Hi everyone, I have been testing “half-cascade” gemini live model and the documentation says it internally generates text as output and make a TTS step. Then, are there any recommendations for the system prompt in order the entire audio-audio pipeline works properly?
Related topics
| Topic | Replies | Views | Activity | |
|---|---|---|---|---|
| How to bias input_audio_transcription with a prompt in the Gemini Live API? | 1 | 97 | July 10, 2025 | |
| Live API Hangs When Using System Prompt with Audio-Only Response Modality | 1 | 279 | June 19, 2025 | |
| A few prompt engineering questions | 4 | 263 | August 20, 2025 | |
| System propmpt behavior | 1 | 114 | July 14, 2025 | |
| Discussion about Gemini model usage, capabilities, and limitations | 0 | 25 | December 30, 2025 |