BUBBLEFollow the rise. Map the bubbles. Find the story between two moments in your sourdough starter’s life.
No account. No photo uploads. A closer look, right in your browser.
Align the jar. Move the surface lines. Then let the image analysis suggest where to look closer.
Same jar, same camera angle. Blue dashed = initial surface. Amber solid = later surface.
Automatic image analysis finds dark, roughly round patterns. Every circle is a candidate, ready for your review.
Select only starter, excluding the lid, jar edges, reflections and labels. Coordinates refer to the later photo after alignment.
Diameters are in the downscaled analysis image’s pixels, not millimeters. Visible marks are not the total bubbles inside the jar. Review is a human correction, not a validated measurement of fermentation strength.
A multi-scale difference-of-Gaussians filter identifies dark rounded regions. Directional contrast checks and overlap suppression reduce repeated detections. Pale bubbles, touching bubbles and glare can be missed or misidentified. No trained AI model or accuracy percentage is claimed.
Measured points stay solid. Estimated curves stay dashed. One feeding cycle at a time.
We fit a quadratic to the latest 5–8 distinct time points over at least 3 hours. A future estimate requires a rising but slowing trend, a concave fit, R² ≥ 0.90 and a vertex within the next 1–4 hours (limited by sampling interval). Leave-one-out fits and half a sampling interval form a sensitivity window. This is not a statistical confidence interval or a validated biological model. A sampled decline is reported separately.
| Time | Height | Source | Edit |
|---|
Your current cycle is saved locally. Previous cycles are archived on this browser (up to 10). Changing conditions requires a new cycle. Old v1 records are left untouched.
Use the same straight-sided jar, diffuse light and camera distance. Align the base and edges; photograph the full starter surface.
Crop reflections out of the analysis region. Check candidate circles. Remove flour spots and jar marks; add bubbles the filter missed.
Note the feeding time, ratio and temperature. Add several measurements as the rise slows. Start a fresh cycle after feeding again.
A photo gives visual evidence. Repeated observations make that evidence more useful.
Gas production and gas retention are different. A very fluid starter may let bubbles escape instead of holding a tall structure. Compare height and visible bubble patterns separately, across consistent feeding cycles.
King Arthur Baking · Sourdough guide ↗Temperature, feeding ratio and the culture’s condition affect timing. A warmer kitchen can mean faster rise. This tool extrapolates your latest height observations; it does not infer biological activity from bubble count, and it does not prescribe a feeding schedule.
King Arthur Baking · Feeding & maintaining ↗A residue line can suggest an earlier rise, but cannot timestamp the peak. The tallest measured point is an observed maximum. A decline after that point supports a past peak somewhere between samples; it does not reveal its exact time or height.
No. Photo reading, analysis, alignment and card creation happen in your browser. The tool stores observation numbers and cycle metadata locally, not photos. Blogger and Google advertising services retain their own data policies.
It means a dark rounded pattern passed the filter. Glare, flour spots and jar texture can still confuse it. Automatic candidates are image estimates. Reviewed marks mean you checked the visible regions, not that the algorithm was scientifically validated.
At least five distinct measurements spanning three hours are required. The rise must slow in a stable curve, and the estimated peak must be nearby. Linear growth, noisy records, changed conditions and too little data suppress the forecast. After a measured decline, we show observed evidence instead of predicting a future peak.
No. This is a visual observation and exploratory forecasting tool. Neither bubble count nor a fitted height curve establishes microbial activity, food safety or baking readiness.