There are nearly endless possibilities when it comes to choosing how to encode your video or audio files, and the choices you make can have a profound effect on the time it takes to process your inputs and the quality & compatibility of your outputs. Pulpit AI supports a wide range of encodings, but the best results are usually achieved by following the recommendations listed here.
Media encoding is already complex, and made more so by the fact that different pieces of hardware and software provide different levels of control over the process. These recommendations describe the ideal content for use in Pulpit AI, but there is a chance that you'll have limited control over some of these parameters.
Every system and workflow is different. If you find that your videos aren't always working as well as you'd like in Pulpit AI and you don't know where to start in applying these recommendations, reach out to support! We're here to help.
Quick Start Recommendations
To make getting started as easy as possible, we've cherry-picked these recommendations based on the more detailed info below. These should allow the majority of uploads to succeed without issue, but if you do encounter problems then read on to find more ways to strategically optimize your files! Check out the below video on how to use Handbrake, our highest recommended video encoder!
Video:
File format: MP4
Resolution: 1920x1080
Frame rate: 30 FPS, with a Constant Frame Rate (CFR)
Bit rate: 8.0 Mb/s
Audio:
File format: MP3
Bit rate: 192 kb/s
Sample rate: 48,000 Hz
General Video Recording Recommendations:
Make sure your speaker is as in-focus as possible!
When uploading files, try to trim them only include speaking portions, and not music/white noise, as they can cause language detection errors.
Use a file you are uploading from a computer! Links from YouTube/Facebook/etc. are not supported due to the Terms of Service of those platforms, and not all uploads from mobile phones/tablets are supported, specifically some Andround
Detailed Encoding Guidelines
General
An uploaded video or audio file must not exceed 12 hours in duration or 50 GB in size.
HD Video
HD video is the most common type to be produced by churches, and will generally result in great-looking outputs when uploaded to Pulpit AI. It also has the advantage of producing much smaller files, leading to faster uploading and processing times.
However, because the clips generated by Pulpit AI will be somewhat zoomed in, sharp-eyed viewers may notice a slight reduction in visual quality (just like if you held a magnifying glass up to a printed page, you would likely notice some reduction in the sharpness of the text on the page). If you have the equipment and want to optimize for quality over efficiency, see the recommendations for 4K video below.
Container format
Container format
Videos should be uploaded in the MP4 container format. Some other formats are accepted (such as MOV, MKV, and MPEG-TS), but certain aspects of these container formats may cause issues.
The MP4 and MOV container formats support what are variously called "edit lists", "edit atoms", or "edit boxes". These are hidden bits of metadata that keep track of how a video file has been edited in a post-production editor (like Final Cut Pro). At this time, edit lists are only partially supported, and we are investigating reports of issues with some files that contain edit lists.
To remove an edit list from a file that contains one, try opening that file in a utility like Handbrake or Adobe Media Encoder and exporting a copy of the file. Ensure that your export conforms to the other recommendations in this article!
Compression format
Compression format
Videos should be compressed according to the H.264 (or AVC) compression format. H.265 (or HEVC), VP8, VP9, and others may also work, but those compression formats are not used as frequently and so their edge cases and nuances are not as well understood.
Bit rate
Bit rate
The combined bit rate of the video, audio, and container overhead should ideally be no higher than 8.0 Mb/s; therefore, the video bitrate itself should probably not exceed 7.5 Mb/s on average.
The video bitrate must not exceed 16 Mb/s for any given GOP (see "Frame properties" below).
Frame properties
Frame properties
Neither the width nor the height of the video should exceed 2048 pixels.
The sample aspect ratio (the ratio of width to height of each individual pixel) should be 1:1. In other words, each pixel should be square.
The frame aspect ratio (the ratio of width to height of each frame of pixels) should be one of the following:
16:9 (most common for TV and web video)
4:3
1:1
9:16
6:13 (the aspect ratio of modern smartphone screens)
The frame rate should be constant (not variable), and the scan type should be progressive (not interlaced nor progressive segmented frame). The rate of frames per second could be anywhere between 5 and 120, but ideally should adhere to one of the following common rates:
23.976
24
25
29.97
30 (most common)
48
50
59.94
60
120
The GOP size should be less than or equal to 20 seconds of video, and the GOP type should be closed.
GOP size might be labeled keyframe interval.
GOP size might be defined as a number of frames rather than a number of seconds. In that case, the GOP size should be less than 20 times your chosen frame rate.
GOP type might be labeled I-frame type. In that case, the I-frame type that corresponds to a closed GOP is IDR.
Color properties
Color properties
The color depth of the video should be 8 bits per channel.
The color space, primaries, white point, and transfer characteristics should adhere to Rec. 709.
The color range should be full.
4K Video
4K packs a lot more detail into each video compared to HD, so as long as you're set up to produce great 4K content and your audience is set up to play 4K content, you can get phenomenally-sharp results from a 4K upload.
However, note that the average person will likely not be able to tell the difference when comparing a 4K to an HD video on their phone. Many devices are not capable of playing 4K content, and many network connections are not fast enough to play 4K content without buffering. Furthermore, trying to produce 4K content will generally cost a lot more time and money than producing equivalent HD content, so there's a definite tradeoff.
If you want to upload 4K video to Pulpit AI, the recommendations for HD video should be applied except as noted below:
Bit rate
Bit rate
Higher bitrates will become increasingly more difficult for your audience to play smoothly. For 4K video, the video bitrate should not exceed 20 Mb/s on average.
The video bitrate must not exceed 40 Mb/s for any given GOP.
Frame properties
Frame properties
Neither the width nor the height of the video should exceed 4096 pixels.
For 4K videos, the frame rate must not exceed 60 frames per second. Otherwise, follow the frame rate recommendations for HD videos.
For 4K videos, the GOP size must not exceed 10 seconds. Otherwise, follow the GOP recommendations for HD videos.
Audio
Pulpit AI supports uploading videos files with audio embedded and standalone audio files. Both should adhere to these recommendations.
Container format
Container format
Standalone audio files should be uploaded in the MP3 container format. Some other formats are accepted (such as M4A or WAV), but certain aspects of these container formats may cause issues.
Audio embedded within a video file is subject to the container format recommendations for video files.
Compression format
Compression format
Audio should be compressed according to the AAC compression format, specifically the AAC Low Complexity profile (often shortened to LC-AAC or AAC-LC).
Bit rate
Bit rate
The combined bit rate of the video, audio, and container overhead should ideally be no higher than 8.0 Mb/s; therefore, the audio bitrate itself should probably not exceed 320 kb/s on average.
Sample rate
Sample rate
Audio uploaded to Pulpit AI should have a sample rate of either 44,100 Hz or 48,000 Hz. The latter is preferred.
Venturing Beyond The Recommendations
Pulpit AI's transcoding system is designed to try to handle anything that may be thrown at it, but it will never be able to handle everything perfectly. If you must upload a file that doesn't match all of these recommendations (for example, you only have video files in the MP1 container format and you don't have time to convert them to MP4), you are welcome to do so, and the file will likely be transcoded without issue. However, it will almost certainly take longer to process, and there are a few known problems you may experience:
Clip generation has a higher-than-average rate of failure for M4V and MOV files, for reasons which are as yet unknown.
MP4 and MOV files with overly-complex "edit list" metadata can generate clips with severe desynchronization between the audio and the video.
Likewise, videos with variable frame rates can generate clips with severe audio/video desync.
If your media production workflow results in files that don't work well in Pulpit AI, the best approach is to either change the settings you're using during the original encoding or pass your files through a transcoder before uploading them. HandBrake is a free and open-source application for transcoding videos, though it cannot process standalone audio files. Adobe Media Encoder (part of the Adobe Creative Cloud tool suite) can process both video and audio files and offers excellent control over the encoding parameters, but comes at a cost. And if you're technically inclined, you could even use FFmpeg, which is by far the most versatile media transcoder in common use, though it comes with a high learning curve.
Tips For Subsplash Customers
If you're coming to Pulpit AI as a Subsplash customer, there are a couple things that you should consider:
If you are live streaming with Subsplash and want to use Pulpit AI to generate clips and other outputs from your live stream, we highly recommend that you capture a local recording alongside your live stream and upload the local recording to Pulpit AI. Most video production systems can do this easily, and the local recording that you capture can be done at a much higher quality than the recording that we capture on our servers.
Similarly, if you're producing edited videos for use in both Subsplash and Pulpit AI, it is recommended that you create two separate files with different encoding parameters. What is considered optimal on Subsplash isn't very optimized on Pulpit AI and vice versa. For best results on both platforms, creating two separate deliverables from your master recording is recommended.