OpenBMB / OpenBMB/MiniCPM-o-Demo

TTFS on full duplex

Open
#3 2 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Dominant language
Python
Stars
386
Forks
81
Avg merge
1h 59m
Merged PRs (30d)
3

Description

Hello, thanks a ton for making this. Has a lot of potential. The TTFS being like 1.5 seconds negates the whole idea of full-duplex though. Any way to improve that? I was not getting much lower than that on a 5090. Should I try on a different GPU?

Also, I noticed it responded a bit faster when I spoke while it was in the middle of speaking. Could there be a way to keep it in that mode, but have it just finish talking and wait, without triggering end-of-turn? This would avoid the TTFS start up time.

Contributor guide

No contributing guide indexed for this repository

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

No files, tests, or code entry points are identified in the issue. Start by locating the full-duplex inference and end-of-turn handling paths, then measure TTFS on the reported hardware and reproduce the speaking-while-speaking behavior. Done should include a defined latency target and verified behavior that avoids unintended end-of-turn triggering.

Written by the indexing model from the issue text.

Assessment

Tech stack
python, pytorch
Domain
ai, audio-video-rtc, performance
Issue type
Feature
Difficulty
5/5
Estimated time
Over a week
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
25/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.