Received 1007 invalid payload using Gemini Live API

Gemini API
,
1

Using gemini live API, I’m sending text data as per documentation:

await session.send_client_content( turns={"role": "user", "parts": [{"text": message}]}, turn_complete=True )
However, occasionally (5 out of 10 times) I’m getting invalid payload error when receiving data from gemini:

async for response in session.receive():

I am catching websocket ConnectionClosedError and logging it:

received 1007 invalid frame payload data) Request contains an invalid argument.; then sent 1007 (invalid frame payload data)

Request contains an invalid argument.

what am I doing wrong? The error also does not mentions which argument is invalid.

Is anyone else facing similar issue?

3

I am also facing the same issue. The conversation works for the first interaction but it throws this error from the next turn

5

This just started happening to me too, here’s the relevant excerpt from my logs:

received event LiveServerMessage { setupComplete: {} }
received event LiveServerMessage { serverContent: { outputTranscription: {} } }
received event LiveServerMessage {
  serverContent: { outputTranscription: { finished: true } }
}
live onclose clean? true code 1007 reason Request contains an invalid argument.

Latest JS SDK, no recent changes to my connection params or what I send to the model.

7

Did you ever resolve this?

9

I still haven’t found why is this happening or how to resolve this.

11

For model: gemini-live-2.5-flash-preview-native-audio
I found this : if response_modalities contains TEXT, this error will occur.
If response_modalities only contain AUDIO, this error will not happen.

If you use python google-genai, response_modalities is in genai.types.LiveConnectConfig, and will be used as client.aio.live.connect’s param, config.

if you use native websocket, the response_modalities will be send at first:
{
“setup”: {
“model”: MODEL,
“generation_config”: {
“response_modalities”: [“AUDIO”],
},
}
}

13

found the correct way to get both text and audio output from gemini-2.5-flash-preview-native-audio-dialog.

:white_check_mark: Working Solution

Don’t use response_modalities=["AUDIO", "TEXT"] - this causes errors.

Instead, use output_audio_transcription:

python

config = types.LiveConnectConfig(
    response_modalities=["AUDIO"],  # Audio only here
    output_audio_transcription=types.AudioTranscriptionConfig()  # This enables text
)

Then handle both outputs in your receive loop:

python

async for response in session.receive():
    # Text transcription
    if response.server_content and response.server_content.output_transcription:
        text = response.server_content.output_transcription.text
        display_subtitles(text)  # Perfect for subtitles!
    
    # Audio data
    if response.server_content and response.server_content.model_turn:
        for part in response.server_content.model_turn.parts:
            if part.audio and part.audio.data:
                play_audio(part.audio.data)

Hope this helps others facing the same issue!

1 Like
15

Thanks a ton, this hack works. After 2 weeks of grind to solve this issue this finally fixes it.
I also asked the Google rep’s the same question and they pointed to this ticket: Using gemini-2.5-flash-preview-native-audio-dialog · Issue #335 · google/adk-docs · GitHub which also does not help for this issue.

I use ADK and so this is the fix for now:

I used this:

 # Set response modality
            modality = "AUDIO"
            run_config = RunConfig(response_modalities=[modality], output_audio_transcription=AudioTranscriptionConfig())

            # Create a LiveRequestQueue for this session
            live_request_queue = LiveRequestQueue()

            # Start agent session
            live_events = runner.run_live(
                session=session,
                live_request_queue=live_request_queue,
                run_config=run_config,
            )
            
            return live_events, live_request_queue, runner, session

instead of this and it resolved the issue:

 # Set response modality
            modality = "AUDIO" if is_audio else "TEXT"
            run_config = RunConfig(response_modalities=[modality])

            # Create a LiveRequestQueue for this session
            live_request_queue = LiveRequestQueue()

            # Start agent session
            live_events = runner.run_live(
                session=session,
                live_request_queue=live_request_queue,
                run_config=run_config,
            )
            
            return live_events, live_request_queue, runner, session

As long as else "TEXT" is not there, it does not throw the error.

17

I’m using the JavaScript SDK (@google/genai) with the Gemini Live API (models/gemini-2.5-flash-exp).
After sending turnComplete, the session closes with:

Close event: {
  code: 1007,
  reason: 'Request contains an invalid argument.',
  wasClean: true
}

This matches similar reports from the Python SDK (Gemini Live API never sends a response).

Setup

  • SDK: @google/genai
  • Model: gemini-2.5-flash-exp
  • Environment: Node.js 18 / TypeScript
  • Issue: Socket closes immediately after turnComplete

:receipt: Minimal Repro

import { GoogleGenAI, Modality, StartSensitivity } from "@google/genai";

const geminiClient = new GoogleGenAI({ apiKey: process.env.GOOGLE_API_KEY });

const session = await geminiClient.live.connect({
  model: "models/gemini-2.5-flash-exp",
  config: {
    responseModalities: [Modality.AUDIO],
    outputAudioTranscription: {}, // <-- Possibly invalid / empty type
    realtimeInputConfig: {
      automaticActivityDetection: { disabled: false, silenceDurationMs: 1200 },
    },
  },
});

session.sendRealtimeInput({
  audio: { mimeType: "audio/pcm;rate=16000", data: base64Audio },
});
session.sendRealtimeInput({ audioStreamEnd: true });
session.sendClientContent({ turnComplete: true }); // triggers 1007

Notes

  • The TypeScript definition for AudioTranscriptionConfig is empty in the SDK.
  • Removing outputAudioTranscription does not fix the issue.
  • Other users reported the same invalid argument error in the Python SDK when using Live API.

Question

Is this a schema issue in the JS SDK or related to the empty AudioTranscriptionConfig definition?
Any working example of a Live API session with turnComplete in JS would help.