Compress Video

About Video Compression Tool

This online tool allows you to compress video files directly in your browser. Reduce video file size while maintaining good quality. Perfect for sharing videos online, uploading to social media, or saving storage space. All processing happens locally for complete privacy. See also our Video Joiner and Change Aspect Ratio.

How do I compress a video without losing quality?

True lossless compression requires re-encoding with a lossless codec (FFV1, x264 --qp 0) and typically produces files larger than the original H.264 source, so it is rarely what users actually want. What people mean by lossless is visually lossless: pick the H.264 (AVC, ITU-T H.264) Balanced preset, use Constant Rate Factor (CRF) 18-20, and keep the source resolution and frame rate. CRF 18 is widely considered the threshold below which most viewers cannot tell the difference from the source on a typical screen. Avoid stacking compression by re-encoding an already-compressed file multiple times. If your goal is just smaller files for web sharing, try H.265 (HEVC) or AV1 at CRF 22-24, which keep visual quality close to H.264 CRF 18 at roughly half the bitrate.

Which output format should I choose, MP4 or WebM?

Pick MP4 (ISO/IEC 14496-14 container with H.264 video + AAC audio) for maximum compatibility. Every modern phone, browser, social platform, TV, and editing program plays MP4/H.264 natively, and hardware decoders mean low battery and CPU use during playback. Pick WebM (VP9 or AV1 video + Opus audio) when you target modern browsers and want better compression for the same visual quality, particularly for web embedding or background loops. WebM is the open-royalty-free choice favored by YouTube and many CDNs. As a rule of thumb: sending to friends, social media, or older devices, choose MP4; embedding on your own modern website where bandwidth matters, choose WebM.

What does CRF mean and which value should I use?

CRF stands for Constant Rate Factor, the rate control mode used by x264, x265, libvpx, and SVT-AV1 to target a perceptual quality level instead of a fixed bitrate. The scale runs roughly 0-51 for H.264/H.265, where lower is higher quality and larger files. Practical ranges: CRF 17-18 visually lossless, CRF 20-23 high quality (recommended default), CRF 24-26 good quality with significant size savings, CRF 28-30 acceptable for low-bandwidth distribution, CRF 32+ visibly degraded. AV1 uses a similar but not identical 0-63 scale, where CRF 30-32 is roughly comparable to H.264 CRF 23. CRF is preferable to fixed bitrate for one-pass encoding because it allocates more bits to complex scenes and fewer to static ones.

Why is my compressed video larger than I expected?

Several common reasons: the source was already heavily compressed (re-encoding cannot magically shrink it further without quality loss), you chose a preset that prioritizes quality (High Quality, CRF below 20), the resolution or frame rate is high (4K60 needs 4x the bitrate of 1080p30 for similar quality), or the content is high-motion (sports, video games, particle effects) which is intrinsically hard to compress. To force smaller output, lower the resolution one step (4K to 1080p, 1080p to 720p), drop frame rate from 60 to 30 fps if motion allows, raise CRF by 4-6 points, or switch from H.264 to H.265/AV1 which typically halve file size at equivalent quality. Audio bitrate also matters: 128 kbps stereo AAC is fine for most content.

What is the difference between H.264, H.265 (HEVC), VP9, and AV1?

These are video codecs at different points on the compression-versus-compute curve. H.264 (AVC, 2003) is the universal baseline: every device decodes it in hardware. H.265 (HEVC, 2013) achieves roughly 50% better compression than H.264 at the same quality but carries patent royalties and uses more CPU to encode. VP9 (Google, 2013) is royalty-free and roughly matches HEVC, widely used by YouTube. AV1 (Alliance for Open Media, 2018) is the newest royalty-free codec, offering 20-30% better compression than HEVC/VP9, but software encoding is very slow without hardware acceleration. For maximum compatibility, choose H.264. For best size-versus-quality on the modern web, choose AV1 if your encoder supports it, otherwise VP9 or HEVC.

Should I use one-pass or two-pass encoding?

Two-pass encoding analyzes the entire video first to map scene complexity, then encodes a second time using that map to allocate bits optimally. This produces better quality at a target average bitrate, especially for videos with varied complexity (talking heads cutting to action sequences). The cost is double the encoding time. Use two-pass when you have a strict file-size or bitrate target, for example uploading to a platform with hard caps or streaming over a fixed-bandwidth connection. Use one-pass with CRF when quality matters more than exact file size: CRF rate control already adapts bit allocation to scene complexity in a single pass and matches two-pass quality at unconstrained file size. For browser-based encoding via FFmpeg.wasm, single-pass CRF is usually the right default because two passes effectively double tab compute.

How do GOP structure, keyframes, and I/P/B frames affect compression?

A Group of Pictures (GOP) is the repeating pattern of frame types between keyframes. I-frames (intra-coded) are full standalone images, the largest and least compressed. P-frames (predicted) encode only differences from earlier frames. B-frames (bidirectional) reference both earlier and later frames for the smallest size. Longer GOPs (more P/B frames between I-frames) yield smaller files but worse seeking and worse error recovery; for web playback, a keyframe interval of 2-4 seconds (60-120 frames at 30 fps) is the standard balance. Streaming (HLS, DASH) often forces a keyframe every 2 seconds to enable segment switching. For social media uploads, keep GOP around 2 seconds; for archival or editing, shorter GOPs (every 1 second or all-I-frame) give cleaner cuts at the cost of size.

When does compression need 10-bit color or HDR-aware encoding?

Standard 8-bit color (sRGB or Rec.709) stores 256 levels per channel, enough for SDR content but prone to banding in dark gradients or skies. 10-bit color (Main10 profile in HEVC, High10 in H.264) stores 1024 levels and effectively eliminates banding, even on SDR content. It also paradoxically compresses better in many codecs because the encoder has finer granularity for prediction. HDR content (HDR10, Dolby Vision, HLG using Rec.2020 color and PQ or HLG transfer functions) requires 10-bit minimum plus HDR metadata signaling. If your source is HDR and you re-encode to 8-bit SDR without tone mapping, highlights clip and the image looks washed out. For browser-based compression, stick to 8-bit unless your source is genuinely HDR and your target devices support HDR playback, since browser HDR pipelines remain inconsistent.