Capture

In order to show my curator some progress, I wanted to make a video of Sloganeer's output. Of course I could have just recorded it with my phone but that doesn't do justice to the app's nice smooth transitions. Yeah there are various freewares around that could work but they all have their downsides. So much better and more interesting to just roll my own capture. OK there went two days. It's actually pretty challenging to capture frames in Direct2D without stalling the GPU pipeline. And 1080p frames are big enough (8 MB) that it's a bad idea to write them to disk in the render thread. So that necessitates a thread-safe queue, and a thread pool of writer threads, each with their own WIC instance. OK better pack a lunch. But it all works now, and it's neatly wrapped in a class CD2DCapture.

This test helps verify that each frame contains what it should: 

	m_pD2DDeviceContext->Clear(m_clrBkgnd);//@@@
	CString s;
	static int iFrame;
	s.Format(_T("%d"), iFrame);
	iFrame++;
	CD2DSizeF	szRT(m_pD2DDeviceContext->GetSize());
	m_pD2DDeviceContext->DrawText(s, s.GetLength(), m_pTextFormat, CD2DRectF(0, 0, szRT.width, szRT.height), m_pDrawBrush);

And for benchmarking OnDraw: 

	CArrayEx	m_aBenchmark;//@@@

	m_aBenchmark.SetSize(5000);
	m_aBenchmark.FastRemoveAll();

	double	t = b.Elapsed();//@@@
	m_aBenchmark.Add(float(t));

in Destroy: 

	CStdioFile	fBench(_T("bench.txt"), CFile::modeCreate | CFile::modeWrite);
	for (int i = 0; i < m_aBenchmark.GetSize(); i++) {
		CString	s;
		s.Format(_T("%f"), m_aBenchmark[i]);
		fBench.WriteString(s + '\n');

Comments

Post a Comment

Popular posts from this blog

Elevator Door

My curator, Anthony, is a big fan of this slogan project, in fact it was he who suggested it. He’s a professional graphic design designer, and an excellent one, hence I always take his graphic design suggestions seriously. And he did suggest something. He suggested a Pac-Man transition. The idea would be a type of reveal, using a radial “radar sweep” over each letter. It’s pretty cool, but what I don’t like is that it will work well for square-ish and round letters but poorly for narrow letters like “I”. Unless you potentially make the sweep elliptical. But that would probably lose the Pac-Man vibe, visually. The larger category here is per-character reveals. The simplest example is a “window blind” reveal, one blind for each character. However, this would not be acceptable for a single line of text, as it would be indistinguishable from the “whole line” vertical reveal that we already do, and would therefore constitute a duplicate. The horizontal equivalent would be a “sliding door”...
Read more

Orthogonality of transitions

It’s annoying that there’s no transition code for the random typewriter effect. Making it a randomly selected variation of typewriter was a misguided choice. All effects should be treated equally in the sense that they all should have codes, so that they all can be scripted from a CSV. The motivation was that random typewriter is too similar to typewriter, but that should be handled by letting the user customize the relative probability of each transition. And in any case, the thinking was inconsistent, as we already have whole groups of effects that are very similar to each other, such as the four scrolling effects, for example. It shall be done, but not this morning. I have reserved this morning for coding color cycling, which will commence shortly; it’s all planned out like the invasion of Normandy. But I had a thought about that too...
Read more

Hollow transition

[to GPT5] I got a decent night sleep and woke up full of ideas. It’s funny how that happens. It’s a human thing. Here’ the first of some ideas that came to me in no particular order. I’ll relay them to you one at a time, so you can respond without feeling pressed for time or space. The melt transition could be adapted into a hollow transition. Instead of drawing the virgin character and erasing its outline, use the virgin character as a mask and draw its outline in the drawing color. The character would start out as a thin outline, which would gradually get thicker until the character became solid. Or in the other direction, the character would start out solid, and hollow out from the middle until it disappeared. Since we already have a melt probe (which even runs in a worker thread at startup!) we already know for each slogan exactly what outline stroke will cause that slogan to completely disappear. And that’s the same data we need for the hollow effect, no? One thing that worrie...
Read more