I did some testing and measure (with a colorimeter) on video test patterns on Infuse, Plex, Home sharing app and LightSpace Connect (which is a TPG - Test Pattern Generator - app), all on my Apple TV 4K 2nd Gen (2021). The results were that the TPG app, Plex and Home sharing have the same outcome while Infuse doesn’t. In particular, except black and 100% white, measured values on Infuse are different on luminance (always brighter than what it should be/reference) and y (higher value than reference, which means more red/green or less blue). So the question is: why?
Does Infuse use a proprietary player/elaborate the source content? Thanks.
Yes, Infuse is a proprietary player.
You may want to run your test video through mediainfo and post the results here. That would help to answer your questions.
When you ask
Does Infuse modify the source content?
Do you mean does it actually rewrite the file on your server? If so, no. If you mean does it interpret the files differently than other players then yes for some cases.
Of course I did not mean that. As the measured values of the patterns displayed via Infuse are different from the other apps, there must be some sort of manipulation (done by the proprietary player) slightly altering the original/intended output (RGB values of pixels), resulting in a ~5% brighter and slightly warmer picture.
There may be but without specifics of the file there’s no way to answer your specific question.
mediainfo returns information of the media file. If you want I may verify the difference with a pixel analyser.
This is interesting…and not so good if confirmed
Without knowing more details about the test content, the Apple TV settings, and the device you are connected to I can’t really offer an explanation for what you are seeing.
But it could it related to this, which affects certain content types.
No James, the test patterns (21 grayscale patterns - in 5% steps) I used are SDR. All of the TVs/monitors (LG OLEDs, Acer/Dell PC monitors) to which I connected my 2nd generation ATV4k give the same results: increase in luminance and y-values (in the CIE diagram) across all scale except 0 and 100%. Maybe some conversion matrix within Infuse’s proprietary player is incorrect, but that’s just my guess.
The issue could be that Infuse internally converts video scale (16-235) to PC scale (0-255) and that causes a luminance mismatch in the final result. I’d run more investigation on that if I had the time…
I always had a feeling that infuse kind of do things differently. I remember seeing someone posted about something related to this from avs forum but I couldn’t find it anymore. Hopefully we’ll find out more soon. Thanks for the research @Anger
Pretty sure you’re talking to them.
@james have you guys at firecore looked into that yet? I’d like to resume my subscription but that mismatch issue stops me. Thanks.
As I got no answer so far, I re-tested Infuse output with some greyscale test patterns. Same results as before, except y is now not exceeding over NIST tolerance (+/- 0.001). Luminance is still higher compared to Apple player (home sharing app & Plex), well over NIST tolerance (+/- 1.5%), therefore the average gamma is lower than expected. I created a resume of the measured data:
Delta Y % is the difference between Apple Player and Infuse measured luminance at the different stimuli in percentage.
and this is what mediainfo says about the files used for the test (I’m sorry, it’s in Italian, but I’m sure you’ll have no problem understanding what is what):
Video
ID : 1
Formato : AVC
Formato/Informazioni : Advanced Video Codec
Profilo formato : High@L4.1
Impostazioni formato : 3 Ref Frames
Impostazioni formato, CABAC : No
Impostazioni formato, ReFrames : 3 frame
ID codec : avc1
ID codec/Informazioni : Advanced Video Coding
Durata : 13s 0ms
Modalità bitrate : Variabile
Bitrate : 75,9 kb/s
Bitrate massimo : 40,0 Mbps
Larghezza : 1.920 pixel
Altezza : 1.080 pixel
Rapporto aspetto visualizzazione : 16:9
Modalità frame rate : Costante
Frequenza fotogrammi : 24,000 fps
Spazio colore : YUV
Sottocampionamento croma : 4:2:0
Profondità bit : 8 bit
Tipo scansione : Progressivo
Bit/(pixel*frame) : 0.002
Dimensione traccia : 120 KiB (96%)
Compressore : x264 core 138 r2358 9e941d1
Data codifica : 2013-12-27 20:28:39 UTC
Data : 2013-12-27 20:28:39 UTC
Intervallo colore : Limited
Colori primari : BT.709
Caratteristiche trasferimento : BT.709
Coefficienti matrici : BT.709
Casella configurazione codec : avcC
I forgot to mention that the issue shows up only when ATV outputs RGB high or YcbCr (both are legal scale 16-235), with RGB low (0-255) is fine. That’s perhaps evidence of some incorrect internal 0-255 to 16-235 conversion.