Microphone was quite mature to have global audio response in the early 20th century. The reasons that those recordings sound broken are 1) at that time the dynamic and directional microphone is not good enough or unavailable yet, 2) the supporting techniques to help environmental effect was not that good 3) the storage media was not good enough to keep all responses, and can be deteriorated because it was not digital.
Amis is the first ever effective AI model to fix the recordings by improving microphone’s state.
What Amis can do:
- Amis keeps all human voice in detail to have all its original emotional sounds.
- Amis also fixes most human-voice clippings and distortions.
- Amis retains all human acoustic fingerprint.
- Amis accepts all kinds of languages because Amis does not use the context at all.
What Amis does not do:
- Amis does not give synthetic robotic sounds because Amis is not an inpainting technique, and it won’t generate non-existent speaking words nor non-existent melodies.
- Most other AI speech enhance technique is to predict its speaking context (what speaker actually says) and then synthesise some-what-human-can-accept sounds.
Amis can be used with Rokuon to improve human voice recording such as radio podcast. As far as I know Amis is the only one effective technique that can also be used in remastering music and singing (vocal).
Using Amis is easy. First, prepare the audio file you want to analyse. This can be done on the PC by QuickTime, Audacity (free mixer software, https://www.audacityteam.org/) or VLC (free media player https://www.videolan.org/vlc/).
Check this out to upload video file directly from iPhone without a PC.
Second, fill out the form and check the agreement. Third, drag the file to the form or select the file.
At last, the analysed results with download links will be sent back by the email. Noted, analysed results will be adjusted to the volume close to the submitted audio, and the download links will be expired after 24 hours.