Why Compression Settings Matter
Video files are enormous without compression. One minute of uncompressed 1080p video at 24 fps takes approximately 5.3 GB of storage. Compression reduces this to 50-200 MB depending on settings — a 25-100x reduction. The settings you choose determine exactly where your video falls on the spectrum between tiny-but-ugly and huge-but-perfect.
The problem is that video compression settings are genuinely confusing. CRF, CBR, VBR, two-pass, presets, profiles, levels, tunes — the terminology is dense, the interactions are complex, and the wrong combination can produce terrible results even when each individual setting seems reasonable.
This guide explains what each setting actually does, how they interact, and which values to use for common scenarios. We focus on H.264 and H.265 since they cover 95% of video encoding needs, but the concepts apply to VP9 and AV1 as well.
Rate Control: How Quality Is Managed
Rate control determines how the encoder allocates bits across the video. This is the single most important compression decision.
CRF (Constant Rate Factor)
CRF is the recommended mode for single-pass, file-based encoding. You specify a quality target, and the encoder adjusts bitrate frame-by-frame to maintain that quality level. Complex scenes (action, confetti, water) get more bits; simple scenes (talking head, static slide) get fewer bits.
H.264 CRF scale: 0-51 (0 = lossless, 51 = worst quality)
- 0: Lossless (enormous files)
- 15-17: Visually lossless (hard to distinguish from source)
- 18-20: High quality (suitable for archival)
- 21-23: Very good quality (recommended for general use)
- 24-27: Good quality (noticeable on close inspection)
- 28-32: Medium quality (visible compression)
- 33+: Low quality (significant artifacts)
H.265 CRF values are shifted roughly 4-6 points higher for equivalent quality. CRF 28 in H.265 produces roughly the same quality as CRF 22 in H.264 — but at 40-50% smaller file size.
The "default" CRF of 23 (H.264) or 28 (H.265) is the encoder's recommended balance between quality and file size. It is a good starting point for most content.
CBR (Constant Bitrate)
CBR produces a fixed bitrate regardless of scene complexity. Simple scenes waste bits; complex scenes lack bits. This produces inconsistent quality but predictable file sizes and bandwidth usage.
Use CBR for:
- Live streaming (platforms require predictable bitrate)
- Broadcast and CDN delivery (network capacity planning)
- Never for local files or uploads (CRF produces better quality)
VBR (Variable Bitrate)
VBR varies the bitrate within defined limits. It is a middle ground between CRF's quality-first approach and CBR's size-predictability:
- 1-pass VBR: Encoder guesses complexity in real time. Better than CBR, worse than CRF.
- 2-pass VBR: First pass analyzes the entire video; second pass encodes with optimal bit allocation. Produces the best possible quality at a target file size. Recommended when you need a specific output size.
Comparison
| Mode | Quality | File Size | Speed | Best For |
|---|---|---|---|---|
| CRF | Consistent | Variable | Fast (1 pass) | Local files, uploads |
| CBR | Variable | Fixed | Fast (1 pass) | Live streaming |
| 1-pass VBR | Moderate | Bounded | Fast | Quick encodes |
| 2-pass VBR | Best at target size | Target | Slow (2 passes) | Delivery when size matters |
Encoding Presets
Presets control how much CPU time the encoder spends optimizing compression. Slower presets find better compression opportunities, producing smaller files at the same quality — or better quality at the same size.
H.264 (x264) Presets
From fastest to slowest: ultrafast, superfast, veryfast, faster, fast, medium, slow, slower, veryslow, placebo.
| Preset | Encode Speed | File Size (relative) | Use Case |
|---|---|---|---|
| ultrafast | 10x faster than medium | 200-300% larger | Real-time capture |
| veryfast | 4x faster | 130-150% larger | Live streaming |
| medium | Baseline | Baseline | General encoding |
| slow | 40% slower | 10-15% smaller | Quality encoding |
| veryslow | 3x slower | 15-20% smaller | Final delivery |
| placebo | 10x slower | 1-2% smaller than veryslow | Never worth it |
The jump from medium to slow provides the best return on time investment. Going beyond veryslow provides negligible improvement and is not worth the encoding time for any practical purpose.
For H.265 (x265), the same preset names apply but encoding is generally 2-5x slower than H.264 at the same preset.
Profiles and Levels
Profiles
Profiles define which encoding features the codec can use. Higher profiles enable more advanced compression tools but require more decoding capability:
H.264 Profiles:
- Baseline: No B-frames, no CABAC. Maximum compatibility (older mobile devices, video conferencing). Worst compression.
- Main: B-frames and CABAC enabled. Better compression. Suitable for most playback devices.
- High: 8x8 transforms and additional prediction modes. Best compression. Required for professional use and recommended for all modern workflows.
H.265 Profiles:
- Main: 8-bit encoding. Standard for consumer content.
- Main 10: 10-bit encoding. Required for HDR content. Also produces slightly better compression for SDR content due to finer quantization.
Levels
Levels define maximum resolution, frame rate, and bitrate combinations. They exist so decoders know their hardware limits:
- Level 3.1: Up to 720p30 at 14 Mbps
- Level 4.0: Up to 1080p30 at 20 Mbps
- Level 4.1: Up to 1080p30 at 50 Mbps (common for Blu-ray)
- Level 5.1: Up to 4K30 at 160 Mbps
- Level 5.2: Up to 4K60 at 240 Mbps
Most encoders auto-select the appropriate level based on your resolution and bitrate settings. You rarely need to set it manually.
Tune Options
Tune parameters optimize the encoder for specific content types:
| Tune | Effect | Best For |
|---|---|---|
| film | Higher B-frame count, lower deblocking | Live-action footage |
| animation | Higher deblocking, lower motion weight | Cartoons, anime |
| grain | Preserves film grain instead of smoothing it | Grainy footage, film scans |
| stillimage | Optimized for static content | Slideshows, presentations |
| zerolatency | Disables lookahead buffer | Live streaming, video chat |
| fastdecode | Disables B-frames and CABAC | Playback on weak hardware |
Quality and Settings Tips
CRF is almost always the right choice for file-based work. Unless you are streaming live or need a specific target file size (two-pass VBR), CRF produces the best quality-to-size ratio with single-pass efficiency.
The preset matters more than you think. Moving from veryfast to slow at the same CRF value can reduce file size by 30-40% — or equivalently, you can increase CRF by 2-3 points at slow and still match the quality of veryfast at a lower CRF, with a much smaller file.
Resolution reduction is free compression. Before tweaking CRF and presets, ask whether your output actually needs to be 4K or 1080p. Downscaling from 4K to 1080p reduces data by 75% before the codec even gets involved. For most web and social media delivery, 1080p or 720p is sufficient. See our 4K conversion guide for resolution decision-making.
Audio compression is often overlooked. Video gets all the attention, but audio settings matter too. AAC at 128 kbps is transparent for most content. For music-heavy video, 192-256 kbps. For speech-only content (tutorials, presentations), 96 kbps AAC is sufficient. See our audio bitrate guide for details.
Test with short clips first. Before encoding a 2-hour video at veryslow, encode a 30-second section with your intended settings. Verify quality visually, check the file size, and adjust before committing to the full encode.
Common Issues and Troubleshooting
File size is unexpectedly large. Check your preset (using ultrafast?), CRF value (below 18?), and resolution. The most common cause is an inappropriate preset — encoding at CRF 23 with ultrafast produces files 2-3x larger than CRF 23 with medium.
Quality looks terrible despite reasonable settings. The source material may already be heavily compressed. Re-encoding a low-bitrate source does not improve quality — it can only maintain or degrade. If the source is a 2 Mbps streaming rip, no CRF value will make it look like a Blu-ray.
Encoding takes forever. Switch to a faster preset. The quality difference between slow and medium is 10-15% file size — often not worth the 40% longer encoding time. For batch encoding, medium is the practical sweet spot.
Player shows "unsupported codec." Check that your profile and level are compatible with the target device. Older devices may not support High profile H.264 or Main 10 H.265. Lower the profile to Main or Baseline for maximum compatibility.
Colors look washed out after encoding. This often indicates a color space or range mismatch. Ensure the output color range (limited/full) matches the source, and that the color primaries (BT.709 for HD, BT.2020 for HDR) are correctly tagged.
Conclusion
For most video encoding: use CRF mode (23 for H.264, 28 for H.265), slow preset, High profile, and let the encoder auto-select the level. This combination produces excellent quality at reasonable file sizes with straightforward configuration. Adjust CRF up for smaller files (web sharing) or down for higher quality (archival), and use two-pass VBR only when you need precise file size control.
Ready to compress? Try our free video compressor with adjustable CRF and preset controls — no registration required.
