Skip to content

Roadmap

Fakhir Ali edited this page Oct 15, 2024 · 1 revision

A loosely made plan for OVC's development

OVC's mission is to be the goto opensource library for making voice bots.

Some core values of OVC are:

  • Usability (very easy to use)
  • Extensibility (very easy to extend)
  • Competitive (featureful)

OVC follows a simplicity first philosophy with regards to development. We believe in the bitter lesson, the final solution to voice agents is an end to end learned general approach. Models will improve over time, there should be no model specific code and porting a model should be very easy. Hence, when models improve so does OVC.

Goals

  • Test everything
  • Improve speed
  • Work towards end to end models

Things to develop

  • General Avatar implementation (featureful)
  • Filler phrases (perceived speed + quality)
  • Pre move llm request (speed)
  • HF-like Audio tokenizer class (e2e)
  • Next audio token predictor (e2e)
  • Endpointing model (speed)
  • Discord (featureful)
  • web client sdk (featureful)
  • speed tests (speed)
  • correctness tests (quality)
Clone this wiki locally