AWS Elemental MediaLive — Reference Guide

AWS Elemental MediaLive is a real-time, broadcast-grade live video encoding service that runs on AWS Cloud. The service ingests live sources from upstream systems (cameras, contribution encoders, AWS Elemental Link devices, etc.), decodes them, and re-encodes them (decode → encode) into multiple streams at different bitrates/resolutions for delivery to any device via downstream systems (origin/packager, CDN).

This is a standalone reference document dedicated to MediaLive — it does not reference the source code of any project. Its content covers: core concepts, input/output types, channel classes, encoding capabilities, key features, detailed setup and usage instructions, reliability & redundancy, pricing model, limits/quotas, and use cases. The original presentation style was bilingual: explanations in Vietnamese with English technical terms preserved for cross-referencing against official AWS documentation; this version has been fully localized into native technical English.

1. What is MediaLive?

AWS Elemental MediaLive is a real-time, broadcast-grade, cloud-based live video processing service. The service encodes/transcodes live video content — decoding the input source and then re-encoding it — to compress the live source into high-quality streams suitable for delivery to playback devices such as smartphones and set-top boxes.

Purpose: Convert live video content from one format/wrapper to another that playback devices can process, and generate multiple versions (bitrate/resolution) for broadcast and streaming.

Position in the workflow: MediaLive sits at the encode/transcode stage:

Upstream system → Input → MediaLive Channel → Output → Downstream system → CDN → Playback device

• The upstream system holds the source content (camera streaming, contribution encoder, AWS Elemental Link device).
• The downstream system receives the output (an origin/packager such as MediaPackage, then a CDN such as CloudFront, then the playback device).
• MediaLive can be used standalone or integrated with MediaConnect (transport), MediaPackage (packaging), MediaTailor (ad insertion), and MediaStore/S3 (storage).

Benefits: broadcast-quality output; automated provisioning (deployment within minutes); high availability across multiple Availability Zones; no upfront investment; pay-as-you-go pricing.

2. Core Concepts

• Channel: the central component; ingests and transcodes source content from an input, then packages it into an output. A channel can be attached to multiple inputs but ingests from only one input at a time.
• Channel class: STANDARD (2 pipelines) or SINGLE_PIPELINE (1 pipeline).
• Input: describes how the upstream system connects to the channel; defines an endpoint on MediaLive (push) or a source IP (pull).
• Input security group: an allow list of IP ranges, attached to a push input. One input is attached to one security group; one security group can serve multiple inputs.
• Output: exists inside an output group; a collection of encodes.
• Output Group: a collection of outputs within a channel; different types exist for different downstream systems.
• Encode: three types — video, audio, and captions.
• Pipeline: one or two parallel processing paths; two pipelines increase resiliency.
• Schedule: a list of actions performed at specific times (input switch, image overlay, SCTE-35).
• Other concepts: Source content, Upstream/Downstream system, Origin service/Packager, CDN, Playback device.

3. Input Types — Push vs Pull

MediaLive supports the following input types: CDI, HLS, Link, MediaConnect, MediaConnect Router, MP4, RTMP Pull, RTMP Push, RTP, SDI, SMPTE 2110, SRT Caller, SRT Listener, TS file.

Description of each type:

• CDI (push): an uncompressed stream inside a VPC delivered to a fixed endpoint.
• HLS (pull): pulls HLS over HTTP/HTTPS, MediaStore, or S3 (the media must be a transport stream).
• Link (push): the AWS Elemental Link hardware device pushes a TS stream.
• MediaConnect (push): pushes a TS stream from a MediaConnect flow (using the flow ARN).
• MP4 (pull): pulls an MP4 file over HTTP/HTTPS or from S3.
• RTMP Pull / RTMP Push: RTMPS is not supported for input; RTMP Push requires a handshake.
• RTP (push): pushes a TS stream over RTP to a fixed endpoint; supports VPC.
• SDI: via MediaLive Anywhere, from a node within the cluster.
• SMPTE 2110 (pull): receives an SMPTE 2110-compliant stream over RTP; supports 2022-7; NMOS is not supported.
• SRT Caller (pull) / SRT Listener (push): via the SRT protocol, which can be encrypted.
• TS file (pull): pulls a static .ts/.m2ts file.

Note: push inputs and pull inputs have different limits and charges.

4. Output Types (Output Groups)

MediaLive supports the following output groups: Archive, CMAF Ingest, Frame Capture, HLS, MediaConnect Router, MediaPackage, Microsoft Smooth, Multiplex, RTMP, RTMPS, SRT, UDP.

• Archive: sends a TS file to S3 (maximum 1 Archive output group per channel) — used for VOD.
• CMAF Ingest: sends to a packager that supports HTTPS PUT (live).
• Frame Capture: a series of JPEGs sent to S3 (maximum 3 per channel) — VOD.
• HLS (TS or fMP4): HTTP/HTTPS, Akamai CDN, MediaPackage (v1 & v2), MediaStore, S3 — live & VOD.
• MediaConnect Router / MediaPackage (HTTPS WebDAV, live) / Microsoft Smooth (live) / Multiplex (MPTS, live) / RTMP/RTMPS (live) / SRT (output is always encrypted, live) / UDP (live).

Note on S3: bucket names must not use dot notation.

5. Channel Classes & Pipeline Redundancy

• STANDARD (2 pipelines): the input has 2 entry points; the upstream sends 2 identical source streams; the 2 pipelines process identically and distribute to 2 destinations; placed in 2 separate Availability Zones → resiliency. If one pipeline fails, the downstream switches to the other output without interruption; MediaLive restarts the failed pipeline within a few minutes.
• SINGLE_PIPELINE (1 pipeline): 1 destination; if the pipeline fails, output stops. Linked channels can be configured (Primary + Follower, same AWS account, one follower per primary) to enable pipeline locking.

• Pipeline redundancy applies to the entire channel and every input; a standard channel is cheaper than two identical single-pipeline channels in the same Region.

• Pipeline locking (output locking): output is frame-accurate across pipelines; enabled by default on a best-effort basis.
  • Mode: Pipeline locking (default) / Epoch locking.
  • Method: Source timecode (default) / Video alignment.
  • Applies to: HLS (Live), MediaPackage, CMAF Ingest, UDP/SRT (segmented); also Microsoft Smooth when using source timecode.

6. Encoding Capabilities: Codecs, Resolution, ABR, Audio

INPUT codecs/formats:

• Video codec: H.264 (AVC), H.265 (HEVC), MPEG-2.
• Audio: AAC, Dolby Digital, Dolby E, Dolby Digital Plus, MPEG Audio, PCM.

OUTPUT codecs/formats:

• Video codec: AV1, H.264, H.265, MPEG-2. (AV1 output is supported on CMAF Ingest and MediaConnect Router.)
• Audio: AAC, Dolby Digital (AC3), Dolby Digital Plus (EAC3), Dolby Digital Plus with Atmos, MPEG-1 Layer II.

Output resolution:

Resolution	Range	Supported codecs
SD	< 720	any codec
HD	> 720 to ≤ 1080	AV1, AVC, HEVC
UHD/4K	> 1080 to ≤ 2160	AVC, HEVC (no AV1)

Codec by output group:

ABR (Adaptive Bitrate): generates multiple outputs at different resolutions/bitrates within the same output group (e.g., HLS) to serve a range of devices; each rung is one video encode; encodes can be shared or cloned.

7. Key Features

• Statmux / Statistical Multiplexing: allocates bits in real time across multiple channels, combining them into a single MPTS; distributed via satellite/cable/terrestrial; Statmux output always goes through MediaConnect; supports H.264/H.265 up to 1080p; maximum 20 programs/MPTS; maximum 72 Mbps; saves 20–40%.
• Input switching: switch on a schedule or dynamically; 3 types: fixed, immediate, follow.
• Scheduling: input switch, input prepare, image overlay (global/per-output), motion graphics, SCTE-35, ID3 metadata, pause/unpause.
• SCTE-35 ad markers: splice_insert, time_signal, return-to-network.
• Captions: Embedded, DVB-Sub, SCTE-27, Teletext, ARIB, Burn-in, WebVTT, SMPTE-TT, TTML, EBU-TT-D, etc.; OCR; accessibility.
• Image overlays: global & per-output, up to 8 simultaneous layers. Motion graphics overlay: up to 1.
• Advanced audio: Dolby Digital/Plus, Dolby Atmos encode, Dolby E decode; audio normalization (ITU-R BS.1770, CALM Act); audio rendition groups; multiple language tracks.
• Nielsen watermarks, KLV/ID3 metadata.
• MQCS (Media Quality Confidence Score): a 0–100 score for CMAF Ingest to MediaPackage; enabled automatically.
• Other: Dynamic inputs, Input clipping, Low latency outputs, Multiplex/MPTS, 3D LUT, Workflow monitor.
• MediaLive Anywhere (new): encoding on on-premises infrastructure, managed from the cloud; channel is single-pipeline; supports on-prem SDI; separate pricing.

8. Setup & Usage

8.0. Three ways to work with MediaLive

• Workflow wizard — quickly creates a working channel; supports some common input/output combinations; automatically creates resources in other services too (e.g., MediaPackage). Suited to newcomers or for quickly scaffolding a setup.
• Tutorial — a step-by-step console walkthrough (RTP input → 1 HLS output → MediaPackage).
• Designing from scratch — full control over every setting; used for production.

8.1. Prerequisites

• AWS account: create an account + an administrator (this step is not billed). Best practice: do not use the root user; for human users, prefer IAM Identity Center + temporary credentials.
• Tools (optional): AWS Management Console, AWS CLI, AWS SDK (Go/JavaScript/.NET/Node.js/Python/Ruby), AWS Tools for PowerShell.
• IAM Role for MediaLive — “trusted entity role” (VERY IMPORTANT): When a channel is running, MediaLive needs permission to operate on your resources (read/write S3, connect to MediaConnect/MediaPackage, etc.). You must configure MediaLive as a trusted entity: create a role (declaring MediaLive as the trusted entity) attached to policies that state the allowed operations + resources. The role is attached at the channel level (each channel can use a different role). Default role name: MediaLiveAccessRole.
  • Creating the role via the console: under General info → IAM role on the Create channel page. If one already exists → Use existing role → MediaLiveAccessRole. If not → Create role from template → Create IAM role. Choose Remember ARN to reuse it later.
  • ⚠️ Modifying the service role’s permissions can break functionality — only edit it when instructed by AWS.
  • Permissions the role may need: S3, MediaConnect, MediaPackage, Link, Secrets Manager (SRT/CDN passphrases), Systems Manager Parameter Store (passwords), CloudWatch + CloudWatch Logs, EC2 (VPC inputs), CloudFront, CloudFormation, Elemental Inference, Resource Groups, MediaLive Anywhere.
• Region considerations: the MediaPackage channel and the MediaLive channel must be in the SAME Region; a MediaConnect flow input must be in the same Region as MediaLive; feature availability (AV1, MediaLive Anywhere) varies by Region.

8.2. Planning — working backward from output to input

The workflow consists of 3 systems: Upstream system → MediaLive → Downstream system. 6-step checklist:

1. Determine the output groups you need based on the downstream system (format + application protocol; delivery over the internet or a VPC; consider Archive/Frame capture groups).
2. Determine the encoding requirements (video & audio) for each output group.
3. Decide the Channel class (resiliency): Standard (2 pipelines) or Single-pipeline.
4. Assess the upstream system: record the Video codec, Resolution tier (SD/HD/UHD), Maximum input bitrate (used for Input specifications); determine the MediaLive input type, audio/caption format, and language(s).
5. Gather identifiers for the source content (e.g., the audio languages that need to be extracted).
6. Coordinate with the downstream party to obtain a destination for each output group.

8.3. STEP A — Create an Input Security Group (for push inputs)

• Purpose: allow-list the IP ranges (CIDR blocks) permitted to push into MediaLive. One input is attached to 1 security group; one security group can serve multiple inputs; maximum 10 rules (CIDRs) per group.
• Procedure: Console → Input security groups → Create input security group → enter one or more IPv4 CIDR blocks (each block with a subnet mask, separated by commas/newlines) → (optional) Tags → Create.
• Example: 192.0.2.0/24 (any IP starting with 192.0.2); 192.0.2.111/32 (exactly one IP).

8.4. STEP B — Create an Input

Concept: an input describes how the upstream and the MediaLive channel connect. Push input: the upstream pushes to the endpoint MediaLive creates. Pull input: MediaLive pulls from the source IP. An input can be attached to only one channel.

Input class & redundancy: STANDARD_INPUT (2 sources) or SINGLE_INPUT (1 source). A PULL input always needs 2 source URLs (for single-pipeline, fill in source 1 and leave source 2 blank). MediaLive ALWAYS creates 2 endpoints for a push input.

B.1. RTMP Push

Inputs → Create input → Input name; Input type = RTMP (push); Network mode = Public; select/create an Input security group; Channel and input class; Input destinations → enter Application name + Application instance (e.g., live/curling) → Create. Result: MediaLive creates 2 endpoints with port 1935, e.g., 198.51.100.99:1935/live/curling and 192.0.2.18:1935/live/curling.

B.2. RTP Push

Inputs → Create input → Input name; Input type = rtp; create an Input security group (e.g., 203.0.113.111/32) → Create. Result: 2 destinations with port 5000, e.g., rtp://198.51.100.10:5000 and rtp://192.0.2.131:5000.

B.3. MediaConnect

Input type = MediaConnect; Channel and input class; ARN for flow A (+ flow B if applicable); Role ARN (e.g., MediaLiveAccessRole) → Create. MediaLive always creates 2 endpoints; it connects to the flow automatically (2 flows → 2 outputs in 2 AZs).

B.4. HLS Pull

Input type = HLS; Input class; Input sources (standard: 2 URLs; single: 1 URL); if credentials are needed → username + password key (Systems Manager Parameter Store) → Create. The HLS source media must be a transport stream; MediaLive pulls it when the channel starts.

8.5. STEP C — Create a Channel

Console → Channels → Create channel.

C.1. Channel and input details (key info)

Channel name; IAM role (see 8.1); (optional) Channel template; MediaLive Anywhere settings (if an on-prem cluster is used); Channel class (STANDARD 2-pipeline / SINGLE_PIPELINE 1-pipeline; Anywhere requires single-pipeline; if SINGLE_PIPELINE, also configure Linked channel settings: Primary/Follower for pipeline locking); Input specifications (Input codec, Input resolution SD/HD/UHD, Maximum input bitrate — used for billing & resource allocation, NOT for decoding; choose a value ≥ the actual value; choosing too low a value causes higher charges or degraded quality); Output delivery (internet/VPC); Tags.

C.2. Attach inputs

Input attachments → Add → select the created input → Confirm. Multiple inputs can be attached (for input switching).

C.3. Finish the settings for each input

Create Audio selectors + Caption selectors (REQUIRED if you want audio/captions included in the output; the video selector is usually automatic); General settings → Global configuration → Input loss behavior (applies to the whole channel). Leave other fields at their defaults; see the Info link on the console.

8.6. STEP D — Create Output Groups, Outputs & Encodes

Structure: an output group contains outputs; an output contains encodes (video/audio/captions). There must be ≥1 output group.

Output group types: Archive, CMAF Ingest, Frame capture, HLS, MediaConnect Router, MediaPackage, Microsoft Smooth, RTMP, SRT, UDP.

D.1. HLS output group

Output groups → Add → HLS → Confirm. HLS group destination:

• Sending to S3: design the path (main manifest .m3u8, child manifest + nameModifier, media segment + nameModifier + segmentModifier + counter + .ts/.mp4; the counter starts at 00001; nameModifier must be unique per output and should start with a hyphen; protocol s3ssl://; all files go into the same folder).
• Sending to MediaPackage (v1 uses WebDAV; v2 uses Basic PUT), MediaStore, or an HTTP server.

Add outputs for the ABR ladder (HLS outputs → Add output). Each output: link Settings → Output settings + Stream settings.

Encode organization (ABR): if the player needs bundled audio → each output = 1 video + all audio; if the player needs an audio rendition group → each video encode gets its own output + each audio encode gets its own separate output.

D.2. MediaPackage output group

• v1 (HLS, channel ID): Add → MediaPackage → Confirm → enter the MediaPackage channel ID (e.g., curling-live).
• v2 (CMAF ingest): MediaPackage destination → Region + channel group name + channel name + Endpoint ID (ENDPOINT_1 or ENDPOINT_2); Additional destinations for redundancy (standard: up to 2, single: up to 1); CMAF: each output carries only 1 stream type (video/audio/captions are kept separate). MediaPackage & MediaLive must be in the same Region.

D.3. Archive output group

Add → Archive → Confirm. Archive can contain only 1 output (do not Add output); it writes a TS file to S3; maximum 1 Archive group per channel.

D.4. RTMP output group

For social media/RTMP servers; supports H.264 + AAC only.

D.5. Configuring an ENCODE

• scratch / share / clone: scratch = enter every field; share = reuse the same encode as-is; clone = copy then adjust a few fields (best to clone within the same codec).
• Video encode: Codec (H264/H265/AV1); Width/Height (blank = match the input); Rate control; Framerate; Profile (H.264 Baseline/Main/High…; H.265 Main/Main_10BIT; AV1 Bit Depth + Level); Timecode; Color space.
• Rate control mode: H.264/H.265 support QVBR/CBR/VBR (default CBR); AV1 supports QVBR (default)/CBR.
  • QVBR: targets a quality level without exceeding the Max bitrate; recommended when paying for bandwidth. Suggested quality levels: Primary screen 8–10 @ 4–6 Mbps; PC/Tablet 7 @ 1.5–3 Mbps; Smartphone 6 @ 1–1.5 Mbps.
  • VBR: average Bitrate + Max bitrate (not available for AV1).
  • CBR: a fixed Bitrate.
  • Billing: the Bitrate field affects output charges; if left blank → billed based on the Max input bitrate.
• Audio encode: Add audio → Create a new audio description → Codec (AAC/AC3/EAC3/EAC3 Atmos/MPEG); Audio selector name (choose the selector created in C.3); (optional) Remix, Audio normalization (CALM Act –24 LKFS, EBU R-128 –23 LKFS); AAC: Bitrate/CodingMode/Profile/SampleRate.
• Captions encode: in Stream settings (the Captions section); requires the corresponding caption selector.

8.7. STEP E — Save & Start the channel

Save (Create channel): MediaLive validates immediately and shows any errors; you CANNOT save a draft or a channel that still has errors. State: Creating → Ready (Idle).

Start: (1) begin streaming from the upstream to the correct endpoint (a push input must push BEFORE starting the channel); (2) Channels → select the channel → Start (Starting → Running); (3) decide whether to enable thumbnails BEFORE starting (this cannot be turned on afterward); (4) preview via MediaPackage if configured.

Start time: most channels start within ≤3 minutes, up to ~10 minutes at most. Recommendation: start ~2 hours before a major event.

8.8. Operations: Start/Stop/Pause & cost

A Running channel (started, even if a pipeline is paused) is billed even when it is not receiving/producing content; an Idle channel (stopped) is billed at the idle rate.

Start is always manual. Stop: Channels → select → Stop. Pause/unpause one or both pipelines using the Pause/Unpause action in the schedule (the channel remains running).

8.9. Working with the Schedule (Schedule actions)

Action types: Input switch; Input prepare; Global image overlay (applies to all outputs); Per-output image overlay (up to 8 layers); Motion graphics overlay (up to 1); SCTE-35; ID3 metadata; Pause/Unpause.

Three timing modes: Fixed (a UTC start time; overlay/SCTE-35 ≥15s, input switch ≥30s, global image activation ≥60s, ≤14 days; this is wall-clock UTC, not a timecode); Immediate (right away; a standard channel internally adds +10s to switch both pipelines simultaneously); Follow (chained after a reference action; the preceding input must be a file input with Source end behavior = Continue and Follow point = End).

Adding an action via the console: Channel → Schedule tab → List/Timeline view → Create → fill in the fields → Create.

SCTE-35: splice_insert (Splice event id, Duration); time_signal (Add Scte35 descriptors); return-to-network.

8.10. Input switching (Multiple-input channels)

Attach more than 1 input + add an input switch action to the schedule. Static vs. dynamic inputs (dynamic file inputs use $urlPath$). The FIRST input switch of a new channel should be an immediate switch (created before the channel starts). Actions can be added without stopping the channel.

8.11. Automatic Input Failover (AIF — failover pair)

• Purpose: protects against upstream/network failures (different from pipeline redundancy). Supported input types: CDI, Link, MediaConnect, RTMP Push, RTP, SRT caller.
• Detection conditions: Input loss, Black video, Audio silence.
• Setup (RTMP/RTP): create 2 inputs of the same type (different network paths, identical content) → attach both → Input attachments → select the primary → Automatic input failover settings → Enable → select the Secondary input → Input preference (EQUAL_INPUT_PREFERENCE: stay on secondary; PRIMARY_INPUT_PREFERENCE: automatically switch back to primary) → enable Failover conditions.
• MediaConnect: the standard setup needs 4 flows (A, B primary; C, D secondary) placed in the correct AZs; if the flow has source redundancy enabled (failover ~500ms) → set the MediaLive input-loss threshold > 500ms.

8.12. Monitoring & Operations

• CloudWatch metrics: retained for 15 months; grouped as Global/Input/Output/Pipeline locking/MQCS. View them on the MediaLive console (Health tab) or the CloudWatch console (namespace AWS/MediaLive; viewing in CloudWatch is billed). Set alarms based on thresholds.
• Channel alerts: view alerts while the channel is running.
• Thumbnails: 1 JPEG every 2 seconds while running (1 for single-pipeline, 2 for standard); always encrypted; must be enabled BEFORE starting the channel.
• Channel logs: sent to CloudWatch Logs for debugging.
• Workflow monitor (2024): cross-service discovery/visualization/monitoring; creates signal maps + CloudWatch alarm templates + EventBridge rule templates (deployed via CloudFormation); supports MediaConnect/MediaLive/MediaPackage/MediaTailor/S3/CloudFront.

8.13. Updating & Cleanup (Editing / Cleanup)

• Edit channel: only possible while Idle; Channels → Actions → Edit → Update channel. You CANNOT change the channel class via Edit. Tags can be edited at any time.
• Delete: only possible while Idle (Stop first if running) → Delete.
• Cleanup to avoid charges: Stop → Delete channel → Inputs → Delete; also delete the input security group if it is no longer used.

8.14. Automation (CLI / SDK / IaC / Wizard)

• Valid --type values: UDP_PUSH, RTP_PUSH, RTMP_PUSH, RTMP_PULL, URL_PULL, MP4_FILE, MEDIACONNECT, INPUT_DEVICE, AWS_CDI, TS_FILE, SRT_CALLER, SRT_LISTENER, MULTICAST, SMPTE_2110_RECEIVER_GROUP, SDI, MEDIACONNECT_ROUTER. Push inputs use --destinations; PULL uses --sources (2 URLs); VPC uses --vpc (+ --role-arn).
• Key create-channel parameters: --name, --channel-class, --role-arn, --input-specification, --input-attachments, --destinations (MediaPackageSettings: ChannelId for v1, or ChannelGroup+ChannelName+ChannelEndpointId+MediaPackageRegionName for v2), --encoder-settings. Tip: use --generate-cli-skeleton to export a JSON skeleton; use --dry-run to validate.
• SCTE-35 timing in the schedule uses 90kHz ticks (1,350,000 ticks = 15 seconds).

SDK / CloudFormation / CDK: resources AWS::MediaLive::Channel, AWS::MediaLive::Input, AWS::MediaLive::InputSecurityGroup; AIF uses AutomaticInputFailoverSettings.

Workflow wizard: creates a CloudFormation stack containing the input + channel + related resources in other services; it is not used to modify an existing workflow.

8.15. Practical tips & common pitfalls

• S3 buckets must NOT use dot notation; all HLS files go into the same folder; protocol s3ssl://.
• A push input must be attached to an input security group; the endpoint MediaLive creates (RTMP port 1935 + /app/instance; RTP port 5000) is DIFFERENT from the IP in the security group — record it for configuring the upstream.
• MediaLive always creates 2 endpoints for a push input; single-pipeline only uses the first endpoint.
• A standard channel requires 2 identical source feeds.
• The channel must be Idle to Edit/Delete; validation happens after Save; drafts cannot be saved.
• You cannot change the channel class via Edit.
• MediaPackage/MediaConnect must be in the same Region as MediaLive.
• MediaPackage v2/CMAF: each output carries only 1 stream type.
• Archive: 1 output per group, maximum 1 group per channel; Frame capture up to 3 per channel; RTMP supports H.264+AAC only.
• Set the input specification at or above the real value; if a Link device is UHD but the input is HD, set Input resolution on the device to avoid being billed at the UHD rate.
• Enable thumbnails BEFORE starting; start ~2 hours before a major event.
• A running channel is always billed → Stop & Delete when not in use; pull inputs are not billed when idle, but push inputs + the channel are.
• MediaConnect AIF + source redundancy: set the input-loss threshold > 500ms.
• Passwords/credentials: Systems Manager Parameter Store; SRT passphrase & CDN identifier: Secrets Manager.

9. Reliability & Redundancy

(a) Pipeline redundancy (dual pipeline/standard channel): handles failures inside the MediaLive encoder pipeline; 2 pipelines run in separate AZs; the upstream provides 2 source instances (which can come from the same encoder).

(b) Automatic input failover (AIF): handles failures upstream of MediaLive (upstream/network); applied per push-input pair (CDI, Link, MediaConnect, RTMP Push, RTP, SRT caller); one active, one standby; detects input loss/black video/audio silence; the two sources come from different encoders.

Underlying infrastructure: redundant infrastructure across multiple AZs, self-monitoring and self-replacing, elastic scaling.

10. Pricing Model

⚠️ NOTE: figures vary by Region and over time; the examples below are from US East (N. Virginia) and illustrate the cost structure only; check 

https://aws.amazon.com/medialive/pricing/

Three options:

• On-demand: billed hourly; rounded up to the minute, 10-minute minimum.
• Reserved: saves up to 75%, requires a 12-month commitment, suited to >180 hours/month.
• Discounted: 3–5 years, contact AWS.

Total cost: sum the cost of each input + output + add-on feature. A channel is billed while running even when it is not receiving/producing content.

• Input pricing: based on codec/bitrate/resolution (AVC, HEVC, MPEG-2; Link is billed by device type; CDI has its own rate).
• Output pricing: based on codec/bitrate/resolution/frame rate (AVC, HEVC, AV1, Statmux).
• Add-ons (one-time per channel): Advanced audio, Audio normalization, Motion graphics.
• Idle resources: only push inputs (RTP/RTMP Push, MediaConnect, Link) and the channel are billed while idle; pull inputs are not billed when idle.

Example (US East, on-demand):

• Standard channel: 2 HD HEVC inputs + 5 AVC outputs + advanced audio ≈ $3.942/hour.
• Single-pipeline ≈ $2.3652/hour.
• Link HD input $0.50/hour; Link UHD $4.583/hour.

MediaLive Anywhere: priced separately based on the codec & highest output resolution, regardless of the number of inputs/outputs, with no add-on surcharges.

11. Limits, Quotas & Best Practices

Inputs: 0–2 push inputs per channel; maximum 20 inputs per channel; maximum 2 Link inputs; SMPTE 2110 0–1; maximum 32 audio+caption selectors; maximum 3 OCR caption selectors per input.

Outputs: maximum 1 Archive output group; maximum 1 MediaConnect Router output group; a CDI input → maximum 1 UHD output encode; maximum 40 audio encodes.

Other: 3D LUT up to 8 files; image overlay up to 8; motion graphics up to 1; multiplex up to 20 programs; schedule up to 1,500 actions; API 20 TPS steady/40 burst.

Best practices:

• Use a standard channel for 24×7 workloads; combine with AIF for critical inputs.
• Use Reserved pricing for >180 hours/month; stop/delete when not in use to avoid idle charges.
• Enable pipeline locking; use Workflow monitor + CloudWatch.
• Integrate MediaConnect + MediaPackage; use input prepare to reduce latency.
• Follow S3 naming rules.

12. Use Cases

• Live streaming events (sports, concerts, conferences).
• 24×7 linear channels.
• OTT/Multiscreen (ABR, combined with MediaPackage + CloudFront + MediaTailor).
• Broadcast distribution (Statmux via satellite/cable/terrestrial).
• Field video contribution (Link, MediaLive Anywhere).
• VOD preparation (Archive/HLS VOD).
• Hybrid/on-premises (MediaLive Anywhere).

Note: All pricing and quota/limit figures in this document may change by AWS Region and over time. The examples are intended only to illustrate the cost structure and configuration approach; always verify against the official AWS pages listed in the References section before deploying to production.

FAQs