[Feature/Model Request]: Support for NPU-accelerated Text-to-Speech (TTS) models
Nobody has claimed this yet.
- Dominant language
- C++
- Stars
- 1.9k
- Forks
- 152
- Avg merge
- 4h 14m
- Merged PRs (30d)
- 11
Description
Suggestion Description
I am requesting support for lightweight Text-to-Speech models (e.g., Kokoro-82M or Piper) compiled for the NPU, alongside an exposed OpenAI-compatible /v1/audio/speech endpoint.
I already use the fastflowLM whisper model for speech to text in openwebui. So an NPU accellerated text to speech model for reading out chat responses would be really useful.
Requested:
TTS Model Suppor - Introduce quantized/compiled versions of highly efficient TTS models (like Kokoro-82M or Piper) optimized for the XDNA 2 architecture.
API Integration - Expose the standard OpenAI-compatible audio generation endpoint (/v1/audio/speech) in the FastFlowLM server mode.
I'd be happy to help test versions as I have an AMD Ryzen AI 9 HX 370 (XDNA 2 / 50 TOPS).
Operating System
Windows 11, Ubuntu 24.04
GPU
NPU: AMD Ryzen AI 9 HX 370 (XDNA 2 / 50 TOPS)
ROCm Component
FastflowLM
Contributor guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
Start by reviewing FastFlowLM server mode and the existing API surface around the /v1/audio/speech entry point. Done means a quantized or compiled Kokoro-82M or Piper model runs on the XDNA 2 NPU and the server exposes OpenAI-compatible audio generation.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- cpp
- Domain
- ai, api
- Issue type
- Feature
- Difficulty
- 5/5
- Estimated time
- Over a week
- Activity status
- Active
- Clarity
- Needs clarification
- Newbie friendliness
- 30/100