How to use the AudioTestResource when faking out a response during testing

[nexxai]

I'm writing a test for my app which uses the audio()->speech() resource and am having trouble figuring out how to build up an instance of the AudioTestResource to do so.

Does anyone have a simple example they can share of how they're doing it?

I tried doing something like this, but obviously it's not working since it doesn't conform to the right interface:

    OpenAI::fake([
        new AudioTestResource(
            new ClientFake(
                ['audio' => 'some text to pretend is an m3'],
                ['model' => 'gpt-4o-mini-tts', 'voice' => $voice, 'text' => $text]
            )
        )
    ]);

For context, this is the method that I'm trying to test:

    public function generate(string $base_text, string $voice): string
    {
        return OpenAI::audio()->speech([
            'model' => 'gpt-4o-mini-tts',
            'response_format' => 'mp3',
            'voice' => $voice,
            'instructions' => $this->systemPrompt(),
            'input' => $base_text,
        ]);
    }

I've tried searching Google for AudioTestResource and even using ChatGPT itself, but got nowhere.

[nexxai]

Ok, I was making it way too difficult on myself.

All I did was this, and it worked fine:

    OpenAI::fake([
        'audio.speech' => 'faking some binary data'
    ]);