Hello,
I’m currently using Gemini API with the following code to generate text-only responses.
We are NOT sending any images, audio, or video files – only plain text is included in the parts parameter and expecting text results.
Here is a simplified version of our code:
const chat = await geminiClient.chats.create({
model: "gemini-2.0-flash-lite",
history: messages.map(msg => ({
role: msg.isFromUser ? 'user' : 'model',
parts: [{ text: msg.content || '' }],
})),
config: {
maxOutputTokens: 8000,
systemInstruction: systemPrompt,
},
});
The Problem
Recently, we noticed unexpected billing charges in our Google Cloud Billing report.
The charges are appearing under the SKU:
“Number of videos generated”, “Veo 3 fast Audio Generation (output)” etc
This is confusing because our code only sends text to Gemini LLM.
We are not using any video generation models or multimedia inputs in our requests.
-
Can text-only Gemini API calls ever incur the “Number of videos generated” SKU?
Our requests send onlyparts: [{ text: ... }](no image/audio/video parts). Under any documented behavior, could this still bill under a video-generation SKU? -
Could a request be categorized under a video SKU without sending video content?
-
Are there known cases where mixed/multimodal models or internal routing can label text calls as “video” in Billing?
-
The function shape is fixed (always text-only). Could a prompt alone (prompt engineering) cause the backend to perform a video generation task and thus bill as video, even without media parts?
-
