Two Compression Modes
WebP is unique among common image formats because it offers both lossy and lossless compression in a single format. JPG only supports lossy. PNG only supports lossless. WebP supports both, and the two modes use entirely different compression algorithms.
Lossy Mode
WebP lossy compression is based on the VP8 video codec's intra-frame prediction. It works by predicting each block of pixels from neighboring blocks, then encoding only the difference (residual). The quality parameter controls how aggressively the residual data is quantized — higher quality means less quantization and more data preserved, lower quality means more aggressive quantization and smaller files.
Lossy WebP is ideal for photographs and complex natural images where a small amount of imperceptible data loss translates to dramatic file size reductions. At quality 80, a lossy WebP is typically 65–70% smaller than the uncompressed original with no visible difference to the human eye.
Lossless Mode
WebP lossless compression uses a completely different algorithm based on spatial prediction, color transforms, LZ77 backward references, and optimized Huffman coding. Every single pixel in the original image is preserved exactly — decompressing a lossless WebP produces a bit-for-bit identical image to the source.
Lossless WebP is typically 26% smaller than PNG for the same image. It is ideal for screenshots, icons, logos, UI elements, technical diagrams, and any image where pixel-perfect accuracy matters. The quality parameter in lossless mode (0–9) controls the compression effort (encoding speed), not the output quality — all settings produce identical output, just at different encoding speeds.
Key distinction: In lossy mode, the quality parameter (0–100) controls output quality. In lossless mode, the "method" parameter (0–6) controls encoding speed. The output of lossless compression is always pixel-perfect regardless of settings.
The Quality Parameter (0–100)
WebP's lossy quality parameter works similarly to JPEG's quality setting but the scales are not directly equivalent. Here is what each quality range produces:
Quality 50–60: Heavy Compression
At these settings, WebP produces very small files at the cost of visible quality loss. Compression artifacts become noticeable: smoothing of fine textures, loss of subtle color gradations, and a "plastic" appearance on skin tones. Fine details like individual hairs, fabric weave, and text in photographs are visibly degraded.
Use case: Extremely bandwidth-constrained environments where file size is critical and image quality is secondary. Low-priority background images on slow mobile connections. Preview thumbnails where the image will never be viewed at full size.
Quality 70–80: Web Standard
This is the sweet spot for web images. At quality 75–80, compression artifacts are imperceptible to most viewers at normal viewing distances and screen sizes. Fine textures are well-preserved, color transitions are smooth, and the file size savings compared to quality 100 are substantial (typically 60–70% smaller).
Quality 75 is Google's default for the cwebp encoder and is the most commonly used setting for web delivery. Quality 80 is the second most popular choice, offering slightly better detail retention for a modest file size increase.
Use case: General web images, e-commerce product photos, blog images, social media content, news photography. This is the correct range for the vast majority of web applications.
Quality 85–90: High Quality
At these settings, WebP preserves virtually all visible detail. Even careful pixel-level comparison against the original reveals minimal differences. The file size is larger than Q75–80 but still significantly smaller than lossless or uncompressed.
Convertio.com's default falls in this range, ensuring that converted images maintain excellent quality for all practical purposes. This is also the recommended range for photography portfolios, high-end e-commerce (where product detail matters for purchasing decisions), and any context where image quality is a priority but bandwidth still matters.
Use case: Photography portfolios, high-end product imagery, medical reference images (non-diagnostic), professional presentations, and situations where the image may be viewed at 100% zoom.
Quality 95–100: Near-Lossless
Quality 100 applies the minimum possible quantization to the DCT coefficients. It is not lossless — some data is still discarded during the prediction and transform steps — but the loss is mathematically minimal. Quality 95–100 files are only 20–30% smaller than the lossless equivalent, which raises the question of whether you should simply use lossless mode instead.
The practical visual difference between Q100 lossy and true lossless is invisible to the human eye. However, for applications where provably zero data loss matters (archival, scientific imaging, pixel art), true lossless is the correct choice because Q100 technically does modify pixel values.
Use case: Archival where lossless is too large, master copies where negligible loss is acceptable, source images that will undergo further processing.
Lossless Mode in Detail
WebP lossless mode is not simply "quality 100+" — it uses a completely different compression algorithm. While lossy WebP is based on VP8 video prediction and DCT, lossless WebP uses a system designed specifically for exact pixel preservation:
- Spatial prediction with 13 modes — far more prediction options than lossy mode's 4 intra modes
- Color transform — decorrelates RGB channels to reduce entropy
- Subtract-green transform — exploits the high correlation between green and the other channels in natural images
- Palette coding — automatically detects images with 256 or fewer unique colors and switches to indexed mode
- LZ77 backward references — finds and replaces repeated pixel patterns
- Huffman coding — optimal entropy coding for the compressed stream
The result is that lossless WebP files are approximately 26% smaller than equivalent PNG files. For images with limited color palettes (icons, simple logos, flat-design UI), the savings can exceed 40% because the palette coding and LZ77 references are extremely effective on these image types.
Lossless WebP guarantees: Every pixel value in the decoded image is identical to the original. This is a mathematical guarantee — the compression is fully reversible. The file size depends on image content: simple graphics compress extremely well, complex photographs compress modestly.
Recommended Settings by Use Case
| Use Case | Mode | Quality | Why |
|---|---|---|---|
| Web photos (general) | Lossy | 75–85 | Best balance of size and quality |
| E-commerce products | Lossy | 80–90 | Product detail matters for sales |
| Photography portfolio | Lossy | 85–92 | Showcase quality is critical |
| Thumbnails | Lossy | 60–70 | Small display size hides artifacts |
| Hero/banner images | Lossy | 75–85 | Large files; size savings matter most |
| UI elements / icons | Lossless | N/A | Sharp edges degrade under lossy compression |
| Logos | Lossless | N/A | Pixel-perfect brand consistency required |
| Screenshots | Lossless | N/A | Text and UI elements need sharp edges |
| Archival / backup | Lossless | N/A | No data loss, 26% smaller than PNG |
File Size at Different Quality Levels
Here is how file size varies with quality for a typical 1200×800 photograph (representative of web usage):
| Setting | File Size | vs Original | Visual Quality |
|---|---|---|---|
| Lossy Q50 | 85 KB | 97% smaller | Visible artifacts, "plastic" smoothing |
| Lossy Q60 | 110 KB | 96% smaller | Minor artifacts on close inspection |
| Lossy Q70 | 150 KB | 95% smaller | Good — artifacts rarely noticeable |
| Lossy Q75 | 185 KB | 94% smaller | Excellent — Google default |
| Lossy Q80 | 230 KB | 92% smaller | Excellent — most popular web setting |
| Lossy Q85 | 290 KB | 90% smaller | Very high — negligible difference from original |
| Lossy Q90 | 395 KB | 86% smaller | Near-perfect — differences only at 200%+ zoom |
| Lossy Q100 | 820 KB | 72% smaller | Minimal quantization — still technically lossy |
| Lossless | 2.2 MB | 25% smaller | Perfect — bit-for-bit identical to source |
Source: 1200×800 photograph, 2.9 MB as PNG. Sizes are representative; actual values vary by image content.
The table illustrates a crucial point: the relationship between quality and file size is not linear. Going from Q50 to Q75 increases file size by about 2x while dramatically improving quality. Going from Q90 to Q100 doubles file size again but produces almost no visible improvement. This is why Q75–85 is the "sweet spot" — it sits at the knee of the quality-size curve where diminishing returns begin.
Lossy vs Lossless: How to Choose
The decision between lossy and lossless WebP depends on your image content and use case:
Choose lossy when:
- The image is a photograph or has complex, continuous tonal variation
- File size reduction is important (web delivery, bandwidth constraints)
- The image will be viewed at normal sizes, not zoomed to 200%+
- A small amount of imperceptible data loss is acceptable
Choose lossless when:
- The image contains text, sharp edges, or geometric shapes (screenshots, logos, icons)
- Pixel-perfect accuracy is required (scientific imaging, pixel art, technical diagrams)
- The image will undergo further editing or processing (lossless prevents cumulative degradation)
- You need a smaller alternative to PNG without any quality compromise
- The image has few unique colors (lossless WebP with palette coding is very efficient)
Simple rule: If the image came from a camera, use lossy (Q75–85). If it came from a computer (screenshot, design tool, code editor), use lossless. This heuristic is correct for the overwhelming majority of cases.
Try Converting with Different Settings
The best way to understand quality settings is to experiment with your own images. Upload a WebP file to Convertio.com and see the result — our converter uses high-quality settings designed to preserve maximum detail while producing a universally compatible JPG file.
When the source WebP uses lossy compression (which is the case for most images saved from websites), converting to JPG introduces minimal additional quality loss because both formats use similar perceptual compression strategies. When the source is lossless WebP, the JPG output will be an excellent high-quality lossy version.