BlueTTS Artifact Benchmark

Deterministic browser-side backend comparisons against an FP32 reference.

Ready to run

Reference backend

Candidates

Voice

Speed 1.00

Steps

CFG scale

Seed runs

Base seed

Prompt corpus

One prompt per line. Each prompt gets deterministic seed variants.

Back to Synth App

Summary

Ready to run

Reference backend: WASM · FP32 · Stock ORT.

Backend	Artifact Score	Verdict	Quiet Floor vs Ref	Quiet High-Band vs Ref	High-Band Ratio vs Ref	Clicks / s vs Ref	Mean Render	Worst Run

No benchmark results yet.

Worker progress and render order

Waiting to run artifact benchmark.

Machine-readable results