server-side ad insertion
Prism X
Server-side ad insertion you can actually run: pre-rolls, VOD that starts instantly, a dashboard humans can read.
Prism X stitches VAST ads into HLS and DASH streams on the server, per viewer, live and on demand. If you know AWS MediaTailor, you know the job: markers go in, personalised manifests come out, ad blockers can’t remove the ads and the player doesn’t stall at ad breaks. Prism X does it in the Noisypeak Cloud or on your own infrastructure (bare metal, VMware, Hyper-V, AWS, Google Cloud, Azure) and bills a fair CPM: only ads that actually reached a viewer count.
It does three things the big-cloud alternative doesn't. It runs pre-rolls, so the first impression is sold too. It has a VOD lazy mode: the asset starts immediately and ad pods are requested while the viewer watches, instead of pre-fetching every ad before the first frame. And it has a session dashboard a human can read, with every viewer, pod and fill rate on one timeline. Prism X is the monetisation stage behind 200+ FAST channels on fastchannels.tv and myfast.ch.
How stitching works
Prism X sits between the origin and the viewer. The source channel arrives as HLS or DASH with ad markers (EXT-X-CUE-OUT/EXT-X-CUE-IN, EXT-X-SCTE35 or EXT-X-DATERANGE). Each viewer gets a session; the session node rewrites the manifest for that viewer, replacing the marked ad break with ad segments. Content segments and ad segments are served as one continuous stream, so the player sees no break and ad blockers have nothing to block.
Ad creatives are transcoded by the integrated UEN2 engine to strictly match the source signal: audio and video codecs, frame size, frame rate, bitrate, codec level and GOP structure. The conformed ad is kept in the ad-content cache (0 to 500 TB) so the same creative is not processed twice. The minimal first segment delivery latency is 700 ms.
- Markers: EXT-X-CUE-OUT/IN · EXT-X-SCTE35 · EXT-X-DATERANGE
- Per-session manifest with per-viewer ad pods
- Ads conformed to source codecs, frame size, fps, bitrate, GOP
- Transcoding on CPU, Intel QSV, NVEnc or Netint
- Output as HLS/DASH, with DRM
Targeting and personalisation
Targeting runs in real time per session. The ad request carries viewer demographics, location, device type and content preferences, so two viewers of the same channel receive different ad pods. The session view shows the session data for each stream: user IP and country, user agent, origin URL and the rendition ladder of the master playlist.
- Demographics, location, device type, content preferences
- Each session identified and traceable in the dashboard
- Prefixes configure many channels over one SSAI config
Live and VOD, with VOD that starts at once
On live channels the ad break is triggered by the marker in the stream and filled as it happens. On VOD, most SSAI systems first resolve every ad pod of the asset (pre-roll and every mid-roll) and only then let playback begin, so the viewer stares at a spinner while the ad server answers. Prism X runs VOD in lazy mode: the asset starts immediately, the pre-roll is served, and later pods are requested while the viewer watches.
Pre-rolls are first-class: the slot before the first frame is a sold slot, on live joins and on VOD, with the same targeting and the same counting as mid-rolls.
- Live: breaks driven by SCTE-35 / CUE / DATERANGE markers
- VOD lazy mode: instant start, pods fetched on the fly
- Pre-rolls on live joins and VOD, counted like any pod
- Same session nodes for live and on-demand
Ad decisioning: VAST, primary and secondary
Prism X speaks VAST and VPAID upstream to the SSP/DSP of your choice; in the Noisypeak FAST pipeline this is typically SpringServe or OpenX. A secondary VAST/VPAID source fills pods the primary leaves empty. The datasheet gives Adbutler as the example backup source, and Adbutler integration is available for evaluation out of the box.
- Upstream protocols: VAST, VPAID
- Primary SSP/DSP plus secondary fill-in source
- Adbutler integration as reference SSP/DSP
A readable dashboard and fair CPM
The SSAI dashboard lists every session with start time, duration, last activity, ad time and average fill rate, and opens a per-session timeline: content in blue, ad pods in red, origin URL, rendition ladder, viewer IP and user agent. When a sales house asks why a pod was empty at 21:14, you answer with one click instead of a log query. Billing follows fair CPM: an ad is counted only when it was actually delivered to the viewer. Every session is logged to ClickHouse; run-time metrics go to Prometheus and Grafana.
- Fair CPM: delivered ads only
- Logs to ClickHouse, metrics to Prometheus/Grafana
- Per-session fill rate, pod count, viewed share
Scaling and deployment
Prism X is delivered as dockerised software with Ansible playbooks for setup and updates. A deployment runs 2 to 50 session processing nodes and 2 to 100 content transcoding nodes on Ubuntu or Amazon Linux, in the Noisypeak Cloud, on bare metal, VMware, Hyper-V, AWS, Google Cloud or Microsoft Azure. The cloud option gives managed scaling; on-premise gives full cost control for operators who want the lowest CPM overhead.
- Docker delivery, Ansible setup and update
- Ubuntu, Amazon Linux
- Noisypeak Cloud, bare metal, VMware, Hyper-V, AWS, GCP, Azure
- Trial from the Noisypeak Cloud with fastchannels.tv playout channels
Features
Does what AWS MediaTailor does, in the Noisypeak Cloud or on your own servers and cloud accounts. Docker and Ansible, 2–50 session nodes, 2–100 transcoders.
Lazy mode: the asset plays at once and ad pods are fetched during playback, so you don't wait for every ad to load before the first frame.
The slot before the first frame is sold too, on live joins and on VOD, targeted and counted like any mid-roll.
Every session, pod and fill rate on one timeline. Finding out why a pod was empty at 21:14 takes one click instead of a log query.
Demographics, location, device and content preferences. Two viewers of the same channel get two different ad mixes.
An ad is counted only when it reached the viewer. ClickHouse logs, Prometheus / Grafana metrics to prove it.
VAST / VPAID to your SSP/DSP (SpringServe, OpenX or any other), with a backup source filling the pods the primary leaves empty.
Integrated UEN2 transcoding conforms every creative to source codec, size, fps, bitrate, level and GOP; cached up to 500 TB. 700 ms to first segment.
EXT-X-CUE-OUT/IN, SCTE-35, DATERANGE in; HLS / DASH with DRM out. Ad blockers can’t remove the ads and the player doesn’t stall at ad breaks.
Prefixes configure hundreds of channels over one SSAI configuration. This is how we run 200+ FAST channels.
Specifications
| Formats | |
|---|---|
| Stream formats | HLS · MPEG-DASH, in and out |
| Output | HLS/DASH with DRM, targeted ads per session |
| Source compliance | Strict: audio/video codecs, frame size, fps, bitrate, codec level, GOP structure |
| Ad markers | |
|---|---|
| Supported markers | EXT-X-CUE-OUT / EXT-X-CUE-IN · EXT-X-SCTE35 · EXT-X-DATERANGE |
| Ad standards | |
|---|---|
| Upstream protocols | VAST · VPAID |
| Secondary source | Yes, pod fill-in from an alternative/backup VAST/VPAID source (e.g. Adbutler) |
| VOD | Optimised: pod requests handled over the duration of playback |
| Counting | Fair CPM: ads actually delivered to end users |
| Scaling | |
|---|---|
| Session processing nodes | 2 to 50 |
| Content transcoding nodes | 2 to 100 |
| Content transcoding | CPU · Intel QSV · NVEnc · Netint |
| Ad-content cache | 0 to 500 TB |
| First segment latency | 700 ms minimal |
| Mass deployment | Prefixes to configure many channels over one SSAI config |
| Integrations | |
|---|---|
| SSP/DSP | Any VAST/VPAID source · SpringServe, OpenX in the FAST pipeline · Adbutler reference integration |
| Logging | Visual logging · ClickHouse |
| Metrics | Prometheus · Grafana dashboards |
| Transcoding engine | Noisypeak UEN2 (Noisypeak Cell) |
| Deployment | |
|---|---|
| Host OS | Ubuntu · Amazon Linux |
| Environments | Noisypeak Cloud · bare metal · VMware · Hyper-V · AWS · Google Cloud · Microsoft Azure |
| Delivery | Dockerised software · Ansible playbook setup and update |
| Evaluation | From the Noisypeak Cloud, with fastchannels.tv playout channels for trial |
Where it is used
Ad breaks from Noisypeak Playout filled per viewer through SpringServe / OpenX. This is how our 200+ channels on 40+ platforms are monetised.
Broadcast channels with SCTE-35 markers, re-monetised per device, location and audience on OTT, pre-roll on join included.
Pre-roll and mid-roll pods on on-demand catalogues, filled during playback instead of before the first frame.
Operators who want cost control run the same stack on their own hardware or cloud account, with fair CPM and no surprise per-request fees.
Talk to us about Prism X
We handle evaluation in the Noisypeak cloud, deployment on premises or in a public cloud, integration and support. One team covers the whole pipeline.