatGPT to fix it, revert the changes, or accept and edit manually.
Improved UI handling of very long prompts.
Renamed the asset from "ChatGPT Script Generator".

2.4.0

Improved the Guiding Prompts. If you added this asset to your project before this update, please reset the Guiding Prompts in the Project Settings for better code generation results.
Lots of UI and quality of life improvements: keyboard shortcuts, context menu items on components, accept/reject changes and many more!

2.3.0

Re-tuned the default guiding prompts to use separate components less frequently.
Handling code markers (```) in ChatGPT's output.
Added "Save to History" toggle on the prompt window.
Fixed per-query temperature slider in the prompt window.
Added temperature of each query to the history.
Added the script generation/editing target to history.
Added per-query timeout option.
Improved the API status messages on the Demo scene.
Storing separate prompt cache for adding and editing modes.
Checking validity of the generated code.
Added "How to write prompts" button in Settings.

AI Toolbox with ChatGPT, DALL·E, Whisper, Gemini etc