Voices - TTS - with Vidnoz

For the OneSixthual dolls I need to assign each doll a particular voice, which fits to its character. I have experimented with several online tools, and here are my results.

Vidnoz for TTS Voices

I have used Vidnoz since 2 years, it was my first venture into AI-generated animations. At first I used their avatar mode: uploaded image of a doll, then made the doll speak by selecting a voice and added the text prompt. The result was then that the avatar spoke. 

This approach had the disadvantage that the avatars were quite unflexible, a bit stiff, and only the head moved together with the speaking. Mouth movement was OK.

So then I devised another more complicated method: Vidnoz also has a mode where audio-only can be generated, without the video. I used this to select from the voices the ones I liked, then entered the text prompt. The resulting wave file I then merged with DreamFace to an existing video snippet. 

Recording the voice

Vidnoz offers more than 1200 voices. It is quite difficult to find a suitable voice, one needs to try one-b-one. In the voice library one can filter by age:

Also one can filter by style: there are 34 different styles (e.g. newscast, whispering, etc.) and 17 scenarios (e.g. Training, Chat, Poetry, etc.), but it is still quite difficult to find a suitable voice, as these categories sometimes are not clearly describing the voice. This filtering is also available in editing mode.

And one can filter by language. It is actually great fun to select a language other than English, e.g. French, and then enter an English text - the voice then speaks English with an accent! 

And one can customize each voice: 

A great feature: one can preview the voice recording without needing to pay - this allows experimentation before the paid final version is recorded.

Using DreamFace for merging video with audio

The app DreamFace costs $30 per year, and credits can be purchased: 30 cr for $5, 1000 cr for $100. 5cr are added automatically weekly, but they expire after 1 week. I forgot how many credits a merging of the video with the audio costs... So I upload the video snippet, then the audio wave file, and I let the app do its magic. 

There are a few issues:

• DreamFace does not support NSFW videos, and it refuses the upload of a video which does contain such content. Therefore, one must only upload just a video close-up of the face that i supposed to talk.
• Multiple faces in a video are not supported. There must only be one clearly visible face.
• When the audio is longer than the video, the video at its and is added playing backwards. This ensures a smooth overall video, but it makes the video unusable in a context where its ending is pre-determined. Therefore, the video snippet should always be longer than the audio snippet.

Dreamface also does have its own voices, and they appear to have a great selection. But I have not used them yet, as I like the choices offered by Vidnoz.

Creating the final video

I then paste / merge the resulting video+audio into my overall video timeline, overwriting the previously non-speaking part with this new speaking part. For this I am currently using Microsoft ClipChamp, in the free version. Works fine for me.