Request allowlist access for gemini-3.1-flash-live-preview — project gen-lang-client-02591885

Gemini API
1

Hi team,

Requesting allowlist access for gemini-3.1-flash-live-preview.

Model: gemini-3.1-flash-live-preview
API surface: Gemini API (AI Studio), not Vertex
Project ID: [gen-lang-client-02591885..]
Project number: [9415414813..]
Billing: [Enabled]

Use case: Real-time voice and vision assistant application (BLOPY AI)
that watches a user’s screen or camera and provides live voice feedback
using audio+video bidirectional streaming via WebSocket. Currently
blocked — API key only has access to gemini-3.5-transcribe-live,
not the conversational Live models needed for two-way audio+video.

Thank you!

2

Hello @Erick_Gtz,

No allowlist is required for accessing gemini-3.1-flash-live-preview. If you are still unable to access this model, can you share the complete error message you are receiving and a minimal code snippet containing the payload details to help debug this issue?

3

Hi Pooja,

Thanks for the quick reply. Here’s the concrete evidence:

1. Billing is now enabled on the project (gen-lang-client-0259188512 /

project number 9415414813..).

2. Calling GET https://generativelanguage.googleapis.com/v1beta/models?key=API_KEY

does NOT return gemini-3.1-flash-live-preview in the list of available

models for this key — the only “live” model returned is

gemini-3.5-transcribe-live.

3. When attempting to connect directly via the Live API using the

@googlegoogle/genai Node SDK, the connection opens but closes immediately

with no useful detail:

const session = await ai.live.connect({

 model: "gemini-3.1-flash-live-preview",

 config: {

   responseModalities: \[Modality.AUDIO\],

   speechConfig: {

     voiceConfig: { prebuiltVoiceConfig: { voiceName: "Aoede" } }

   },

   systemInstruction: "...",

 },

 callbacks: {

   onclose: (event) => console.log("closed:", JSON.stringify(event)),

 }

});

Result: session closes within ~1 second, onclose event detail is an

empty object ({}), no error code or reason provided.

Could this be related to something other than allowlisting — e.g. API

version, region restriction, or a project-level setting that needs to be

enabled separately from billing? Any pointer on what to check next would

help a lot.

Thanks!

4

Hello @Erick_Gtz ,

  • Regarding models.list not returning the model: The API defaults to 50 results per page, and this model might be located on page 2. Try retrieving the full list with: curl "https://generativelanguage.googleapis.com/v1beta/models?pageSize=100&key=YOUR_API_KEY".

  • Regarding the Node.js SDK disconnect: In Node.js, JSON.stringify(event) outputs {} because CloseEvent properties are prototype getters. Inspect event.code and event.reason directly in onclose (e.g. console.log(event.code, event.reason)) to view the server’s actual closure message. You might want to consider using Gemini (or any publicly available LLM) to help answer such code related queries.

  • Also check that project gen-lang-client-02591... is actively linked to a Cloud Billing account in Google Cloud Console to prevent Free Tier rate limits (65k tokens/minute).

5

Found it — turns out it was an unrelated invalid API key issue on my end, not an allowlist problem. Thanks for the help pointing me to check the close event details directly!