In addition to key technical specifications, content creators need to know how to prepare content for each platform, what the common pitfalls are, and how to ensure quality control and compliance.

Content Duration, Top And Tail

Professional broadcast delivery sometimes requires calibration slates, countdown clocks, black frames and such like to help the play-out scheduler. User submitted video should not have anything other than the program content. The scale of user-contributed content uploading is such that human intervention to trim content is just not feasible. Aside from some AI inspection to identify prohibited material, the footage is transcoded and mounted for presentation as is.

The following should be omitted:

• Advertisements.
• Bars and tone.
• Test patterns.
• Production slates.
• Textless material.
• Leading and trailing black frames.
• Black interstitial frames unless included for creative reasons.

If black frames are included as part of the narrative, they should not exceed two seconds duration.

Maximum permitted lengths vary. Some platforms allow video durations of up to 240 minutes. 

Recommendation:

• Keep your footage short to improve user engagement.

Letterboxing Or Pillar Boxing

If you are presenting widescreen content in a 4:3 aspect ratio or vice-versa, the black space filling areas can remain. The transcoding process identifies these areas and should automatically crop the image rectangle to remove them. The final encoding will only include the active pixel area.

Ensure that any blank areas at the sides or top and bottom are symmetrical or the cropping may remove some active pixels.

Recommendation:

• Do the cropping yourself and upload without the black bars.

Vertical Interval Timecode (VITC) Lines

You should crop the visible content rectangle to remove any Vertical Interval Timecode (VITC) and technical data. This includes traditional line 19 or 21 subtitle data and Teletext code. You may also find Macrovision copy protection from old analog video tapes embedded in the same area. That should all be removed.

In ST 2110 coded video, this is called VANC data and should be removed and encoded in a separate auxiliary stream. There may also be proprietary data embedded in these lines which you would rather not share.

Recommendation:

• Crop any non-active picture area lines.
• Remove VANC data.
• Also remove the HANC data in the horizontal blanking period if it is present.

Aspect Ratios

Display aspect ratios include these values. Some platforms will only accept a sub-set of them:

Some oddball aspect ratios are rejected by at least one major platform although Facebook and Instagram are OK with them. Do not ship content with these aspect ratios unless you know the platform accepts it:

You should be able to fit your content into one of the preferred aspect ratios for the target platform. A small amount of scaling in the horizontal axis to accurately fit might be OK if it is not visible. Otherwise introduce a tiny amount of cropping if necessary.

A pixel aspect ratio of 1:1 is preferred although other pixel aspect ratios allow for anamorphic scaling. Anamorphic scaling should not be applied unless the content was prepared with that in mind otherwise image distortions will occur (e.g. circles become ovals).

YouTube suggests that some content may be available in a widescreen and also a pan-and-scan SD format. They would like both versions uploaded together.

Recommendation:

• Most TV receivers these days have 16:9 screens. This is a good format to use.
• Use 4:3 if your content is formatted like that.
• Choose others if they suit your viewing user’s device display format better.

Picture Sizes

Standard definition NSTC or PAL are different sizes. Both are usually accepted. These might also be interlaced.

High definition 2K format (1920 x 1080) is probably the most popular format although the larger 4K UHD size is relevant for high quality productions and movies.

Be careful to choose an appropriate picture size. If you upload a music track with a stationary image, picture sizes larger than enhanced SD (720 lines progressive) may just be wasting bitrate.

Upscaling a lower resolution source video to a higher resolution is not acceptable. It wastes bandwidth for no visible improvement in picture quality. YouTube suggests that 1920 x 1080 is optimal for most applications.

Some platforms may mandate a minimum picture size for SD and HD content.

Recommendation:

• HD 1920 x 1080 is a good general-purpose size.

Picture Orientation

Videos intended to be viewed on vertically oriented mobile phones need to be formatted in portrait mode. Landscape mode is more useful for videos intended for repurposing to other uses. Some platforms such as Instagram and Facebook support multiple orientations for different purposes. They also allow square formats that will look the same regardless of the device orientation.

Recommendation:

• Landscape viewing is more comfortable for the user.

Frame Rates

These should be constant throughout the program. Players are resilient to frame rate changes if they are properly encoded into the compressed stream but this presents problems when transcoding into the different alternative formats for presentation.

The platform will likely reject your footage if variable frame rates are detected. Your editing process should be able to fix this when compiling the program and rendering out the final version. Mixing frame rates in your content or retiming footage to speed up or slow down the action requires inter-frame interpolation which can reduce the image quality if the rendering tools are not robust enough.

These are the most popular frame rates:

Some platforms may only accept a sub-set of these frame rates.

Recommendation:

• Film based content should be delivered at 24 or 48 fps to avoid the need for frame interpolation.
• Choosing 25, 30 or 60 fps is good for digital content.
• The fractional rates are a historical legacy from analog interlaced TV. They could be transcoded to retime them at the editing stage and are not recommended for deployment.
• Higher frame rates allow for better motion representation. Sporting events look very good at 100 or 120 fps.

Scan Type

Progressive scanning is preferred. If the content has originated from a Standard Definition (SD) original taken from an archive library, it may be interlaced.  This is often also supported but the transcoding results may not be perfect. High resolution (4K) content can only be delivered with a progressive scan type.

Telecine converted footage with pulldown is usually not accepted. Converting this to a progressive scan is difficult and often requires human intervention. It should be inverse-telecine corrected back to a native frame rate before uploading. Pulldown on interlaced content is especially hard to deal with.

The field order for interlaced scanning is sometimes described as field dominance. Getting it wrong causes unpleasant jitter artefacts. Make sure the field order is correctly specified in the metadata. Field order may be described as top or bottom first or odd/even fields.

Macro-Block-Adaptive Frame/Field (MBAFF) coding can be used in some platforms if the transcoding engine can decode it. This can significantly reduce the bit rate required for H.264 AVC coded content.

Recommendation:

• Progressive only.
• Remove telecine pulldown carefully.
• Interlaced content should be carefully transcoded to remove the combing artefacts.

Picture Content

It is permitted to include advertisements if they are genuinely part of the narrative. Product placement and sponsorship is permitted, but commercial ad breaks are not.

It has become common practice for studios and networks to burn in a small logo artwork in the corner of the image. These are not permitted. There should not be any bugs, DOGs or watermarks.

Recommendation:

• Only deliver the program content because the platform may want to add interstitials.
• Never add your own logo graphics.

High Dynamic Range (HDR)

If you deliver High Dynamic Range content, then the OpenHDR format is preferred. This is sometimes described as HDR-10.

You may also need to deliver a Standard Dynamic Range (SDR) version of the same footage. Both SDR and HDR versions must be identically conformed to the same duration. The only permitted difference is the color space used to implement HDR. Although HDR can be used with many formats, it may only apply to 4K image sizes on some platforms. Amazon specified the following constraints for HDR 4K:

• Rec. 2020 Container or DCI P3 Primaries.
• D65 White Point.
• Color range should be Legal YUV.

Non-HDR content should use the Rec 709 color space. The chroma sub-sampling should be at least 4:2:2 format with a 10-bit sample size. Higher specifications may also be acceptable.

Recommendation:

• HDR is good for long-form movie content but less necessary for short documentary style material.

Coding Formats

These video codec and container combinations are popular:

• ProRes 422 HQ profile in a QuickTime container.
• H.264 High profile in an MPEG-4 file or MPEG-2 transport stream.
• MPEG-2 Main or High profile in an MPEG file or transport stream.
• H.265 (HEVC) in an MPEG-4 file is not yet widely used but may become more popular. Some platforms do support it.

Platforms may mandate a minimum bit rate to ensure sufficient quality. Platforms also recommend various profiles, levels and keyframe intervals that they prefer.

Recommendation:

• H.264 (AVC) is often specified. Make sure you code this to a high enough bit rate and use the High profile for the best quality.
• ProRes 422 HQ is also popular but you need the right software to produce it.

Audio Coding

Audio may be embedded into a video file. In that case, at least one platform dictates that only a single language shall be used for all embedded tracks. This implies that multiple foreign language tracks cannot be embedded. Commentary tracks should also be separated out to another file.

The audio must all be finished program sound tracks. Separate tracks for music and effects are not generally acceptable. There is no guaranteed standard way of arranging these and any conventions are unlikely to be rigorously applied. Therefore, the transcoding cannot anticipate the track organization. The metadata should describe the track assignments.

Multi-track audio may be provided in separate files to support different languages or accessibility support. These must all be conformed to the same single video program.

Any related audio and video files must be precisely the same duration.

All audio tracks in a given file must be encoded to the same bit depth. A file containing a 24-bit surround-sound track must also format a stereo mix-down as 24-bits. Files with different bit depths may be rejected.

Surround channels must be labelled and arranged correctly. Platforms should conform to the same conventions but they may not. Their individual specifications should be checked to ensure your audio tracks are correctly assigned.

These audio codecs are popular when combined with specific video formats:

Audio is sampled at 44.1 kHz for CDs but for video, a 48 kHz sample rate is used. Make sure your audio export settings output 48 kHz as 44.1 will likely be rejected. A bit depth of 16- or 24-bit samples is preferred.

Multi-channel assignments follow the usual conventions although some platforms may assign channels differently:

Sometimes, you want to upload sound only without a picture. YouTube prefers these audio formats when there is no accompanying video.

Recommendation:

• More channels are fine.
• Lossy formats might be OK but can introduce artefacts when the sample rate is converted.
• Deliver at 48 kHz 16-bit as a minimum but 24-bit is better.
• Describe the channel mapping in the metadata.

Subtitling And Other Text

Do not include any burned in subtitles. These need to be delivered separately. A VTT file containing timed-text is probably the most versatile across the different platforms.

Do not include references to external website with onscreen URLs.

Forced Narrative subtitles may or may not be allowed depending on the platform.

If you provide the VTT tracks for timed-text, the duration of the file must conform to the video and audio. All files must have the same duration.

Recommendation:

• Provide multiple VTT timed-text tracks in a separate file.
• Split forced narrative subtitles from dialog and descriptive subtitles.

Metadata

Metadata may be supplied separately or embedded into the files depending on the platform requirements. It is probably more useful when delivered in a separate XML or JSON formatted file.

It should correctly describe the following attributes:

• Scan type (progressive or interlaced and MBAFF).
• Field order (top or bottom first for interlaced).
• Display aspect ratio.
• Audio channel mapping.
• Anamorphic stretching from 4:3 or 16:9 aspect ratio must also be flagged.

Recommendation:

• Provide as much metadata as possible. It will be useful somewhere in the pipeline and even more helpful in the user interface dashboard.

Container Types

Footage may be delivered in a variety of container types and video codecs. These containers are the most popular. Your platform may prefer delivery in a specific subset of these:

Interestingly LinkedIn supports a larger variety of file formats but specifically excludes the formats that are preferred by other platforms.

Gotchas And Other Things That Go Bump In The Night

There are some very subtle problems with media production. The content may look fine and play back perfectly well on your editing system. Later on though, the transcoding process may fail to process the file for a variety of reasons, including:

• Playback scaling in QuickTime.
• Inactive audio tracks.
• Unsupported containers.
• Bad file names.

Playback Scaling In QuickTime

The QuickTime player is a useful tool for making quick top and tail trimming edits. Some versions of the player also allow you to specify a zoom factor in the display size parameter. This causes the player to scale the video on playback. This value should be set to the same value as the native size of the footage for a 1:1 scaling before saving or exporting the footage. This scaling value is stored in the metadata embedded in the video file and can upset the transcoding processors.

Inactive Audio Tracks

QuickTime player can deactivate audio files to silence them in .mov and .mp4 files. If any audio tracks are disabled, your content might be rejected. Ensure all audio tracks that are needed are enabled and silenced tracks are removed before saving or exporting the media.

Unsupported Containers

These ingest formats may be problematic. They are rejected by at least one platform although others may be OK. Check the target platform specs:

Bad File Names

Do not use special characters in upload file names. These are either poor choices or completely rejected; they can completely crash a target system in some contexts or be misinterpreted as dangerous wildcard matches. This is especially dangerous if automation processes are running with elevated privileges. The most dangerous ones are listed as ‘shell-metacharacters’ in the command line shell documentation:

Other Common Problems

These are some more common problems with uploaded content that trigger a rejection:

• Unsupported display aspect ratios.
• Picture sizes smaller than SD, HD, or UHD specifications.
• Too many audio channels provided.
• Not enough audio channels provided.
• Different video and audio durations.
• Unsupported frame rates.
• Long silent periods in the audio tracks.
• Long durations of black frames in the video.
• Incorrect packaging of coded audio or video in containers.
• Unsupported container types used.
• Incorrect coding profile and level used.
• Wrong keyframe interval when coding.
• File names contain illegal characters.
• Video contains illegal color values that are out of gamut.
• Duration is too long or too short.
• Uploaded files are too big or too small.

Quality Control

There are various third party checking services and software applications that can verify your content before uploading it. This is functionality that is present in most of the leading commercial monitoring & compliance applications. This can warn you early in the production process and saves needing to rework the content.

Regulatory Frameworks

In addition to technical checkboxes, there are also important philosophical matters related to user-generated content for these channels:

• Taste and decency issues.
• Intellectual Property and Copyright.
• Truth vs. faked content.

These are less problematic for professional broadcast and streaming organizations because of the regulatory frameworks and supervision. Ofcom for example provides guidance to avoid distributing harmful content.

This is not the case with user-uploaded material which is regulated with poorly managed moderation and takedown processes. These are haphazardly applied long after harmful content has been deployed and already viewed many times by the public.

Matters Concerning Taste & Decency

A major downside of user-generated content is also its biggest strength. Literally anyone can publish media for consumption by anyone else.

Because there are few constraints, a lot of content many consider dubious, illegal, inflammatory and quite dangerous gets uploaded along with all that is good, edifying and wholesome. We want to be entertained and we want the online services to be a safe place for young people to visit. There is other material that many people don’t want to see and feel should never be presented in a public forum.

Defending our rights to free speech and opposing censorship does not excuse the publishing of harmful material. Distinguishing between the good and bad content and taking down the harmful material immediately is the User Generated Content industry’s biggest challenge.

Inflammatory content drives up traffic levels and increases advertising revenue. Therefore, self-regulation by the social media companies does not seem to work very well.

This means that user generated content implies a certain level of responsibility. Some members of the public show a significant lack of restraint in what they post. It should be possible to trace uploaded media back to an individual to hold them accountable if it is found to be harmful. This might deter but it won’t stop them all.

Regulation cannot solve this on its own. Eventually some kind of enforcement will be necessary despite the arguments about censorship and free speech.

AI technology may help find content after it is deployed. It really needs to be used to intercept harmful material at the point of ingest if possible. That could provide 100% moderation but raises counter arguments about censorship and free speech.

We will look at fake vs. real content another time when we examine machine generated content in a future series.

Matters Concerning Rights And Intellectual Property

From the perspective of a creative participant in any industry, the ethics are very simple. We own the fruits of our labors! We may license that work to other creatives, publishers and broadcasters to deploy and in return we may expect to receive some remuneration. We may choose to give it away for free but that is our decision. Long established worldwide copyright regulations exist to protect this status quo. There is a moral and regulatory obligation for broadcasters to respect the intellectual property rights of creators.

Conclusion

This guide is published during a time of great change. There seems to be a convergence underway that may well make the entire concept of outputting content from broadcast infrastructure to social media platforms obsolete. The current technological struggle to integrate streaming and broadcast services may become irrelevant if streaming completely dominates TV viewing.