Project

General

Profile

Actions

Feature #273

open

W3: Send audio buffer to backend

Added by Malek chrifi alaoui about 1 month ago. Updated 15 days ago.

Status:
Resolved
Priority:
Normal
Start date:
11/20/2025
Due date:
11/27/2025 (28 days late)
% Done:

0%

Estimated time:
2:30 h

Description

Implement logic in Unreal Engine to send the recorded microphone buffer to a backend server using HTTP or WebSocket.
Ensure proper data formatting and handle network errors.
This step is essential for triggering the STT pipeline.

Actions #1

Updated by Malek chrifi alaoui about 1 month ago

  • Status changed from Re-opened to New
Actions #2

Updated by Malek chrifi alaoui 24 days ago

  • Status changed from New to In Progress
Actions #3

Updated by Malek chrifi alaoui 23 days ago

  • Status changed from In Progress to Resolved
Actions #4

Updated by Malek chrifi alaoui 15 days ago

"Send audio buffer to backend" – Unreal Engine 5

For this feature, I implemented the microphone-recording system inside Unreal Engine 5.
The goal was to reliably capture microphone audio and export it so the backend can process it later.

  1. Microphone Recording System

I created a dedicated Blueprint called BP_AudioManager which is responsible for handling microphone input.

Start Recording

Uses an AudioCapture component.

Calls StartRecordingOutput to begin capturing audio.

Stores the recorded data in a SoundWave variable.

Activates the microphone and starts listening.

Stop Recording

Calls FinishRecordingOutput, which finalizes the recording.

Unreal automatically exports the recording to a .wav file inside:
Saved/AudioCaptures/

The file path is stored in a Blueprint variable so it can be accessed by other systems.

This ensures that every recording session produces a usable audio file.

  1. Push-to-Talk Input Logic

To control recording, I added a Push-to-Talk input action using the Enhanced Input System:

First press → Start recording

Second press → Stop recording

A Blueprint boolean (bIsMicRecording) toggles between the two states.
This creates a simple and reliable record/stop workflow.

  1. What Is Completed for This Ticket

The following tasks are fully implemented:

Microphone audio is captured inside UE5

Audio is exported to disk as a WAV file

Recording can be triggered manually through a Push-to-Talk input action

The system is stable and ready to be connected to the backend in the next step

Actions

Also available in: Atom PDF