Every adult video platform makes the same three decisions - what to transcode to, how to deliver it, and how to stop the files being taken - and the order matters, because the ladder determines the bandwidth bill and the bandwidth bill determines whether the business works.
The ladder is a cost decision disguised as a quality one
A transcoding ladder is the set of renditions each upload produces. Too few and mobile viewers buffer; too many and storage and processing multiply for renditions nobody selects.
| Rendition | Typical bitrate | Who receives it |
|---|
| 360p | 0.6-0.8 Mbps | Poor cellular; the floor that prevents abandonment |
| 480p | 1.0-1.5 Mbps | Average mobile - usually the most-served rung |
| 720p | 2.5-3.5 Mbps | Good mobile and most desktop |
| 1080p | 5-6 Mbps | Desktop on fixed lines |
| 4K | 15-20 Mbps | Rarely justified; verify demand before enabling |
Start with four rungs and read the analytics after a month. Most adult platforms discover the 480p and 720p rungs carry the majority of watch time, which makes the 4K rung an expensive vanity feature.
Segment sizing, and the trade nobody explains
Adaptive streaming cuts video into segments. Short segments - two seconds - switch quality quickly and start fast, at the cost of more requests. Long segments - ten seconds - are more efficient and slower to adapt, which on a fluctuating mobile connection means visible buffering.
Four to six seconds is the workable middle for this category, where the audience is mostly mobile and the first ten seconds decide whether they stay. Measure time-to-first-frame rather than throughput: a platform that starts in under two seconds keeps viewers a competitor with better peak quality loses.
Signed URLs, and what they actually protect
Without signing, the video file is a public link the moment the page renders - anyone can read it from the network tab and hotlink it. Signed URLs bind the link to a short expiry and usually to an address or session, so a copied link stops working.
Be honest about the limit: signing prevents casual hotlinking and lazy scraping. It does not prevent someone who can play the video from recording it. Treat it as a bandwidth-theft control rather than as content protection, and set the expiry short - minutes, not hours - with renewal handled by the player.
The egress arithmetic
This is the calculation that decides viability, and it is arithmetic rather than opinion. One thousand viewers, twelve minutes average watch, 1.5 Mbps average rendition, works out to roughly 135 GB a day - about four terabytes a month. At a rate that looks trivial per gigabyte, that is a serious monthly line, and it scales linearly with success.
Three levers move it. The ladder, because serving 720p where 480p suffices costs double for a difference most mobile viewers cannot see. The cache hit ratio, since every miss is paid twice - origin to edge and edge to viewer. And preview handling: auto-playing previews on a listing page can consume more egress than the actual views, and they are the first thing to make click-to-play when the bill arrives.
Origin protection is not optional
If the origin is reachable directly, someone will find it and pull video around the CDN - at your cost and without your cache. Lock the origin to the CDN's address ranges, require a shared secret on origin requests, and never let the origin hostname appear in markup or playlists.
This is the most common expensive mistake we see on adult video platforms: the CDN is configured correctly, the origin is wide open, and the first large bill is the discovery mechanism.
Storage grows faster than anyone plans
Every upload becomes its renditions plus thumbnails plus previews - typically two to three times the source. A library that grows steadily produces a storage line that outpaces the traffic line within a year.
Two policies keep it sane. Tier cold content to cheaper storage after a defined period, accepting slower first-play on old material. And delete source files once renditions are verified, or accept that you are paying to store masters you will never serve. Decide which before the library is large, because retrofitting a tiering policy across a hundred thousand assets is its own project.
Codec choice, and why it is not free
H.264 plays everywhere and is the safe default. Newer codecs cut bandwidth substantially for the same quality, which is tempting when egress dominates the bill - but they cost more to encode, and support is uneven on older devices, which in this category are a real share of the audience.
| Codec | Bandwidth vs H.264 | Trade |
|---|
| H.264 | Baseline | Universal playback, highest egress |
| H.265 / HEVC | Roughly 30-40% less | Licensing questions, patchy browser support |
| AV1 | Roughly 30-50% less | Expensive to encode, newer devices only |
The pragmatic pattern at scale: keep an H.264 ladder for compatibility and add a second ladder in a modern codec for the devices that support it. That costs storage and encoding to save egress, which is the right trade only once egress is genuinely the largest line - measure before committing.
Live streaming changes the shape entirely
Recorded video is a caching problem; live is a distribution problem with no second chances. Ingest has to be resilient because a dropped stream is lost rather than retried, latency budgets are measured in seconds because interaction depends on it, and the cache offers far less help because everyone is requesting the same segment at the same moment.
If the roadmap includes live, choose the delivery stack for live from the start. Retrofitting live onto a stack designed for recorded playback normally means replacing the ingest and the player, which is most of the work twice.
Common questions
How many renditions should the ladder have?
Four to start - 360p, 480p, 720p, 1080p - then read a month of analytics. Most adult platforms find 480p and 720p carry the majority of watch time, which makes higher rungs expensive to keep.
What segment length is right?
Four to six seconds for a mostly-mobile audience. Shorter starts faster and adapts quicker at the cost of more requests; longer is more efficient and buffers visibly when the connection fluctuates.
Do signed URLs stop piracy?
No. They stop hotlinking and casual scraping, which is a bandwidth control rather than content protection. Anyone who can play a video can record it; set short expiries and treat signing as cost defence.
What dominates the monthly bill?
Egress. A thousand viewers watching twelve minutes at 1.5 Mbps is roughly four terabytes a month. The ladder, the cache hit ratio and auto-playing previews are the three levers that move it.
Related reading