Learn FFmpeg - Video Encoding & Codecs

#Introduction

After episode 3 you can run your first transcode — format conversion and stream copy — this episode goes into the heart of quality: video encoding. This is where the difference between "playable" results and "production-worthy" results is decided.

You won't memorize all of FFmpeg's options; what you need is the right mental model. Think of encoding like tuning an espresso machine: there are several interrelated parameters, and you must know which to turn first, which to sacrifice, and which should never be touched carelessly.

#Choosing a Codec: The Language of Video

In episode 0 you learned that a codec is a compression algorithm. Now let's look at the four video codecs you'll encounter most often and their encoders in FFmpeg:

The choice is like choosing a language: H.264 is the universal language understood by all devices; HEVC saves more space but not every player understands it; AV1 is the most economical for the same quality, but needs far more compute power to encode. For general content, H.264 is still the compatibility champion.

Check which encoders are available in your build:

ffmpeg -encoders | grep -E "libx264|libx265|libvpx-vp9|libsvtav1"

To see all options of an encoder, run ffmpeg -h encoder=libx264 — the full parameter list with descriptions.

#Specifying the Codec: -c:v

The -c:v option selects the video codec; -c:a for audio. The full command:

ffmpeg -i sample.mp4 -c:v libx264 -crf 18 -preset slow out-h264.mp4

ffmpeg -i sample.mp4 -c:v libsvtav1 -crf 30 -preset 8 out-av1.mp4

Notice the -crf and -preset pair that always accompanies libx264 — these are the two main levers you must understand.

#-crf: The Quality Lever

CRF (Constant Rate Factor) is the most intuitive way to control quality: smaller value = higher quality = bigger file. Its typical scale:

CRF is constant: FFmpeg adds bitrate in complex scenes and reduces it in calm scenes. Think of it like a cameraman adjusting compression so every moment always looks good — not using one rigid setting for the whole video.

#-preset: The Speed Lever

-preset sets the speed vs efficiency trade-off. Slower presets give better compression at the same file size:

For final results that will be stored, -preset slow is usually worth it. For previews or quick tests, use -preset fast. Again: choose according to the situation, not habit.

#Bitrate: -b:v and Two-Pass

CRF is great for local files because it chases quality. But for streaming with a specific bitrate target — e.g., max 2 Mbps over a limited network — you need bitrate mode:

ffmpeg -i sample.mp4 -c:v libx264 -b:v 2M -maxrate 2.5M -bufsize 5M out-2mbps.mp4

Quality in this mode drops on complex scenes — bitrate is capped, so quality "gives up". For the best results in bitrate mode, use two-pass: the first pass maps the video's complexity, the second pass allocates bitrate optimally:

ffmpeg -y -i sample.mp4 -c:v libx264 -b:v 2M -pass 1 -an -f null /dev/null

ffmpeg -i sample.mp4 -c:v libx264 -b:v 2M -pass 2 out-2pass.mp4

The first pass only analyzes without producing a video file (directed to -f null /dev/null), then the second pass uses the analysis result. The result: best quality for a given bitrate — the industry standard for bandwidth-targeted distribution.

#Video Settings: Resolution, FPS, and GOP

Besides codec and bitrate, these three settings determine the character of the output video:

Combined example: downscale to 480p, keep 30 fps, and GOP 250 (roughly every 8 seconds at 30 fps):

ffmpeg -i sample.mp4 -c:v libx264 -crf 23 -preset slow -vf scale=854:480 -r 30 -g 250 out-480p.mp4

GOP matters for streaming: keyframes are the "entry points" for new viewers — too-long GOPs make seeking or stream joins slower. We cover keyframes and streaming in depth in the HLS episode.

#Quality vs Size: The Eternal Trade-off

All the options above ultimately boil down to one balance: quality vs size. There is no "most correct" setting — it all depends on the goal:

#Conclusion

In episode 4, you've mastered the video encoding map: choosing H.264, H.265/HEVC, VP9, and AV1 codecs, controlling quality through -crf and -preset, understanding the -b:v bitrate mode and two-pass, and setting resolution, fps, and GOP.

Key takeaways:

• H.264 (libx264) for compatibility; HEVC and AV1 for saving space.
• -c:v selects the codec; -crf low = high quality = large file.
• -preset trades speed for efficiency — slow for final results.
• Two-pass for the best quality at a target streaming bitrate.
• Resolution (scale), fps (-r), and GOP (-g) determine the video's character.

In the next episode 5, we enter the world of basic filters — scale, crop, rotate, fps, trim, and setpts — along with how to build filter chains with -vf to process video precisely. See you in episode 5!