Gemini Native Audio (Tier 1) – Response Cutoff & Unclear RPD Limits

## Description
We are using the **Gemini Native Audio API (Tier 1)** and encountering two main issues:

### 1. Response Cutoff
- Mode: **audio, video**
- After some time, both **transcription and audio response get cut off**.
- The cutoff does **not happen at the start**, only towards the end of the response.
- **Expected:** The model should return a full transcription/audio output without truncation.

❓ Is this due to a **model limitation** or related to **token usage/quota**?

### 2. Quota / RPD Limits
- Tier 1 quota appears to **expire before reaching 50 requests per day** (in practice, cutoff happens after ~10–20 hits).
- Documentation does not clearly mention **RPD limits for Tier 2 and Tier 3**.

ALSO - if we use native audio, as proactivity mode it expire even before RPD limits
❓ If we upgrade to Tier 2 or 3, will we face the **same inconsistency and early cutoff**?
Can you clarify the actual quota limits for each tier?

config:

# config
CONFIG = types.LiveConnectConfig(
response_modalities=["AUDIO"],
media_resolution=types.MediaResolution.MEDIA_RESOLUTION_LOW,
# context_window_compression=(
# # Configures compression with default parameters.
# types.ContextWindowCompressionConfig(
# trigger_tokens=28000,
# sliding_window=types.SlidingWindow(target_tokens=13774),
# )
# ),
speech_config=types.SpeechConfig(
language_code="en-US",
voice_config=types.VoiceConfig(
prebuilt_voice_config=types.PrebuiltVoiceConfig(voice_name="puck")
)
),
realtime_input_config=types.RealtimeInputConfig(
automatic_activity_detection=types.AutomaticActivityDetection(disabled=False),

activity_handling=types.ActivityHandling.NO_INTERRUPTION
),
input_audio_transcription=types.AudioTranscriptionConfig(),
output_audio_transcription=types.AudioTranscriptionConfig(),
system_instruction=SYSTEM_INSTRUCTION,


# proactivity=types.ProactivityConfig(
# proactive_audio=True
# )
)
---

## Environment
- API: Gemini Native Audio
- Tier: 1
- Mode: audio, video

---

## Steps to Reproduce
1. Send multiple audio/video inputs via Gemini Native Audio (Tier 1).
2. Observe that transcription/audio responses **truncate towards the end**.
3. Notice that quota expires **before 50 RPD** (sometimes after only 10–20 requests).

---

## Expected Behavior
- Model should complete transcription/audio responses without early cutoff.
- Quota/RPD limits should be clear and consistent, especially for production use.