UE5.8: No audio Detection

[Maria_Nastase]

What do you need help with?

Question

Installation method

FAB

Issue area

Microphone / Voice Input

Impact

Major - a core flow is broken

Unreal Engine version

5.8

Convai Unreal Engine Plugin version

4.0.0-beta.24

Character ID

bae217ea-cb98-11f0-b519-42010a7be027

Project setup

Blueprint

Target platform

Unreal Editor

Where does it happen?

Custom project

Short summary

The Convai character is unable to detect voice input

What happened?

I followed one of the more recent Convai tutorials titled Real-Time AI Conversations & Facial Animation for MetaHumans to set up a Convai character in the sample FirstPerson project.

Once I finished setting everything up, I tested the voice interaction via Push To Talk, and nothing happened. I also tested the feature where you type to communicate with the Chatbot, and it did respond to that.

I checked, and I selected the proper input. Also, the Unreal Engine Editor is has allowed microphone access.

I have another project in UE5.7 with the Convai-3.6.9-hotfix-2 version, and the chatbot is responsive there, with the same microphone.

One thing I also noticed in the new version was that the Test button for the Mic is not visible in the Settings Widget, which I technically fixed by setting the Size box (where TestMicRecord_btn_1 is located) to be Visible. But now when I test any mic, I get the warnings:

ConvaiPlayerLog: Warning: StartRecording: already talking!
ConvaiPlayerLog: Warning: FinishRecording: did not start recording

On my other Convai version it returns that it recorded for a certain number of seconds.

I didn’t change the BPs much, I only added some prints to check whether the Start Talking and Finish Talking get triggered, and whether the Muted variable is indeed toggled.

Also, I do make sure that I release the T key a few seconds after I am finished talking.

Screenshots, recording, logs, or minimal repro

Push-To-Talk-And-Mic-Test-Logs.txt (25.8 KB)

Type-Message-Logs.txt (47.1 KB)

I checked the relevant documentation, resources, and similar forum posts before posting.

on

I confirm that I did not include API keys, tokens, passwords, secrets, private character data, private conversation logs, or customer data.

on

[Anmol]

Hi, thanks for the detailed report and the logs, they made this easy to pin down. Here’s what’s going on:

The two mic warnings are a red herring. In plugin v4 the microphone works differently than in 3.6.9. In 3.6.9, Push-To-Talk started/stopped a recording each time you held the key. In v4 the mic is always streaming to the server from the moment the session connects, and PTT just gates that stream. The old “Test Mic” button still uses the 3.6.9-style StartRecording/FinishRecording, which is why you see:

StartRecording: already talking!
FinishRecording: did not start recording

That button is hidden by default in v4 on purpose — it’s incompatible with the new streaming model. Making it visible is what surfaced those warnings. Please re-hide it; it isn’t related to your problem. (Also: the “talked total 4.09s” print in your Blueprint just measures how long you held the T key — it doesn’t mean audio actually reached the server.)

The real cause is a microphone device mismatch. Your logs show the live voice stream is capturing your Windows default input device, which is:

SteelSeries Sonar - Microphone (SteelSeries Sonar Virtual Audio Device)

Your actual headset mic — Microphone (Arctis Nova Pro Wireless) — only appears in the log the moment you hit “Test Mic,” never in the streaming path. In v4, the device you pick in the Convai settings widget only ever fed that (broken) Test button. The live stream started on the Windows default device at connect and never switched to your selection.

So: you speak into the Arctis, but the plugin is streaming the SteelSeries Sonar virtual mic, which is delivering silence → the server hears nothing → no response. Text works because it never touches the mic. This is also why 3.6.9 works fine on the same hardware — it opens your selected device fresh on every PTT press; v4 does not.

Fastest fix — try this first:

1. Windows → Sound settings → Input → set Arctis Nova Pro Wireless as the default input device.
2. Restart Play-In-Editor and talk. This alone will very likely fix it.

If you’d rather keep SteelSeries Sonar as default: open the SteelSeries Sonar app → Mic tab, and make sure its virtual Microphone’s input source is set to the Arctis and isn’t muted. Sonar’s virtual mic frequently ships pointed at the wrong source or silent.

We will also push the fix for underlying plugin bug with Convai (device selection in the widget doesn’t apply to the streaming path). But steps above will get you talking today. Let me know how step 1 goes.