search

erew123 / alltalk_tts

AllTalk is based on the Coqui TTS engine, similar to the Coqui_tts extension for Text generation webUI, however supports a variety of advanced features, such as a settings page, low VRAM support, DeepSpeed, narrator, model finetuning, custom models, wav file maintenance. It can also be used with 3rd Party software via JSON calls.
GNU Affero General Public License v3.0
1.17k stars 123 forks source link

Support for MARS models #285

Closed lofcz closed 4 months ago

lofcz commented 4 months ago

Is your feature request related to a problem? Please describe. No.

Describe the solution you'd like To my knowledge, MARS5 and the upcoming MARS6 are the sota English TTS models with open weights available - https://github.com/Camb-ai/MARS5-TTS Given that AllTalk TTS already supports multiple models, I'd like to know whether the maintainer would be interested in adding support for MARS in v2.

Describe alternatives you've considered Do nothing :/

erew123 commented 4 months ago

Hi @lofcz

Pretty much any TTS engine can be added in and there is a new engine template folder with instructions in the code base https://github.com/erew123/alltalk_tts/tree/alltalkbeta/system/tts_engines/template-tts-engine

And throughout all the template files, they are documented with instruction e.g. https://github.com/erew123/alltalk_tts/blob/alltalkbeta/system/tts_engines/template-tts-engine/model_engine.py

The only considerations are that something installed by one engine could over-lap/over-write other engine's requirements, but thats not an issue if you DONT want to use both engines (should there be a conflict).

The instructions for the TTS engine you list one seem pretty easy and are documented on the main page of it, so it should be easy to add in Id think.

I can take a look at it some time, or youre welcome to have a go yourself.

Thanks

lofcz commented 4 months ago

Thanks for the quick reply. Adding MARS is something we want to do, I've checked the instructions and they seem clear to me, however, I have no knowledge of the codebase, and implementation on our side would probably be something hacky, to just get it working for our use case. Hence the inquiry on whether we should go ahead, or if this is something you would be open to adding.

erew123 commented 4 months ago

@lofcz Sure, no problems. I have a few TTS engines to add in at some point, so ill take a look soon.