Pitch & resonance (formant dispersion) over the recording · shaded bands = your targets
PitchFormant dispersionTarget
Vowel space · F1 × F2
Each cluster is a vowel category · larger spread = more articulation
Per-vowel resonance · F2
Higher = brighter / more fronted
Studio
Pick a recording from the library to see its pitch, resonance, and quality mapped on an interactive timeline — or analyze a new one.
or drop a file anywhere · WAV / MP3 / AIFF
Benchmark Dashboard
Regression Test Results Over Time
Each bar shows how many audio files passed, warned, or failed their metric tolerance checks for a benchmark run. Click a bar to see the full results.
Recent Runs
Click a run to see its full results and baseline comparison.
Date
Branch
Suite
Pass
Warn
Fail
Run Detail
Audio File
Corpus
Pitch Class
Formant Class
Pitch Median (Hz)
F1 (Hz)
Formant Disp. (Hz)
HNR (dB)
No benchmark runs recorded yet. Run uv run python -m benchmarks.run to generate data.
Analysis Speed Over Time
How long it takes to analyze the benchmark audio corpus. "Both modes" runs whole-recording and per-vowel analysis; "Whole only" runs whole-recording alone.
Peak Memory Usage
Maximum resident memory (RSS) during analysis.
No performance benchmarks recorded yet.
Track how a specific metric changes across benchmark runs for a single audio file. Useful for detecting regressions or verifying fixes.
Metric Over Time
Select a metric and audio file above to view trends across runs.