GPT-4 API and image input

API
44

Okay, let’s consider other APIs with file input, it will not be just path. Based on description by people with access to ChatGPT with image analysis enabled, it seems you can upload several images. From other post, it seems under Chat completions API…

const completion = await openai.chat.completions.create({
    messages: [
      { role: "system", content: "You are a helpful assistant." }, 
     { role: "user", content: "Here are two pictures of vegetables I took from the supermarket, which one will be good for curry?" }
],
    files: [
        fs.createReadStream("IMG0001.png"),
        fs.createReadStream("IMG0002.png")
    ],
    model: "gpt-4-vision",
  });

Of course, these are just wild guesses, for science! Can’t wait until DevDay when they unveil everything.

1 Like