What it may mean
The program source may be objectively louder during mixed moments, or voice dynamics may be too inconsistent for the current balance.
What it does not prove
Separate meters or a quiet menu do not prove the balance during the loudest representative scene.
Evidence to inspect
Record three controlled segments: voice alone, program audio alone, then both together. Compare the same sentence on headphones and phone speakers.
One conservative free test
Lower only the program source by a small documented amount, keep the microphone unchanged, and repeat the same mixed segment before adding compression or a limiter.
Rollback first
Record the original program gain value. Restore it if intelligibility does not improve or the program becomes unacceptably quiet.
Help us test this symptom against the finished output.
There is no public download or paid checkout today. The initial research scope targets Windows 10/11 x64 and OBS 28+.
Join the research cohort →