We've released a new version of the /media/enhance v1.1 API.

Changelog

• The content types supported by the Media Enhance API have been expanded to give processing additional guidance on how to tune the results. Each content.type configures the audio parameters automatically to apply processing that is most suitable to the characteristics typical of that type of recording.
  • If you don't specify a content.type the algorithm will choose an estimated best fit but for better results the type should be identified as an input parameter.
  • New processing profiles have been added based on these types:
    • studio - for recordings captured in a controlled studio-like environment. Generally, the subjects are close to the microphone with minimal background noise.
    • mobile_phone - where the recording was captured on a mobile phone. Frequently, the subject is giving priority to the camera in recording a video which is not an optimal position for the microphone and sound capture.
    • voice_recording - may also be on a mobile device, but voice is the priority and other sounds seen as a distraction.
    • conference - includes recordings where there are multiple talkers in a larger room setting
    • Note that voice-over has been renamed to voice_over.
  • Adjusting a specific audio attribute will take precedence over the setting applied from the content type chosen.
• Noise reduction has made significant improvements in performance for cases where the background noise is changing dynamically. This improves content where:
  • The environmental noise is changing over time, eg. talkers who are moving around within a location so sometimes closer or further away from a microphone or source of noise.
  • Multiple talkers with different background noise levels have been cut together. This is especially prevalent in content using video conferencing to connect multiple remote guests or presenters.
• Improved accuracy of progress reporting returned when getting the status of a job. The progress percentage is now updated more frequently to provide a better indication of how much of the input file has been processed.