The Speech2Text APP can train audio files, use the tensorflow model to analyze the selected audio files, and output its category to reach speech to text.
[Operation steps and instructions]
1. Prepare dataset
The data set used by the APP is an audio file with three characters of bed, cat and happy, placed in the english_word/train folder, and select english_word in the Select Dataset.
If you want to use your own dataset, please copy the english _word folder and place it on the same level as english _word, delete all files and folders in the train folder, name the folder name with each word, and put it in wav audio files, each audio file takes about 1 second in length.
Press Train to start training.
If you need to set a different Batch Size or training times, please fill in by yourself.
The trained model is placed in the model folder.
The check of Load Weight is whether to load the weight.
If it is the first training model, or there are other training words, for example, to train 3 words into 4 words, please uncheck.
If you have already trained the model, but want to continue training, and there are no new categories in the training folder, you can choose to load the weight to shorten the training time.
There are three kinds of inferences, inferring a single audio file, inferring a folder, and inferring a microphone.
If you need to select a model for inference, please select or enter the file name in the Inference Model Path area.
Select any file, it is normal that Weight Path only shows cp-XXX.ckpt. If the user wants to input the file name by himself, please input according to this format. Do not input cp-XXX.ckpt.index or cp-XXX.ckpt.data-00000-of-00001.
(1) Inferring a single audio file
Press the icon to select the wav file you want to infer.
(2) Inference folder
Press the icon to select the wav audio file folder location to be inferred.
(3) Inference microphone
Press the pattern, turn on the microphone to record for 10 seconds, infer the content of the audio file per second within 10 seconds.
Please set the recording length in Record Length, in seconds.
Contact Us and How to Buy
Welcome to contact us. Please refer to the following link: