5uck1ess

5uck1ess / tts-bench

Public

Speed and quality benchmark: for all types of TTSs on Windows/Mac/Linux.

19
1
89% credibility
Found May 25, 2026 at 19 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Python
AI Summary

TTS-Bench is an open-source benchmarking tool that compares 25 different text-to-speech models on your own hardware, measuring speed, audio quality, and voice cloning ability, then generates interactive reports with audio samples so you can find the best model for your needs.

How It Works

1
🔍 Discover TTS-Bench

You hear about a tool that lets you compare 25 different text-to-speech models on your own computer, so you decide to check it out.

2
💾 Install the tool

You download and install the software with one simple command — it automatically sets up all the TTS models in separate environments so they don't interfere with each other.

3
⚡ Run the benchmark

With one command, your computer tests every single TTS model using the same phrases, measuring how fast each one speaks and how much memory it uses.

4
Choose your focus
🏃
Speed comparison

See which model generates audio fastest on your specific hardware

🎵
Quality scoring

Listen to samples ranked by an objective quality score

🎭
Voice cloning

Test how well each model can mimic your own voice from a short recording

5
👂 Listen and compare

The report shows inline audio players so you can hear every model rendering the same phrase, making it easy to spot differences in naturalness and clarity.

6
🌐 Share your results

You publish your benchmark results to a website where others can listen to the same audio samples on the same hardware you tested.

🎉 Find your perfect model

You discover which TTS model sounds best and runs fastest on your specific computer, saving you weeks of testing each one individually.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 19 to 19 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is tts-bench?

TTS-bench is a benchmarking framework for local text-to-speech models. It runs the same prompts through 25 different TTS engines and measures three things: how fast each one generates audio (TTFA and RTF metrics), how good the output sounds (using an objective NAQ quality score), and lets you listen to side-by-side samples. The tool runs on Python, supports Windows/Mac/Linux, and works across CPU, NVIDIA CUDA, and Apple Silicon GPUs. You install models via a single script, run `python bench.py`, and get an HTML report with sortable tables and embedded audio players.

Why is it gaining traction?

Most TTS comparisons are scattered forum posts with cherry-picked examples. TTS-bench gives you reproducible, hardware-specific numbers across a model zoo you can actually compare without installing anything. The NAQ (Naturalness-Artifact Quotient) score is particularly useful -- it breaks quality into artifact absence and naturalness presence, so you can see whether a fast model sounds robotic or expressive. The live demo at the project site lets you hear every model on the same hardware before you commit to installing anything.

Who should use this?

Backend developers building voice features who need to pick a TTS engine for production. DevOps evaluating inference costs on specific GPU hardware. Researchers comparing voice cloning quality across OmniVoice, ChatterBox, and IndexTTS. If you're evaluating TTS for an AI agent or accessibility feature, this saves weeks of manual testing across models.

Verdict

With a 0.8999999761581421% credibility score and only 19 stars, this is a one-person project with impressive scope but limited community validation. The documentation is thorough and the HTML reports are polished, but test coverage is unknown and the project is young. Worth using as a reference benchmark -- run it on your own hardware and publish your results to the community. Don't treat the numbers as gospel without verifying on your specific rig.

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.