Processing

Transcoding vs passthrough: when to re-encode

Every live channel either passes through untouched or is decoded and encoded again. The right answer differs channel by channel — and getting it right saves both hardware and picture quality.

Key takeaways

  • Passthrough is the cheapest and highest-quality option whenever the source already suits your devices.
  • Transcoding is justified for codec or resolution changes, bitrate reduction, deinterlacing or audio fixes.
  • Often only the audio needs processing — the video can still be copied untouched.
  • Deciding per channel, not per platform, gives the best balance of capacity and quality.

What the two terms mean

Passthrough (also called stream copy or remuxing) takes the compressed audio and video from the source and places them in a new container without decoding them. The bits that leave the server are the same bits that arrived; only the packaging and timing information change.

Transcoding decodes the source to raw pictures and audio samples, optionally processes them — scaling, deinterlacing, loudness or channel adjustments — and encodes them again, possibly in a different codec, resolution or bitrate. Between the two sits a common hybrid: copying video while transcoding only audio.

The cost of transcoding

Passthrough needs very little processing: a single server can typically remux hundreds of channels, with network bandwidth becoming the limit long before the CPU. Transcoding is different. Real-time encoding of one 1080p channel can occupy several CPU cores, and HEVC or 4K multiplies that cost.

Hardware acceleration (GPU or dedicated encoder chips) can raise density dramatically, but it adds hardware, drivers and its own quality trade-offs. Either way, transcoding capacity is something you plan and pay for, while passthrough capacity is almost free.

Transcoding also adds a little latency, because the encoder needs a lookahead buffer to make good decisions, and it introduces another component that can fail and must be monitored.

Software or hardware encoding

Software encoders running on general-purpose CPUs offer the most tuning options and usually the best quality per bit, at the price of higher power and core count. Hardware encoders deliver many more channels per server and lower power per channel, but expose fewer settings and typically need a somewhat higher bitrate for the same perceived quality. A common approach is to use hardware for large volumes of standard channels and software encoding for premium channels where every bit of quality matters. Whatever the choice, measure real throughput under full load — an encoder that occasionally drops below real-time speed will produce stutter that no player can hide.

Quality and generation loss

Lossy codecs throw information away. Each time a stream is decoded and re-encoded, the new encoder works on an already imperfect picture and discards a little more — this is generation loss. Blocking, banding and blurred fine detail tend to accumulate, especially when the new bitrate is not much higher than the original.

A transcoded stream can never be better than its source. If a broadcaster already delivers a well-encoded 1080i H.264 feed at a reasonable bitrate, passing it through will almost always look better than re-encoding it at the same bitrate.

Good reasons to transcode

  • Codec conversion. Source arrives as MPEG-2 or HEVC, but target devices need H.264.
  • Bitrate reduction. A contribution feed at 15–20 Mbps is far too heavy for distribution to many viewers.
  • Resolution changes or creating several renditions for adaptive bitrate delivery.
  • Deinterlacing interlaced broadcast sources for devices and players that display them poorly.
  • Audio fixes. Converting audio to AAC for broad compatibility, repairing broken timestamps or normalising loudness.
  • Track mapping. Selecting which video, audio and subtitle tracks to keep from a multi-track source.

Many "transcoding" needs are actually audio needs. Re-encoding only the audio to AAC while copying video keeps picture quality untouched and costs a fraction of a full transcode.

When passthrough is the better choice

  • The source codec, resolution and bitrate already suit the target devices.
  • You want the lowest possible latency and processing cost.
  • You need to carry many channels on modest hardware.
  • The source is already high quality and any re-encode would only degrade it.

Passthrough still benefits from smart server behaviour. Because players can only start decoding at a keyframe, a streaming server that keeps the most recent keyframe and stream headers ready can send them immediately to a new viewer — giving a near-instant picture on channel change even without re-encoding.

A per-channel decision guide

SituationRecommended
H.264 HD source at distribution bitratePassthrough
Video fine, audio in a format some devices cannot playCopy video, transcode audio to AAC
MPEG-2 or very high-bitrate contribution feedTranscode to H.264 or HEVC
Interlaced source with visible combing on target devicesTranscode with deinterlacing
Multi-track source with unwanted tracksRemap tracks; copy if codecs are suitable
Delivery over the open internet to mixed devicesTranscode to an adaptive bitrate ladder

The key point is that this decision is made channel by channel. A platform that forces one mode for everything either wastes processing power or leaves quality and compatibility problems unsolved.

How SaurusRex handles it: every channel can be passed through or processed individually — with HEVC, track mapping, an all-codecs mode, deinterlacing and AAC audio normalisation — and channels start with a keyframe-aligned burst for an instant picture. Per-channel CPU and memory guards stop one heavy feed from affecting others. Details on the platform page.

Let's talk

See SaurusRex running on your own network.

Tell us about your channels, sites and subscribers. We'll walk you through a live demo and propose an architecture that fits.