EasyLottie Research
Lottie File Statistics 2026
We parsed 2,090 Lottie JSON files to measure their size, duration, layers, animation features, easing, and image-asset usage. After exact-byte deduplication, the corpus contained 605 unique files.
2,090
JSON files parsed
No parse failures
605
unique contents
Deduplicated by SHA-256
5.3 KB
median file size
22.7 KB at P90
2.0 s
median duration
7.0 seconds at P90
Key findings
Most files were short and small
The median file was 5,336 bytes and 60 frames, or 2 seconds. At P90, files were 22,674 bytes, 240 frames, and 7 seconds. The largest file was 322,096 bytes; the longest ran for 40 seconds.
Image layers dominated this corpus
Image layers represented 73.34% of top-level layers. The corpus included 851 image-only files, 138 vector-only files, and 1,101 mixed files. This describes this corpus, not every Lottie library.
Complex structure was not rare
Precompositions appeared in 27.80% of files. Parenting appeared in 23.35%, and masks or track mattes appeared in 20.62%. Supporting only flat transforms would miss a material share of these animations.
Two easing patterns accounted for 48.88%
Across 109,588 examined keyframes, 27.18% were classified as linear. The two most frequent non-linear handle pairs together accounted for 48.88% of every examined keyframe.
Feature prevalence
Raw file prevalence can be inflated when the same animation is copied across products. The unique-content column deduplicates exact file bytes with SHA-256, so both views are shown.
| Feature | All files | Unique contents | Occurrences raw / unique |
|---|---|---|---|
| Precompositions | 581 (27.80%) | 175 (28.93%) | 3,091 / 945 |
| Masks or track mattes | 431 (20.62%) | 153 (25.29%) | 1,355 / 554 |
| Parenting | 488 (23.35%) | 131 (21.65%) | 3,032 / 1,029 |
| Trim paths | 328 (15.69%) | 68 (11.24%) | 418 / 83 |
| Gradients | 69 (3.30%) | 41 (6.78%) | 151 / 98 |
| Effects | 28 (1.34%) | 14 (2.31%) | 752 / 104 |
| Text layers | 5 (0.24%) | 3 (0.50%) | 57 / 29 |
Layer architecture
Top-level counts describe only each document's root layers. Recursive counts also traverse layers in referenced precompositions.
| Layer type | Top-level count | Top-level share | Recursive count | Recursive share |
|---|---|---|---|---|
| Image | 12,501 | 73.34% | 17,708 | 69.90% |
| Shape | 2,537 | 14.88% | 4,020 | 15.87% |
| Precomposition | 1,727 | 10.13% | 3,091 | 12.20% |
| Null | 211 | 1.24% | 360 | 1.42% |
| Solid | 70 | 0.41% | 99 | 0.39% |
| Text | 0 | 0% | 57 | 0.22% |
Animated transforms
Scale was the most frequently animated transform. Position paths were often curved: 16,308 of 23,668 animated-position keyframes, or 68.90%, included spatial tangents.
| Transform | Animated layers | Share of recursive layers |
|---|---|---|
| Scale | 9,550 | 37.69% |
| Opacity | 7,595 | 29.98% |
| Position | 7,240 | 28.58% |
| Rotation | 5,136 | 20.27% |
| Anchor point | 265 | 1.05% |
Assets, timing, and easing
Frame rates
25 fps: 1,270 files
30 fps: 446 files
60 fps: 269 files
Other: 105 files
Image assets
External assets: 1,900 files
Embedded assets: 15 files
No image assets: 175 files
Anchor points
Centered: 91.02%
Bottom: 2.05%
Top: 0.36%
Other: 6.57%
Most common easing handles
The leading handle pair was outgoing (.333, 0) and incoming (.667, 1), observed 27,700 times. The second was (.167, .167) to (.833, .833), observed 25,867 times.
Methodology and limits
- We attempted to parse 2,090 candidate JSON files and successfully parsed all 2,090.
- Exact file bytes were hashed with SHA-256, producing 605 unique contents and 1,485 duplicate copies.
- Layer-type totals use both a root-only view and a recursive view that follows referenced precompositions.
- Feature, transform, shape, and easing counts use the recursive view.
- Linear easing means handles were absent, or all four handle coordinates were within 0.02 of 0.5.
- Anchor normalization includes only static anchors on image, solid, and precomposition layers with known dimensions.
Important sampling limitation
This was an engineering corpus, not a random sample of every Lottie file on the public web. It contains repeated production assets and is unusually image-heavy. The results describe the analyzed corpus and should be used as implementation evidence, not as a universal market-share estimate. Source files are not redistributed; the downloadable JSON contains aggregate statistics only.
How to cite this analysis
You may quote the aggregate numbers with attribution and a link to this page. For questions about the methodology or a correction, use the contact page.
Inspect your own Lottie
Open a JSON, ZIP, or .lottie file locally in your browser to inspect dimensions, frame rate, duration, layers, and assets. Your file is not uploaded.