Detect sounds in audio file using tensorflow and display as annotations in wavesurfer.js

Create a method, locally in Javascript, to detect specific words from an audio file using [login to view URL], hopefully making use of previously trained models. Make use of [login to view URL] and output is a list of times when the sound occurs that is displayed in [login to view URL] as a series of regions. [login to view URL] I will provide the sample words to be detected to the chosen freelancer. Requirements: User would be able to upload an audio file. User may be able to "train" the model by selecting examples of certain common words or sounds that need to be removed from within the uploaded file. Program would process the audio, search for and create regions on the audio file where these sounds occur and may display a confidence level for each in [login to view URL] Notes: using the available tensorflow demos, it is possible to train them to recognize these samples/transfer learning with speech command libraries. There may be better models. References: [login to view URL]   [login to view URL]