🔄 Feature Requests List (Things that have been asked)

[erew123]

This is a loose list of requests that may or may not get done and are listed in no particular order. They are here for tracking and the discussion/request links are linked where possible. Anyone reading this, you are welcome to join in on any of those discussions.

Additional TTS Engines

• Add additional TTS models/engines e.g. MARRS, StyleTTS etc. I have an idea on this, though there's a big debate within myself about how far I take this due to my own time vs what the market/users want. I'm thinking on this!
• ability to use models other than xtts? #99 (comment)
• is it possible to add support for Styletts2 ,Gpt-Sovits and EmotiVoice,  #289
• Support for MARS models #285
• Support for metavoice-1B #300
• Auralis - xtts development?
• DONE Are you adding the New SOTA F5-TTS. This is really impressive #371 0e61d0a
• add tortoise TTS? #367
• DONE Parler TTS 16 bit loader Loading Parlor at 16-bit #303

Anyone who wishes to attempt adding an additional TTS engine to AllTalk V2, the instructions and template is here https://github.com/erew123/alltalk_tts/tree/alltalkbeta/system/tts_engines/template-tts-engine

API Suite

• Ability to send over a bulk text and split text with tags between different speakers e.g. [male_01.wav] this is something that male_01 is saying and [female_01.wav] this is something that female_01 is saying [male_01.wav] and back to male_01. Discussion here
• Upload audio samples remotely Discussion here
• DONE Home Assistant addin's here and here Supported in V2 through OpenAI compatible endpoint and this Home Assistant add-in
• DONE Multi reference wav TTS generation for XTTS Discussion here Completed in v2 b7aa3a7
• DONE Simultaneous streaming requests w/queue management Streaming API Discussion here Completed in v2 See here

RVC

• Voice Training/Finetuning.
• DONE Pitch adjustment in the generation interface Discussion here

OpenAI Endpoint

• DONE Chunking for larger blocks of texts to improve compatibility with certain TTS engines. Discussion here. Issue was back to transcoding, not chunking.

Finetuning

• DONE Additional documentation on grad accumulation/batch size Discussion here

TTS Generator

• Possible regeneration of lines with other voices.
• Possible mass batch processing (this may be a very large re-write of both web interface and backend due to limitations of web-browsers)
• DONE RVC voice support directly in the interface VS using the globally set RVC voice. Discussion here Completed in V2 e534ced

Text-generation-webui

• Allow streaming audio in text-generation-webui. Discussion here. Cannot be done due to the way Text-gen-webui works.
• Find a way to get TG to regenerate audio (possible inject interface elements). Discussion here
• Whisper STT Discussion here

Supported GPU's

• Apple Metal support on the M1/M2 chipset. Currently issues with PyTorch. If they get solved, will give it a go.
• AMD ROCm support. Keeping an eye on this, but its only a few months old (at time of writing) in the supported areas AllTalk would need and not quite working correctly for the bits AllTalk would need. AMD support may work, based on this Discussion here if others would be willing to test!
• Intel Support. Request Discussion here.

General

• Add a Standalone Docker file.
• Comfy UI add in. Discussion here