Chromaster v2.x : The ultimate companion Topaz Video / Video AI deserves

[Alpheratz]

Chromaster v2.10.3 α

Black bar detection is now very accurate

The detector was measuring crop on 68-pixel thumbnails. On a 4K source that’s 32 lines of slack, no matter the threshold. It now measures the full-resolution frame.

Black bar detection

• Pixel-exact — measured on the actual frame, not a thumbnail. Down to a single line.
• It follows your selection — leave the toggle on, click any thumbnail, the crop updates. No more off-and-on.
• It says why it failed — five distinct causes, five distinct messages, instead of one that named the wrong thing.
• It works on reopened files — the detector was dead on any file opened a second time. The thumbnail cache restored the strip without writing the files it reads.

MKV metadata

• MKVextract now ships with MKVmerge — the installer downloaded the whole toolkit and kept a single binary. Without the second one, reading embedded HDR and quality data back out of your MKVs failed silently.
• MKVextract is a proper dependency — its own row, path, version and status, like every other tool.

Startup

• Hardware detection is twice as fast — the four probes ran one after another. They now run together (900 ms down to 450 on my hardware)
• Status lines stop guessing — decode, NVENC and Vulkan rows said “not found” while the probe was still running. They say “detecting” now.

Download Chromaster v2.10.3 α

[obelinkin]

Aspect ratio
Is Non-square pixels preserved through the pipeline, including through a resize; or square up anamorphic pixels, force a display aspect, or letterbox to a target size?

[Alpheratz]

Honest answer, two halves.

What works today

• Source pixel aspect is read on open. SAR, DAR and PAR are shown in the Crop and Size tabs, with common names recognised (PAL 16:9, NTSC DV…).
• H.266, ProRes and FFV1: FFmpeg writes the file end to end, so anamorphic sources come out correct.

What doesn’t

• AV1 (SVT and NVENC) and AV2 go through a pipe and are put in a container afterwards. Nothing on that path carries the aspect, so the file plays back with square pixels.
• Real case: a PAL DVD at 720x576 meant for 16:9 comes out squeezed in AV1, correct in ProRes, same settings. Nobody chose that, the two paths just differ.
• No automatic square-up, no forced display aspect, no letterbox to target size.
• Workaround for now: resize to a square-pixel resolution yourself in the Size tab. Resizing keeps the display aspect, so it works.

Plan

1. Write the aspect at mux time for the three codecs that lose it. That’s the actual fix.
2. Optional next: a “square up” button in the Size tab, one click to the equivalent square-pixel resolution.

Letterboxing is a separate feature, not planned yet. Tell me if you’d use it.

[obelinkin]

Thanks for the detailed answer.

@Alpheratz

Letterboxing

It’s either cropping the frame (cutting off heads) or having bars on the sides. There’s no other way around it.

[Alpheratz]

Show in picture plz

Working on it

[Alpheratz]

Chromaster v2.10.4 α

It wasn’t easy to design, but I hope I’ve included these features with the usual comfort and ergonomics

FFmpeg’s “Starship Troopers” sample test file

Anamorphic sources, handled properly

Aspect ratio

• Non-square pixels now survive every codec
  AV1, AV2 and NVENC dropped the aspect on the way out. A 16:9 PAL DVD came out squeezed on AV1, correct on ProRes, same settings. Now written at mux time.

• The GUI shows the corrected image
  Every preview, thumbnail and comparison displayed stored pixels. An anamorphic source looked wrong inside Chromaster before you even encoded it.

• Rectangular / square pixel switch
  Pick what the file actually contains. Rectangular keeps the source dimensions with the aspect written alongside; square really widens the image and needs no aspect at all. Widens, never shrinks, and the scaling factor then works from that new size.

Rectangle px preserved	Rectangle px squared

• Crop presets mean what they say
  The 16:9 button targeted 16:9 of stored pixels. On an anamorphic DVD that produced a 2.5:1 image. It now targets what you see.

• AV2 warns instead of lying
  The format can’t carry aspect at all. You’re told before you encode, with the direction of the distortion.

Ratio names

• PAL and NTSC were swapped
  32:27 is NTSC 16:9, not PAL. Two labels had each other’s value. Your NTSC DVD was named after the wrong continent.

• Half the table was unreachable
  Every pixel taller than wide, which includes NTSC 4:3, the most common DVD format ever pressed. A portrait-video trick was flipping them before the lookup.

• Names are examples now, not claims
  2.39:1 (CinemaScope). The ratio is the statement, the name is context. No resolution inside the parentheses — that’s what made the label lie.

• Graded proximity
  Exact, ≈, ~, or nothing above 5%. You can see how close the match actually is.

• Fractions stay readable
  13924727:7802865 is technically correct and completely useless. Capped at four digits per term, marked approximate.

VOB and older sources

• VOB opens directly
  No more remuxing a DVD rip just to look at it.

• Thumbnails work on untagged sources
  Any SDR file that doesn’t declare its transfer had an empty strip. Silently. The filter graph refused at init and 22 extractions returned in a third of a second.

• MPEG PS duration is flagged
  The format can report half the real length. Thumbnail spread and ETA both rely on it.
  

Also fixed

• Failures are no longer silent
  Nine sites launched a tool and never read its exit code or its error output. A pipeline log now keeps the reason, timestamped, so “22/24 thumbnails failed” has somewhere to point.

• The thumbnail tally counted the wrong thing
  It reported what was requested, not what was produced. {OK} 22 thumbs printed directly under 22 failures.

• Resize comparison misaligned on anamorphic sources
  Both halves showed different regions, visible as a break at the cursor.

• Application group sat too low in Settings
  One row of phantom padding, inherited from a group that never used it.

Download Chromaster v2.10.4 α