How to optimize choppy / interrupted speech output

General
1

Hi,

I’m trying to get the example scripts to run, especially No. 19 voice active dog (standard script, no camera etc.). Every example up to no.19 works, stt, tts, cam etc.

I’m using a Raspi 5 (8GB, booting into commandline) and a local ollama (pidog connected to a WLAN Hotpspot of a M2 Ultra with running ollama, no other devices in that WIFI network. Distance ca. 2 meters).

What I’m experiencing is a rather disappointing speech output, choppy and interrupted.
Is there a way to optimize that?
(And: what is causing the delays?)

Thx

2

Example 1-18 runs normally, indicating that the hardware and audio system are functioning properly. The issue is most likely caused by the timing of the LLM response in Example 19.

To improve fluency, we recommend:

Using a smaller model (e.g., a 3B model)

Limiting the maximum response length: for example, ask it to reply within 20 words

The delay is mainly due to the LLM generation time (not WiFi), which is normal for local models.