Import analysis options

Every import dialog — Log file, CSV file, Twitch VOD, Live capture — shares the same set of analysis options. These control how the app treats messages after they've been read in: which chatters appear, what counts as a duplicate, whether emote-only spam is filtered out, and how highlights are detected.

This page is the single reference for what each option does. Every import page links here so you don't get the same explanation repeated in six places.

Defaults at a glance

The defaults are tuned for typical streams (1–4 hours, moderate chat). For longer or much busier streams, see the per-option notes below.

Minimum messages

Filters out chatters with very low message counts. By default, anyone who sent fewer than 5 messages is hidden from the chatter list, charts, and category breakdowns — but their messages are still counted in the totals.

Raise this to focus on regulars only and trim the long tail of drive-by viewers. Lower it (down to 1) if you specifically want to see every chatter, including the one-message visitors.

Duplicate detection

The app spots messages that closely match earlier ones, so you can find copypasta and spam waves without manually scanning the log.

Similarity threshold

How alike two messages have to be to count as duplicates. Default 90% — typos, extra spaces, and minor punctuation variations are forgiven, but unrelated messages with the same few words won't match.

• Raise it (e.g. 95%) for stricter matching — only near-identical text
• Lower it (e.g. 75%) to catch loosely-similar copypasta variants

Minimum word count

Short messages like "yes", "lol", and "gg" are skipped from duplicate detection so they don't dominate the duplicates list. Default 3 words. Raise it if "no thanks" and "i agree" keep showing up as matches; lower it if you have a specific reason to flag short repeated phrases.

Discard emote-only messages

When on, messages that consist entirely of emotes are excluded from analysis — they're still counted in the Messages total on the Overview tab, but they don't appear in any of the per-category grids or contribute to engagement tier classification.

Useful when you want stats focused on actual conversation rather than hype reactions. Most users leave this off — emote-only messages are real audience activity and their patterns are interesting in their own right.

Highlight detection

The app scans the chat for moments where activity spiked far above the stream's typical rate. See the Highlights page for what the algorithm actually does — the per-option notes below cover the dials.

Enable highlight detection

On by default. Turn it off if you only care about overall stats and want a slightly faster import.

Highlight window

The length of each time bucket the app analyses, in seconds. Default 60 seconds.

• Smaller windows (e.g. 30 s) catch sharper, briefer spikes — good for fast-paced gameplay reactions
• Larger windows (e.g. 120 s) smooth out brief bursts and surface sustained hype moments instead

Highlight sensitivity

How much busier than the baseline a window has to be before it qualifies as a highlight. Default 1.5. Lower surfaces more moments; higher keeps only the biggest spikes.

Max highlights

A cap on how many highlights end up in the Highlights tab. Default 10. Raise it if you want everything that crossed the threshold; lower it for a tighter shortlist.

When to leave defaults alone

For most analysis runs the defaults give a useful picture. The two options worth tuning are:

1. Highlight sensitivity — for very long or very short streams
2. Minimum messages — when you specifically want to include or exclude one-message viewers

Everything else can usually stay where it is unless something looks wrong in the results.

Next: Main window tour →