Automatic peak detection and timing analysis web app

AKA

Martin had been manually notating audio (.WAV) files to find the timing differences between beats.

The TIGHTINATOR turns potentially hours of highly detailed tedium into a quick drag-and-drop web workflow.

As described in the videos, Yan stood up and said "I will take the ring to Mordor." He needed a computer and an environment, so I set him up on mine while I played some marble machine/mole themed DnD (5e.)

A few hours later, Yan had a viable analyzer completed; WAV file input, peak detection, CSV generation, and some initial plotting/graphing. We went over the work with Martin, and with positive feedback—and getting pushed on stage to present on the livestream—I set about configuring a web environment to handle uploads, take user configuration, handle errors, and display results.

In production a Flask app is served using gunicorn. The machine is a tiny Digital Ocean droplet, and every so often I have to delete the uploads directory to ensure the disk doesn't fill up. Easy enough to automate. I built a custom SELinux context on this machine, and somehow got the app processing all the way through.

With the small remaining bits of my vacation, we enhanced the user workflow, adding the ability to rerun an analysis, display previous summaries, alter the graph legends, and added logging and monitoring. The final steps were a bit of landing page branding, and settling on a name.

Once TIGHTINATOR was agreed upon, I looked for the cheapest available domain, and completed the migration.

Behold! The TIGHTINATOR!

Use with Huygen drive:
As featured in the prototype series:

Drive mechanism research.


The first time the tool was shown.


He's really stretching his legs out in this one.


Martin doesn't show it in the above video, but a lot of a back and forth happened about the best way to record these tests, and how to differentiate between test results he can and cannot trust. In the above video he mentions that one single transient looks exactly like the similarness plot, which increases his confidence in the automatically computed standard deviation. In saying that he is alluding to previous analyses that did not have similarness lots that were as consistently stacked, or as clearly defined. It does clearly communicate that this graph, while not providing any valuable statistical analysis, is representative of the overall quality of this set of results.

Traffic Stats

Tightinator traffic stats

Wow Martin gets a lot of reach. I'm surprised this many people actually click into the app this much. I guess Americans are pretty curious. That Swedish traffic is probably entirely Martin. I hope I don't need to scale this.