On this Page

AI Nodes

Meadow provides visual scripting nodes for integrating AI capabilities directly into your experiences. You can send text and image prompts to large language models and use text-to-speech to generate audio — all from within your visual scripting graphs.

Requirements

All AI nodes require an API key from the respective provider. You are responsible for your own API usage and costs. Keep your API keys secure and be mindful of rate limits.

• ChatGPT / ChatGPT TTS: OpenAI API key
• Claude: Anthropic API key
• Gemini / Gemini TTS: Google AI API key

Chat Nodes

The chat nodes send a text prompt (and optionally an image) to an AI model and return a text response. All three chat nodes share the same interface pattern.

ChatGPT

Sends a request to OpenAI’s ChatGPT API.

Inputs:

• Message: string — The user prompt to send.
• Image (optional): Texture2D — An image to include with the prompt (vision).
• System Prompt (optional): string — Instructions that guide the AI’s behaviour.
• API Key: string — Your OpenAI API key.
• Model: ChatGPTModel — Which model to use. Options:
  • GPT-5.2, GPT-5.1, GPT-5, GPT-5 Mini, GPT-5 Nano
  • GPT-4.1, GPT-4.1 Mini, GPT-4.1 Nano
  • o4 Mini
• Max Tokens: int — Maximum length of the response (default: 1024).

Outputs:

• Success → Control flow on successful response.
• Failure → Control flow on error.
• Response: string — The AI’s text response.
• Error: string — Error message if the request failed.

Claude

Sends a request to Anthropic’s Claude API.

Inputs:

• Message: string — The user prompt to send.
• Image (optional): Texture2D — An image to include with the prompt (vision).
• System Prompt (optional): string — Instructions that guide the AI’s behaviour.
• API Key: string — Your Anthropic API key.
• Model: ClaudeModel — Which model to use. Options:
  • Claude Sonnet 4.5, Claude Opus 4.5, Claude Opus 4.1
  • Claude Sonnet 4, Claude Opus 4
  • Claude 3.5 Haiku
• Max Tokens: int — Maximum length of the response (default: 1024).

Outputs:

• Success → Control flow on successful response.
• Failure → Control flow on error.
• Response: string — The AI’s text response.
• Error: string — Error message if the request failed.

Gemini

Sends a request to Google’s Gemini API.

Inputs:

• Message: string — The user prompt to send.
• Image (optional): Texture2D — An image to include with the prompt (vision).
• System Prompt (optional): string — Instructions that guide the AI’s behaviour.
• API Key: string — Your Google AI API key.
• Model: GeminiModel — Which model to use. Options:
  • Gemini 3 Flash (Preview), Gemini 3 Pro (Preview)
  • Gemini 2.5 Flash, Gemini 2.5 Pro
  • Gemini 2.0 Flash
• Max Tokens: int — Maximum length of the response (default: 1024).

Outputs:

• Success → Control flow on successful response.
• Failure → Control flow on error.
• Response: string — The AI’s text response.
• Error: string — Error message if the request failed.

Text-to-Speech Nodes

The TTS nodes convert text into spoken audio, returning an AudioClip that you can play with an AudioSource component.

ChatGPT TTS

Generates speech using OpenAI’s TTS API.

Inputs:

• Text: string — The text to convert to speech (max 4096 characters).
• API Key: string — Your OpenAI API key.
• Model: ChatGPTTTSModel — Which TTS model to use. Options:
  • GPT-4o Mini TTS
  • TTS-1
  • TTS-1 HD
• Voice: ChatGPTTTSVoice — The voice to use. Options:
  • Alloy, Ash, Ballad, Coral, Echo, Sage, Shimmer, Verse, Marin, Cedar
• Speed: float — Playback speed (0.25 to 4.0, default: 1.0).

Outputs:

• Success → Control flow on successful response.
• Failure → Control flow on error.
• Audio Clip: AudioClip — The generated speech audio.
• Error: string — Error message if the request failed.

Gemini TTS

Generates speech using Google’s Gemini TTS API.

Inputs:

• Text: string — The text to convert to speech.
• API Key: string — Your Google AI API key.
• Voice: GeminiTTSVoice — The voice to use. Options:
  • Aoede, Charon, Fenrir, Kore, Puck, Leda, Orus, Zephyr

Outputs:

• Success → Control flow on successful response.
• Failure → Control flow on error.
• Audio Clip: AudioClip — The generated speech audio.
• Error: string — Error message if the request failed.

Playing TTS Audio

To play the audio clip returned by a TTS node:

1. Add an AudioSource component to a GameObject in your scene.
2. After the TTS node’s Success output, use a Set AudioSource Clip node to assign the audio clip.
3. Use a Play AudioSource node to play it.

Tips

• System prompts are useful for setting the AI’s personality, language, or constraining its output format. For example: “You are a friendly museum guide. Keep responses under 50 words.”
• Vision works by connecting a Texture2D to the Image input. This could be a camera capture, a rendered texture, or any image in your scene. The AI will describe or respond based on the image content.
• Smaller models are faster and cheaper. Use GPT-5 Nano, Claude 3.5 Haiku, or Gemini 2.0 Flash for quick interactions. Save the larger models for complex reasoning.
• API calls take time. Consider showing a loading indicator while waiting for responses, especially on slower connections.
• All AI nodes are found under the XREF → AI category in the visual scripting fuzzy finder.