[RFC]: Does the official team of vllm-omni plan to support the integration of the index-tts model?

[ericzw-ai]

Motivation.

At present, vllm-omni has not yet integrated with the index-tts2 model. Community users have clearly expressed their expectations for this integration, so this RFC is hereby submitted to clarify the timeline for the integration.

Proposed Change.

Integrate the index-tts2 model into the vllm-omni framework, enabling users to call text-to-speech (TTS) capabilities natively within vllm-omni, and achieve seamless collaboration with existing large language model (LLM) inference workflows.

Expected Feature Scope

Support batch text input for parallel TTS inference.
Provide configurable parameters (e.g., speech speed, pitch, speaker voice type) to meet customized needs.
Maintain consistent inference performance with the standalone index-tts2 model (low latency, high throughput).
Add corresponding usage examples and API documentation in the vllm-omni official documentation.

Feedback Period.

No response

CC List.

No response

Any Other Things.

No response

Before submitting a new issue...

• Make sure you already searched for relevant issues, and asked the chatbot living at the bottom right corner of the documentation page, which can answer lots of frequently asked questions.

[hsliuustc0106]

yes, you can find some RFC for tts models. Some community members are working on it. #334 #271 #498

[hsliuustc0106]

if you are interested in contributing to tts models, feel free to contact me in slack sig-omni channel

[divyanshsinghvi]

Hi @ericzw-ai The problem with these models is that they don't follow the standard config.json path that is generally expected and too many modules that are currently not implemented in other standard repos. Once my PR #498 is approved, I will try to update my PR #334 along similar structure.