Web TV Software

1. Ingest. The source in the studio or in the field sends the broadcast to the server over RTMP or SRT. For teams connecting from the field, SRT holds up better on unstable mobile connections. The ingest bit rate sets the ceiling on quality; it cannot be raised afterwards.

2. Encoding. The single incoming stream is split into several qualities on the server. That work needs processor power and is the real cost item in video publishing. The panel lets you choose how many qualities are produced: three is usually enough for a news broadcast, while four is preferred for sport.

3. Distribution. Unlike audio, video cannot be distributed from the server alone; as viewer numbers grow, a distribution layer becomes essential. The broadcast is split into segments and served from the point nearest the viewer. The real variable in choosing a package is the simultaneous viewer count multiplied by the average watch time.

4. The player. The player measures the viewer’s connection and chooses the quality itself. Ad breaks are inserted through the player, and subtitle and audio track selection happen there too. The same player works on the site, in the mobile app and when casting to a television.

Three of the four stops — everything but the ingest — are managed on the KEYDAL side. Encoding capacity, storage and distribution are included in the package price; when the broadcast drops, the panel raises an alert and keeps an incident record.

Pre-roll. This is where viewers drop off most. The panel lets you set the length of the pre-roll advertisement and whether it can be skipped, and the report shows separately how many people watched it through and how many left. A short, skippable break usually works better on news clips, and a longer one on archive episodes.

Mid-roll. An advertisement placed mid-way through a long episode is abandoned less often than a pre-roll. The panel lets you place the break by content rather than by time: you mark the natural break points within the episode and the advertisement falls there. Break length and repeat count are recorded and can be compared with the limits in the legislation.

Sponsorship. Programme sponsorship is subject to a separate regime and must be announced. In the panel the sponsored episode is flagged and the sponsor’s name and announcement text are attached to it, with the opening and closing announcement added to the player automatically. We covered the detail in the sponsorship article .

Archive revenue. After a new episode is published, viewing falls but does not go to zero. Viewers arriving from search engines reach archive episodes even months later and watch the advertisement. The panel’s viewing report shows that long tail separately, so you can see which older episodes are still earning.

A programme broadcast live goes on living in three further places once it is recorded: as an episode in the archive, as a clip embedded in a news story on the site, and as a short cut for social media. The panel produces all three uses from a single recording, so you do not need to upload separately for each.

Trimming is done from the panel. You choose the start and end second of the recording and write a title and description; the new episode takes its place in the archive and is split into quality versions automatically. As many cuts as you like can be produced from the same recording, and the source file is left intact.

Archive pages are marked up with video structured data on the search engine side: the duration, publication date, thumbnail and description appear in the search result. The same markup is used for clips embedded in news stories, so stories with video stand out in search results.

Storage cost grows as the archive grows. The panel moves older episodes whose viewing has fallen to a cheaper storage tier; the content is not lost, it simply takes a second or two longer to start. Which episode is on which tier is visible in the panel.

Time to first frame depends on how quickly the player is ready and on the size of the first segment. The panel deliberately keeps the first segment small: the picture opens quickly and the quality rises within a second or two. The viewer is not kept waiting, and the loss of quality is not permanent either.

The player continuously measures the viewer’s connection and adjusts the quality itself. When the connection weakens, the picture drops to a lower quality rather than freezing, which is the single most effective behaviour for keeping a viewer on the stream. Quality changes appear in the report, so regions with frequent drops can be identified.

The page itself is kept light. The player loads without waiting for the rest of the page, cover images are produced in WebP and AVIF, and archive lists load page by page. So the video page opens quickly despite the heavy component running on it.