# andimarafioti/faster-qwen3-tts

Real-time text-to-speech with Qwen3-TTS

Repository: https://github.com/andimarafioti/faster-qwen3-tts
Canonical: https://ross.abutalabs.com/products/faster-qwen3-tts
Language: Python
License: MIT
License Family: permissive
Last push: 2026-08-25T13:44:23+00:00

## Health v2 (maintenance only)
Score: 78/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 99, release rhythm 87, longevity 14
- inputs: {"age_days": 198, "days_push": 8, "days_rel": 8, "gap_med": 40, "n_releases_24m": 4}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1330, forks 192 (observed 2026-08-28T04:04:23.782910+00:00)

## What it is
A Python library for real-time text-to-speech inference with Qwen3-TTS using manual CUDA graph capture, requiring no Flash Attention, vLLM, or Triton. It supports both streaming and non-streaming generation, with an experimental GGML backend via qwentts.cpp.

## Use cases
- generate speech from text in real time on an NVIDIA GPU
- stream TTS audio with low latency
- run Qwen3-TTS without vLLM or Triton
- design voices with an instruct prompt like 'warm, confident narrator'
- run TTS with a quantized GGML backend
- deploy text-to-speech on T4 or Blackwell GPUs

## When to choose
- you need fast, real-time Qwen3-TTS inference on NVIDIA hardware
- you want streaming TTS without heavyweight inference stacks
- you want a simple pip-installable TTS library with CLI support

## When to avoid
- you have no NVIDIA GPU or an old CUDA driver
- you need CPU-only or non-NVIDIA inference (except experimental GGML)
- you need a production-hardened, long-term-stable TTS solution

## Facets
- artifact type: library
- maturity: active
- function: tts, llm-inference, audio-processing
- domain: speech-processing, machine-learning
- platform: python, cli
- tags: text-to-speech, cuda-graphs, qwen3, streaming, real-time, nvidia-gpu, audio, linux, gpu

## Member repositories
- andimarafioti/faster-qwen3-tts (main) score 78

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:04:23.782910+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T04:45:23.680820+00:00, confidence not recorded.
  - readme: https://github.com/andimarafioti/faster-qwen3-tts (fetched 2026-08-28T04:04:23.782910+00:00, sha bf529fd013b1)
  - registry_pypi: https://pypi.org/pypi/faster-qwen3-tts/json (fetched 2026-08-29T12:04:30.941124+00:00, sha d8e2b253136b)
- Data as of 2026-08-30T08:39:29.467469+00:00.