What Happens When A Video Is Re-Encoded

Compressed video does not survive being compressed again without loss. Each pass removes information the next pass can never recover, and the damage compounds rather than merely repeating.

The second encode works from a damaged original

Decoding a compressed file produces pixels, but not the original pixels. It produces the encoder's approximation, complete with rounding errors and blocking.

A second encoder treats that approximation as truth. It spends its budget faithfully preserving artifacts alongside genuine picture content, because it has no way to distinguish them.

So generational loss is not simply loss applied twice. The second pass wastes effort on the first pass's mistakes, which leaves less budget for what actually matters.

Artifacts are expensive to encode

Compression works best on smooth, predictable content. Blocking edges and mosquito noise are sharp and irregular, exactly the sort of detail that consumes data.

That inversion is what makes repeated encoding punishing. Damage from earlier passes is harder to compress than the picture it replaced, so quality falls faster with each round.

Noise behaves the same way. Grain, sensor noise and dithering are random by nature, and random detail is unpredictable and therefore costly.

Changing parameters multiplies the damage

Re-encoding at a different resolution requires resampling, which blurs and then re-sharpens edges. The artifact grid from the first encode no longer aligns with the second encoder's blocks.

Frame rate conversion is worse, because frames must be blended or dropped to fit a new cadence. Blended frames contain motion from two moments and compress poorly.

Even keeping every setting identical does not avoid loss, since the block grid may shift and quantization rounds again. Identical settings reduce the harm rather than eliminating it.

Copying streams avoids the problem entirely

Some operations move compressed data into a different container without touching it. Nothing is decoded and nothing is re-encoded, so the picture is bit-for-bit unchanged.

This works only when the destination container accepts the codec already in use. It also constrains cutting to positions where a self-contained frame begins.

When an operation completes almost instantly and produces a file of nearly identical size, that is usually what happened. A long processing bar means the picture was rebuilt.

Workflow order limits how many generations accumulate

Each additional export, upload or format conversion is another generation. A clip shot compressed, edited, exported, uploaded and re-shared has passed through several encoders.

Keeping a high-quality master and producing every delivery version from it holds the count to two. Deriving each new version from the last version compounds indefinitely.

Platforms re-encode uploads regardless, so the practical goal is arriving at the platform with as few prior generations as possible. What you hand over is the best it will ever be.