Optimization
Content detection
How Inverity classifies each asset, and why it matters.
Content detection is the step that makes everything else work. Before Inverity compresses anything, it decides what kind of content it's looking at — because the right compression for a photograph is the wrong compression for a logo.
Why one setting can't work
Consider three files that might sit in the same folder:
- A product photograph. Continuous tone, lots of fine gradient detail, no hard edges. Lossy compression works extremely well here; the eye doesn't track small errors in noisy detail.
- A UI screenshot. Large flat areas, hard edges, and text. The same lossy settings that flatter a photograph will put visible ringing artifacts around every letter.
- A logo with transparency. Few colours, hard edges, an alpha channel. It needs a format that preserves transparency, and it compresses best with a completely different approach again.
Apply one quality setting to all three and you either wreck the screenshot or under-compress the photo. Inverity classifies first, then picks.
Image content types
| Type | What it looks like | What drives the choice |
|---|---|---|
| Photo | Camera imagery, product shots, lifestyle | Continuous tone; tolerates aggressive lossy encoding |
| Screenshot | UI captures, dashboards, documents | Text and hard edges; needs edge-preserving treatment |
| Icon | Logos, glyphs, simple marks | Few colours, small dimensions; often palette-friendly |
| Transparent graphic | Overlays, badges, cut-outs | Alpha channel must survive the round trip |
See Image optimization for what each one means in practice.
Video content types
| Type | What it looks like | What drives the choice |
|---|---|---|
| Talking head | Interviews, webinars, presenters | Static background, small moving region |
| Screen recording | Demos, tutorials, screencasts | Text legibility, long static stretches |
| Motion graphic | Animation, explainers, titles | Synthetic colour, sharp edges, hard cuts |
| Archive master | High-resolution source footage | Fidelity matters more than size |
See Video optimization.
When detection is ambiguous
Some assets genuinely sit between categories — a screenshot with a photograph inside it, or a motion graphic with live-action footage cut into it. Inverity picks the classification that best fits the asset overall and, because every result still goes through the quality check, an imperfect classification produces a smaller saving or a skip rather than a damaged file.
You can't override the classification
Detection is automatic and there's no per-asset override. What you can control is the overall aggressiveness, using quality profiles — set per connected source, so you can treat your DAM masters differently from your marketing imagery.
Practical consequences
- Mixed folders are fine. You don't need to separate photos from screenshots before connecting a source. That's the whole point.
- Savings vary by folder, and that's expected. A screenshot-heavy docs library will show lower average savings than a photography library. Compare each folder against itself over time, not against another folder.
- Odd content skips more. Highly noisy, already-compressed, or unusual assets are more likely to be kept as originals. See Results and savings.