Tags: python

Enough progress has been made on the Cherokee IMS-Toucan TTS project to start providing full sets of audio lesson material.

This audio is 100% computer generated audio. No humans were used in the making of these audio files.

I've continued working with using the IMS-Toucan to create a TTS for Cherokee. Much progress has been made and the system appears usable enough to go ahead and switch over to creating lesson audio. For those interested, I maintain a fork of the repo at Cherokee Language IMS-Toucan.

I've attached readings of The Wolf and the Crawdad, The Search Party, and The Turtle and the Rabbit from the book Cherokee Reference Grammar - Brad Montgomery-Anderson (2015) as well as readings of the Two Hunters story from the Cherokee English Dictionary.

I've been working with using the IMS-Toucan to create a TTS for Cherokee. My previous attempts with the Tacotron model kept resulting in problematic models that had a tendency to "ramble" on with "mumblings" in a not entirely predictable fashion. I maintain a fork of the repo at Cherokee Language IMS-Toucan.

The results I've obtained from the IMS-Toucan FastSpeech2 model are much more reliable.

I've attached a copy of three different TTS voices reading the Two Hunters story from the C.E.D., pp354-355. Cherokee Dictionary

Enough progress has been made on the Cherokee TTS project to provide a full set of audio only exercises.

This audio is 100% computer generated audio. No humans were used in the making of these audio files.

Your help reviewing Cherokee language audio is greatly needed.

As mentioned in a previous post [here], I've been working on getting a Cherokee TTS system operational. I now have initial output ready for a quality check review.

If you have at least some ear for the sounds of Cherokee either as a second language learner or you are a first language speaker your help would most definitely be appreciated.

Simply visit [Audio Quality Vote], login, review audio, and vote. That's it! Simple and easy! Nothing to download or install, it all happens in the browser.

I've abandoned trying to use the espeak-ng speech synthesizer. After spending many weeks working it, it still will not correctly apply stress rules (adding stress to unstressed words) and has other issues I've run into that are very problematic, such as where "d", "g", "n" sound so similar that it is easy to mishear these consonants.

Instead I've switched to using a Tacotron 2 TTS system published by Tomáš Nekvinda for speaking Cherokee. While more work definitely needs done, I think this is a good start, and generally sounds much better than espeak-ng. Samples follow.