How to manage the wide diversity of file formats for ingest, archiving and delivery purposes? Example of Ina’s policy (MARCHAND)
1. Preservation, Digitization and Migration
Problems and Solutions
How to manage the wide diversity of file formats
for ingest, archiving and delivery purpose ?
Example of INA policy.
2. Introduction
The early years of video files management
New challenges of the 2000’s
Ingest
Normalize all input to a unique conservation format ?
Keep all input as it is ?
Using a selection of approved formats ?
Outgest
Workflow
INA Workflow
Transcode/normalize to a supported format
Approved formats at INA
3. The early years of video files management
In the 1990’s, the issue was much simpler :
• In and out : tapes
• Encoding and decoding : hardware based
• Not many solutions/formats available
INA : MPEG-2 8Mb/s was the archive format
Encoder Decoder
MPEG-2 files
storage
Ingest
Digital
archive
Outgest
FIAT – Parallel sessions – 20 october 2017
4. New challenges of the 2000’s
FIAT – Parallel sessions – 20 october 2017
1990’s file formats are no longer good enough. The original tapes’ sound and picture are in superior quality and the
customers need the best quality available.
INA chose JPEG-2000 in 2011 as a new conservation format for tape digitizing.
There is now a much greater variety of carriers and file formats coming in.
mxf
mov
mp4
mpg
avi
Optical disc
Flash drive
Hard drive
Data tape Network transfer
Source media Wrapper Video codec Audio codec
+ metadata
5. Ingest - Normalize all input to a unique conservation format ?
Easy management once the conservation format is created
Very demanding in transcoding ressources
Possible quality loss from very high-quality sources
Possible data space waste from low-quality sources
What happens when the chosen format becomes obsolete ?
FIAT – Parallel sessions – 20 october 2017
Transcoding servers
movmxf
mp4
mpg
avi
Conservation
format
Digital
archive
6. Ingest - Keep all input as it is ?
Simpler ingest
No quality loss of space wasted
Makes quality control very complicated
Potential compatibility issues
Outgest is going to be very complex
Kind of a mess…
FIAT – Parallel sessions – 20 october 2017
Digital
archive
Format A
Format C
Format B
Format X
movmxf
mp4
mpg
avi
7. Ingest - Using a selection of approved formats ?
If one format is not enough and all formats is too much…pick the most relevant formats for your usage :
What you’re most likely to get as input formats
What you’re most likely to be asked as output formats
What will provide an acceptable quality, compatibility and sustainability (wide adoption by professionals and/or
public)
FIAT – Parallel sessions – 20 october 2017
Format A
Format C
Format
D
Format
B
Input by format, bigger is more files
Digital
archive
Transcoding
Format A
Format C
Format C
8. Identified format
Ingest - INA Workflow
FIAT – Parallel sessions – 20 october 2017
movmxf
mp4
mpg
avi
Unsupported format
File specs
analysis Worth
creating a
new format ?
Yes
No
Transcoding to nearest
supported format
Format AAA
Format CCC
Format BBB
Digital
archive
Each chosen format is given a 3-letter word to make the identification easier.
This "trigram" is inserted in the video filenames and is referenced in our MAM tools and transcoding presets.
???
AAA
9. Ingest - Matching a supported format
A supported format have characteristics that have to be perfectly matched :
Container
Video and audio codecs
Framerate
And others that can tolerate some deviation :
Bitrate
Number of audio tracks
Encoding profile
Ideally, you must be able to play and transcode from and to all the chosen formats.
FIAT – Parallel sessions – 20 october 2017
mov
4:2:2
mov
4:2:2 HQ
Source format Supported format
10. Ingest - Transcode/normalize to a supported format
When transcoding into a supported format, preferably :
- Only recode what needs to be recoded, keep intact what can be (even if sometimes slightly different from target)
- Choose target codecs whose profile/bitrate will match the quality of the source codecs
- Don't convert to an inferior resolution
- Don't mix framerates, use dedicated formats for each
FIAT – Parallel sessions – 20 october 2017
Source format Target format
rewrapping
straight copy
mov mxf
11. Approved formats at INA
A few INA trigrams…
FIAT – Parallel sessions – 20 october 2017
Code Video Audio
EXP MPG MPEG2 8M CBR 4:2:0 8 bit 720x576 MPEG1-Layer2 16 bit 48kHz 192kb/s
AAA MXF JPEG 2000 100M VBR 4:2:2 10 bit 720x576 PCM 24 bit 48kHz 2304kb/s
PKY MOV Apple ProRes HQ 185M VBR 4:2:2 10 bit 1920x1080 PCM 24 bit 48kHz 2304kb/s
PCO MXF XDCAM HD422 50M VBR 4:2:2 8 bit 1920x1080 PCM 24 bit 48kHz 2304kb/s
VIS MP4 H264 600k VBR 4:2:0 8 bit 512x384 AAC 48kHz 64kb/s
All the differents formats cover :
Resolutions ranging from 400x300 to 4096x3112
Bitrates ranging from 300kb/s to 10Gb/s
Color subsampling schemes (4:2:0; 4:1:1; 4:2:2; RGB)
Framerates (25; 29,97; 23,98)
Etc…
12. Outgest workflow
Current processing capacities allow fast on-demand transcoding at delivery stage.
FIAT – Parallel sessions – 20 october 2017
Format AAA
Format CCC
Format BBB
Digital
archive
Customer
Deblocking/
denoising
Upscaling
Transcoding
Processing
Even if the customer can deal with our archive format, we prefer processing it ourselves to match the target use.
We know best how to present our archive material.
• Older compression formats may need some deblocking.
• SD material may need high-quality upscaling.
• 4/3 material needs to be fit into a 16/9 frame while avoiding picture cropping and stretching.
MPEG-2 SD 8Mb/s XDCAM HD 50Mb/s
13. Thank you !
Etienne Marchand
emarchand@ina.fr
FIAT – Parallel sessions – 20 october 2017