Amazon IVS Streaming Configuration

Amazon Interactive Video Service (IVS) allows developers to easily deliver low-latency video to viewers worldwide. With Amazon IVS, streamers need to handle only stream production, then send the stream to Amazon IVS. Amazon IVS handles video processing (ingesting and transcoding), delivery, and playback to viewers using the Amazon IVS player.

There is a wealth of solutions for live streaming. Whether you have a studio equipped with multiple cameras, visual switchers, graphics compositing, and a variety of audio mixing equipment, or you plan to start your first stream off a smartphone, you need to deal with some of the same concepts and encoding parameters.

This document describes how to configure video encoders to stream to Amazon IVS. The audience for this document is developers who want to build streaming functionality into their applications.

Note that audio-only input is not supported for IVS low-latency streaming.

Prerequisites

Follow the steps in Getting Started with IVS Low-Latency Streaming, to create a channel and set up streaming. In the process, a channel ARN (Amazon Resource Name) and stream key are assigned, along with URLs for ingesting and playing back a stream. You will need to point your streaming application to the ingest URL.

Before reading this document, you should be familiar with:

• Amazon IVS basics: Read What is IVS Low-Latency Streaming and Getting Started with IVS Low-Latency Streaming

• Amazon IVS API: Understand the IVS Low-Latency Streaming API Reference.

Reducing Latency

Amazon IVS low-latency streaming is compatible with most streaming applications and requires only minor changes to your streaming-application configuration. For the lowest possible latency, you must use the Amazon IVS player; third-party HLS video players are not supported. See the Amazon IVS Player SDK documentation.

To prepare your streaming application for low-latency streaming, do the following. (Note: not all these options are available on every streaming application.)

• On the video encoder, set IDR/Keyframe to a 2-second interval (or 1 second, for even lower end-to-end latency).

  IDR/Keyframe directly affects the timing of stream startup and the latency of related EventBridge events (Stream Start and Recording Start). If IDR/Keyframe is 2 seconds, stream-start latency will be approximately 6-7 seconds. If IDR/Keyframe is 1 second, stream-start latency will be approximately 3-4 seconds. Your video will be available for viewers and auto-recording to Amazon S3 only after the initial stream-start latency period.

  The shorter, 1-second keyframe interval has some QoS tradeoffs. It can cause the Amazon IVS Player’s adaptive bitrate streaming (ABR) to switch resolution more often; the segment size is smaller, so the ABR check happens more often. Buffering may increase due to increased resolution-switching and/or if the viewer’s network cannot download the segments fast enough. Evaluate these tradeoffs when deciding between a 1- or 2-second keyframe interval.

  Avoid setting IDR/Keyframe to values higher than 5 seconds. Not only will the stream-start latency be higher than when using 1 or 2 seconds, but IVS will be unable to guarantee that every segment generated for playback will begin with an IDR/keyframe. Segments not beginning with an IDR/keyframe may result in decode errors or visual distortions when viewers start playback or change renditions.

• If available, set your encoder to zero-latency tuning within an x264 configuration.

• Ensure that buffer size (VBV) does not exceed the average bitrate (kilobits-per-second) of the stream.

Avoid Third-Party Streaming/Forwarding Services

We strongly recommend you do not use third-party service to restream or forward content to Amazon IVS. This will incur extra latency. For low latency, stream directly to Amazon IVS.

Encoder Settings

Stream Ingest: Codecs and Ingest Protocols

Codecs: Amazon IVS supports H.264 for video and AAC (LC) for audio.

Ingest protocols: Amazon IVS supports the most common ingest protocols used in streaming software and hardware: RTMPS (Real-Time Messaging Protocol over a TLS/SSL connection), RTMP, and SRT (Secure Reliable Transport). Amazon IVS streaming through RTMPS requires TLS version 1.2 or later.

RTMPS/RTMP

Your video encoder must connect to Amazon IVS ingest over the RTMPS protocol associated with outbound port 443/TCP. To ensure this, specify an IVS ingest server, which includes the port in the path:

rtmps://<IVS-ingest-server>/<IVS-stream-key>

For example:

rtmps://a1b2c3d4e5f6.global-contribute.live-video.net:443/app/<IVS-stream-key>

IVS channels also can be configured to allow insecure RTMP ingest, though we recommend that you use RTMPS unless you have specific and verified use cases that require RTMP. When streaming RTMP, ensure that the protocol is set to rtmp:// and remove the :443 port. For example:

rtmp://a1b2c3d4e5f6.global-contribute.live-video.net/app/<IVS-stream-key>

SRT

Your video encoder must connect to the ingest endpoint using the SRT protocol at port 9000. To ensure this, specify an ingest endpoint, which includes the port and passphrase in the path:

srt://<ingest-endpoint>:<port>?streamid=<stream-key>&passphrase=<passphrase>

Use a passphrase only if insecure ingest is not enabled for the channel.

For example:

srt://a1b2c3d4e5f6.srt.live-video.net:9000?streamid=sk_us-west-2_abcd1234efgh5678ijkl&passphrase=ZU5A3yrjGAkghUNDr0c5NXBhsPrjlmtcKMNBluh7oImwJQ3ijeyClvMKxlpPcGAMziICJ

To optimize SRT stream performance, see this Haivision blog: How to Configure SRT Settings on Your Video Encoder for Optimal Performance.

Resolution/Bitrate/FPS

The stream’s resolution largely determines its bitrate and frame rate (frames-per-second, or FPS). Use the following guidelines; these are our recommendations. Note the resolutions shown below are landscape orientation (horizontal x vertical), so reverse these for portrait orientation.

	Acceptable Quality (SD) 480p (852x480)	Good Quality (HD) 720p (1280x720)	High Quality (Full HD) 1080p (1920x1080)
Bitrate	Up to 1500 Kbps	Up to 4500 Kbps	Up to 8500 Kbps
FPS	30	30 or 60	30 or 60
Keyframe interval	2 seconds	2 seconds	2 seconds

Bitrate, FPS, and resolution are interrelated. The optimal values depend on circumstances and can be complicated to determine. Our best guidance is to start with the values above and experiment if desired. The goal is clear and smooth motion of video components during streaming and good resolution within the available bandwidth. Increasing frame rate and/or resolution increases overall video quality, but this is necessarily limited by bandwidth.

Amazon IVS supports framerates up to 60 FPS (including European PAL 25 and 50 standard frame rates). The higher the framerate, the better the quality -- as long as there is adequate bitrate bandwidth. Depending on the application, a low framerate can be fine; e.g., for a security camera.

Channel Types

Channel type determines the allowable resolution and bitrate. If you exceed the allowable input resolution or bitrate, the stream probably will disconnect immediately.

There are four channel types: STANDARD, ADVANCED_SD, ADVANCED_HD, and BASIC. When you create a channel, the default type is STANDARD.

There are two types of video processing, transcoding and transmuxing. This is determined by the channel type, whether the channel is configured for multitrack video input, and whether the broadcaster uses a multitrack-enabled client. (Multitrack video is configured with the multitrackInputConfiguration API property of the Channel data type.)

• Video on STANDARD (without multitrack input) and ADVANCED channels is transcoded: multiple qualities are generated from the original input, to automatically give viewers the best experience for their devices and network conditions. Transcoding allows higher playback quality across a range of download speeds. Transcoding is the best option for broadcasters with limited first-mile internet connectivity and/or limited device capabilities (e.g., mobile phones instead of desktop PCs).

• Video on STANDARD (with multitrack input enabled and the broadcaster using a multitrack-enabled client) and BASIC channels is transmuxed: Amazon IVS delivers the original input to viewers. Similar to transcoding, transmuxed multitrack input delivers viewers the best experience for their devices and network conditions.

All transcoded channels have transcode presets, which determine which renditions are produced. Think of these as ABR ladders. They allow you to trade off available download bandwidth and video quality, to optimize the viewing experience.

• STANDARD channels have one, default transcode preset.

• ADVANCED channels have two, selectable transcode presets:

  • Constrained bandwidth delivery uses a lower bitrate than STANDARD for each quality level. Use it if you have low download bandwidth and/or simple video content (e.g., talking heads).

  • Higher bandwidth delivery uses a higher bitrate for each uality level. Use it if you have high download bandwidth and/or complex ideo content (e.g., flashes and quick scene changes). This is the default.

STANDARD Channels

Single-Track Video Input

STANDARD channels are transcoded. The highest video resolution produced is full HD, 1080p. This is the default channel type.

• Transcode presets: There is one, default transcode-preset ladder.

• Audio: For renditions 360p and below, audio is transcoded. For other renditions, original audio is passed through.

Input Resolution and Maximum Bitrate	Ladder Details
1080p60 at 8.5 Mbps	1080p60 at source bitrate 720p60 at 3.4 Mbps 480p30 at 1.4 Mbps 360p30 at 0.63 Mbps 160p30 at 0.23 Mbps
1080p30 at 8.5 Mbps	1080p30 at source bitrate 720p30 at 2.4 Mbps 480p30 at 1.4 Mbps 360p30 at 0.63 Mbps 160p30 at 0.23 Mbps
Less than 1080p60 and greater than 720p60, at 8.5 Mbps	Source passthrough 720p60 at 3.4 Mbps 480p30 at 1.4 Mbps 360p30 at 0.63 Mbps 160p30 at 0.23 Mbps
Less than 1080p30 and greater than 720p30, at 8.5 Mbps	Source passthrough 720p30 at 2.4 Mbps 480p30 at 1.4 Mbps 360p30 at 0.63 Mbps 160p30 at 0.23 Mbps
720p60 at 8.5 Mbps	720p60 at 3.4 Mbps 480p30 at 1.4 Mbps 360p30 at 0.63 Mbps 160p30 at 0.23 Mbps
720p30 at 8.5 Mbps	720p30 at 2.4 Mbps 480p30 at 1.4 Mbps 360p30 at 0.63 Mbps 160p30 at 0.23 Mbps
Less than 720p30/60 and greater than or equal to 480p30/60, at 8.5 Mbps	480p30 at 1.4 Mbps 360p30 at 0.63 Mbps 160p30 at 0.23 Mbps

Multitrack Video Input

STANDARD channels are transmuxed when the input is multitrack video. The highest video resolution produced is limited by the multitrackInputConfiguration.maximumResolution property. The specific renditions are dynamic depending on the broadcaster’s system and environmental requirements.

ADVANCED-HD Channels

ADVANCED-HD channels are transcoded. The highest video resolution produced is HD, 720p.

• Transcode presets: There are two, selectable transcode-preset ladders.

• Audio: For renditions 360p and below, audio is transcoded. For other renditions, original audio is passed through.

Input Resolution and Maximum Bitrate	Ladder Details
720p60 up to 1080p60, at 8.5 Mbps	Transcode preset: higher bandwidth delivery (default): 720p60 at 3 Mbps 480p30 at 1.3 Mbps 360p30 at 0.7 Mbps 160p30 at 0.27 Mbps Audio-only at 0.08 Mbps Transcode preset: constrained bandwidth delivery: 720p60 at 2.2 Mbps 480p30 at 0.8 Mbps 360p30 at 0.4 Mbps 160p30 at 0.22 Mbps Audio-only at 0.08 Mbps
720p30 up to 1080p30, at 8.5 Mbps	Transcode preset: higher bandwidth delivery (default): 720p30 at 2.3 Mbps 480p30 at 1.3 Mbps 360p30 at 0.7 Mbps 160p30 at 0.27 Mbps Audio-only at 0.08 Mbps Transcode preset: constrained bandwidth delivery: 720p30 at 1.9 Mbps 480p30 at 0.8 Mbps 360p30 at 0.4 Mbps 160p30 at 0.22 Mbps Audio-only at 0.08 Mbps
Less than 720p30/60 and greater than 480p30/60, at 8.5 Mbps	Transcode preset: higher bandwidth delivery (default): Source transcoded at 2.3 Mbps 480p30 at 1.3 Mbps 360p30 at 0.7 Mbps 160p30 at 0.27 Mbps Audio-only at 0.08 Mbps Transcode preset: constrained bandwidth delivery: Source transcoded at 1.9 Mbps 480p30 at 0.8 Mbps 360p30 at 0.4 Mbps 160p30 at 0.22 Mbps Audio-only at 0.08 Mbps
480p30/60 at 8.5 Mbps	Transcode preset: higher bandwidth delivery (default): 480p30 at 1.3 Mbps 360p30 at 0.7 Mbps 160p30 at 0.27 Mbps Audio-only at 0.08 Mbps Transcode preset: constrained bandwidth delivery: 480p30 at 0.8 Mbps 360p30 at 0.4 Mbps 160p30 at 0.22 Mbps Audio-only at 0.08 Mbps

ADVANCED-SD Channels

ADVANCED-SD channels are transcoded. Available renditions are capped at input quality, with no up-conversion.

• Transcode presets: There are two, selectable transcode-preset ladders.

• Audio: Audio is transcoded.

Input Resolution and Maximum Bitrate	Ladder Details
480p30/60 up to 1080p30/60, at 8.5 Mbps	Transcode preset: higher bandwidth delivery (default): 480p30 at 1.3 Mbps 360p30 at 0.7 Mbps 160p30 at 0.27 Mbps Audio-only at 0.08 Mbps Transcode preset: constrained bandwidth delivery: 480p30 at 0.8 Mbps 360p30 at 0.4 Mbps 160p30 at 0.22 Mbps Audio-only at 0.08 Mbps

BASIC Channels

BASIC channels are transmuxed. A single rendition is produced.

• Transcode presets: NA

• Audio: Audio is transcoded.

Input Resolution and Maximum Bitrate	Ladder Details
Greater than 480p30/60 and less than or equal to 1080p30/60, at 3.5 Mbps	Source encoding parameters (no ladder)
480p30/60 at 1.5 Mbps	Source encoding parameters (no ladder)

Video Settings

We recommend the following settings. They are available to most H.264 video-encoding software or hardware APIs.

• On the video encoder, set IDR/Keyframe to a 2-second interval (or 1 second, for even lower end-to-end latency).

• H.264 level: Main

• Scene change: Off (preferred)

• Chroma subsample: YUV420P

• CABAC: Preferred

• ColorSpace: BT.709 (recommended for maximum compatibility across HDTVs and computer displays). Amazon IVS video transcoding supports ColorSpace pass-through; advanced users can use other ColorSpace video and full-range video.

Audio Settings

We support the following settings:

• Codec: AAC (LC)

• Bitrate: 96 Kbps to 320 Kbps

• Sample rate: 44.1 Khz or 48 Khz (it is best to match your production audio flow)

• Channels: Maximum 2 - Stereo (1: mono or 2: stereo audio channel support)

Use CBR, Not VBR

Always use CBR (Constant BitRate), not VBR (Variable BitRate), as the rate-control method for encoders. CBR is better suited for the fixed-bandwidth nature of networks, and it produces more predictable, stable video playback for client devices. With a consistent bitrate, it is easy for viewers to select a quality level that their connection can handle over time.

Depending on the complexity of the scene, VBR can result in spikes in bitrate, which can cause frame drops before the video reaches Amazon IVS and/or buffering in client players.

We strongly recommend you only use CBR. If you use VBR, your streams will be more subject to buffering and playback that is not smooth.

Use Progressive Signals

Use progressive signal flows; avoid any interlaced video in production flow and/or encoding. Progressive stream signals yield much better playback quality displaying a whole frame at a time, avoiding any motion artifacting that is produced when displaying an interlaced signal.

Network Requirements

You must have a stable internet connection that can maintain an adequate, constant upload stream. An unstable internet connection could result in stream stuttering and lagging for your viewers.

Use wired connections. WiFi and LTE connections can be spotty or suffer from interference or latency due to bad QoS/packet-queue prioritization. Whenever possible, rely on a hardwired connection for streams.

Plan to allocate 50% more bandwidth than the minimum required. The overhead is added to compensate for the bitrate fluctuations in encoding of a video bitstream.

Use a dedicated Internet VLAN to encoding machines. Keeping the encoder on a separate network prevents potentially disruptive effects, including: pollution by traffic, bandwidth bottlenecks and adverse security factors.

Closed Captioning

IVS supports closed captioning. As a streamer, if you want to offer captions to your audience, you must transmit caption data in an accepted format, either embedded in your stream or alongside your stream, through your video encoder.

Amazon IVS accepts captions in line 21 CEA-708/EIA-608 format (also referred to as 608 over 708). You can transmit captions using one of the following methods:

• CEA-708/EIA-608 embedded in the video elementary stream, as described in ATSC A/72 (SEI user_data). This format is common among television broadcast encoders.

• CEA-708/EIA-608 transmitted via RTMPS onCaptionInfo script/AMF0 tag. This format is common among Internet broadcast encoders and media servers like Elemental Technologies and Wowza. The Amazon IVS Player SDKs support one language; they do not support multi-track captions playback.

Note: The Amazon IVS Player SDKs support caption data only in the CC1 NTSC field 1. They do not support multi-track captions playback.

When transmitting via RTMPS, the payload must contain an ECMA array with two element pairs:

• A string named type that contains the characters 708.

• A string named data that contains a base64-encoded CEA-708/EIA-608 payload.

For example:

 
00000000  12 00 00 69 00 00 00 00  00 00 00 02 00 0d 6f 6e  |...i..........on|
00000010  43 61 70 74 69 6f 6e 49  6e 66 6f 08 00 00 00 02  |CaptionInfo.....|
00000020  00 04 74 79 70 65 02 00  03 37 30 38 00 04 64 61  |..type...708..da|
00000030  74 61 02 00 3c 74 51 41  78 52 30 45 35 4e 41 4e  |ta..<tQAxR0E5NAN|
00000040  4c 41 50 79 55 72 76 79  55 49 50 79 52 51 50 7a  |LAPyUrvyUIPyRQPz|
00000050  49 35 66 7a 73 37 50 7a  76 4c 50 77 67 56 50 7a  |I5fzs7PzvLPwgVPz|
00000060  33 36 66 7a 30 34 2f 78  6f 67 50 79 55 4c 2f 38  |36fz04/xogPyUL/8|
00000070  3d 00 00 09 00 00 00 74                           |=......t|

If you use the Elemental video encoder, set it up as follows:

• Set caption embed to “capture 608 Field 1.”

• Embed captions with onCaptionInfo as the RTMPS tag in the output group.

For more information, see the blog post Adding Closed Captions to an Amazon IVS Live Stream.

Stream with FFmpeg

FFmpeg is a free, open-source project comprising a vast suite of software libraries for handling video, audio, and other multimedia files and streams. It can be used with many operating systems and devices.

See the FFmpeg website for installation and other information about FFmpeg. Use the latest static build (do not compile).

After installing, choose an audio/video input source for FFmpeg. You can look up what is available, as follows:

ffmpeg -list_devices true -f dshow -i dummy.

For more information, see here. Depending on what is available and what capture method is targeted, you should be able to capture the video/audio (embedded) directly from your selected device and encode the signals with FFmpeg. For example:

• Webcam — To capture output from the Logitech C920 webcam:

  ffmpeg -f dshow -video_size 1920x1080 -framerate 30 -i video="HD Pro Webcam C920":audio="Microphone (HD Pro Webcam C920)" -c:v libx264 -b:v 6000K -maxrate 6000K -pix_fmt yuv420p -r 30 -s 1920x1080 -profile:v main -preset veryfast -g 120 -x264opts "nal-hrd=cbr:no-scenecut” -acodec aac -ab 160k -ar 44100 -f flv rtmps://<IVS-ingest-server>/<IVS-stream-key>

• Video file — FFmpeg works with many video-file formats and capture cards. Here is an example of streaming based on a MP4 input:

  ffmpeg -re -i input.mp4 -c:v libx264 -b:v 6000K -maxrate 6000K -pix_fmt yuv420p -s 1920x1080 -profile:v main -preset veryfast -force_key_frames expr:gte(t,n_forced*2) -x264opts "nal-hrd=cbr:no-scenecut” -acodec aac -ab 160k -ar 44100 -f flv rtmps://<IVS-ingest-server>/app/<IVS-stream-key>

For more information about what to enter for <IVS-ingest-server> and <IVS-stream-key>, see the information about setting up live-streaming software in Getting Started with IVS Low-Latency Streaming. For example:

• Ingest server: rtmps://jds34ksdg3las.global-contribute.live-video.net/app/

• Stream key: sk_us-west-2_abcd1234efgh5678ijkl

Stream Takeover

Stream takeover allows a user to replace an ongoing stream on a channel they own with a new stream. During this process, the previous stream never disconnects, it is simply replaced by the new stream. This allows users to seamlessly connect to a new stream without having to wait until the ongoing stream disconnects entirely.

The stream-takeover process extends an ongoing stream session but does not initiate a new one. This maintains stream continuity without requiring viewers to refresh the player, though they may experience a brief buffering state. There are no discontinuities in recordings of stream sessions during which a stream takeover occurs.

To initiate a stream takeover, append the priority URL parameter to the user's stream key. The stream key becomes <IVS-stream-key>?priority=<priority>, where <priority> is a positive integer between 1 and 2,147,483,647.

The URI syntax for using stream takeover with the RTMPS protocol is:

rtmps://<uri>/<streamkey>?priority=N

For SRT ingest, the URI syntax for stream takeover is:

srt://<uri>?streamid=#!::u=<streamkey>,priority=N&passphrase=foobar

A takeover succeeds if the priority integer provided for the new stream is greater than the priority integer for the ongoing stream, or if no previous priority integer was set. Also, the old and new stream must share the same resolution, video codec, audio codec, and number of tracks.

By default, up to 100 takeovers can be done in a single stream, as long as a greater priority integer is used for each successive takeover. The maximum number of stream takeovers is adjustable per AWS account (see Service Quotas). Once the stream is over, the channel retains no memory of previous priority integers or how many takeovers were done, so any priority integer can be reused in future streams. Also, if encoder settings were changed for stream takeover, the stream session retains no memory of previous encoder settings, displaying only the latest settings.

If the auto-reconnect feature is enabled, the IVS mobile broadcast SDKs use stream takeover to automatically reconnect when a broadcaster switches networks (for example, from WiFi to cellular). To enable auto-reconnect:

• On iOS, set config.autoReconnect.enabled = true on your IVSBroadcastConfiguration object.

• On Android, set config.autoReconnect.setEnabled(true) on your BroadcastConfiguration object.

Considerations for Using Auto-Reconnect and Stream Takeover Together

When mobile broadcast SDK customers enable auto-reconnect as described above, the ongoing streamer (Broadcaster A) will try to reconnect up to 5 times following a network disruption, starting with priority=1 and incrementing the priority with each reconnect attempt. This process allows the broadcast to automatically recover on unstable networks by gradually increasing the priority with each successful reconnect.

However, due to the incremental nature of the auto-reconnect behavior, it becomes challenging for another broadcaster (Broadcaster B) to successfully use stream takeover when the original broadcaster is using auto-reconnect. The priority value required to ensure a successful takeover will be unpredictable, as reconnect attempts by Broadcaster A increment the priority value with each retry over the stream duration.

Note: We do not recommend using stream takeover to override a stream from the mobile broadcast SDK when auto-reconnect is enabled, as you will need to manage or keep a record of the priority required for takeover. While setting a large priority value may work initially, it could create challenges if another takeover is needed later. We recommend maintaining auto-reconnect for Broadcaster A during network instability and stream takeover by Broadcaster B as distinct use cases.

Stream with the Amazon IVS Broadcast SDK

The Amazon IVS broadcast SDK is for developers who are building Android, iOS, or Web applications with Amazon IVS. See the broadcast SDK documentation in the Amazon IVS User Guide, starting here. There are subpages with guides for Android, iOS, and Web streaming. The broadcast SDKs enable you to customize bitrate, frame rate, and resolution.

Testing the Stream

Always verify that your stream works.

Navigate to the video stream in the Amazon IVS console, to watch what is being streamed and manage the live stream.