Taigi SST Voice

Currently, data volume is limited and training resources are insufficient, so there are some accuracy challenges.

To improve the accuracy of the SST system, we are currently seeking contributors for data collection and annotation.

Go to the contributor application form

After uploading a Taigi audio file, the AI will perform transcription (SST) and provide 5 suggested results. The displayed text can also be played back using AI voice synthesis.

You can also try the recording version directly.

Upload audio file

WAV / MP3, up to 3 MB

FAQ

Is the recognition accuracy poor?
Currently, there is not enough training data. If the audio contains noise or is of poor quality, the recognition accuracy may be reduced.<br>Please try to upload clear audio recordings (with minimal noise).<br>The better the audio quality, the better the recognition results.<br>Currently, audio sources such as TV programs or YouTube videos, which are relatively high quality, tend to produce more stable recognition results.<br>
What is the difference between the recording version and the upload version?
The recording version transmits audio directly from your browser's microphone for recognition. The upload version is useful for batch processing of existing audio files (such as WAV, MP3).
What is SST (Speech-to-Text)?
SST (Speech-to-Text) is an AI technology that automatically converts audio data into text.<br>This website’s Taigi SST system recognizes Taigi speech and provides 5 transcription candidates.<br>The transcribed text can also be played back using AI voice synthesis.<br>It can be used for Taigi learning, subtitle production, and document creation.<br>We are continuously working to improve recognition accuracy.<br>
Will Romanization (Tailo) be displayed?
The current model does not yet support displaying Romanization (Tailo).<br>The recognition results currently only display Taigi characters.<br>A new model is under development, and Romanization display will be supported in the future.<br>