Two-pass encoding is the encoding strategy where the encoder watches the entire video once before encoding it. In the first pass, the encoder doesn't produce any output file — it just scans the video and records statistics about how complex each scene is, where the cuts are, what the motion looks like, how much detail lives in each shot. In the second pass, the encoder actually produces the output file, but armed with the full picture of what's coming. The result: bits get allocated where they're really needed, and the final file is meaningfully smaller (or higher quality at the same size) than a single-pass encode would have produced.
The benefit comes from knowing the future. A single-pass encoder, when it reaches minute 20 of a film, has no idea whether the next hour is mostly dialogue or mostly action — it has to guess and adjust on the fly. A two-pass encoder already knows there's a 10-minute battle scene coming at minute 35 and a 5-minute slow conversation at minute 47, so it can hold bitrate in reserve for the battle and spend less on the conversation, balancing out to the target average. The improvement over single-pass VBR is typically 10–20 % bitrate savings at the same VMAF, sometimes more on content with very uneven complexity.
For a product team, two-pass is the default rate-control mode for premium VOD libraries — Netflix, Disney+, Apple TV+ all use it (or more sophisticated multi-pass variants). The cost is doubling encoder time, which matters for big catalogues but is irrelevant per title — you encode once and serve millions of times. Two-pass doesn't make sense for live streaming (no opportunity for a first pass) or for ultra-fast turnaround (TikTok-style content where speed beats final efficiency). For everything in between, two-pass is the standard "do it right" setting.
