Archive for August, 2010

Canon SD780 IS: The cheapest art-capable HD camera on the market

My super-tiny Canon SD780 IS arrived today (also known as IXUS IS 100)! The charger for it is bigger than the camera! I got it with the red color, since that color was cheaper at Amazon (it was $159 last week when I bought it, it’s $169 as I write this). With it, I also bought the Zeikos lens adapter for Canon digicams, so this way I could use my ND filter. See, small-lens digicams tend to record video outdoors in high shutter speeds (lens/algorithmic limitation), so a strong ND filter can help with the situation by forcing the camera to open up its aperture and slow down the shutter speed.

So when I received the camera I set its “Color” settings to flat (contrast, sharpness, saturation, and skin tone set to minimum). These flat settings also maximized the dynamic range of the camera. Then I disabled digital zoom, so I don’t jump into its nasty look by mistake while shooting. Then, while shooting, I reduced 1-2 steps of exposure compensation and locked it there (these digicams tend to over-expose outdoors, so I made sure mine didn’t). I also turned OFF image stabilization, even if I was handheld (to gain a bit more video quality).

The result was a beautiful-looking video of my garden, very clean-looking, and definitely better than many prosumer SD camcorders that used to cost $5000 just a few short years ago! And all that with less than 200 bucks, and at a tiny package. Color grading was very easy, since the “flat” settings created a very nice image that scaled well with color grading.

Check the video below (switch to 720p quality).

Right-click to download an un-processed sample video (14.5 MB), with its original 24mbps video encoding. The only thing I did to it was to re-wrap it to MP4 (from MOV, without lossy video re-encoding), just so it becomes compatible with the PS3 and the XBoX360. You see, that’s where you can really appreciate its quality: on your TV. I just played this unprocessed file on our plasma HDTV via the PS3, and it looked like a million bucks! No one would ever believe that what they see on screen was shot with such a small camera.

If you prefer to watch it on your computer, it’s best watched with any player, except Quicktime (it uses the wrong gamma value with h.264, making videos look washed out). If you must use Quicktime, then load the clip in it, click “Window” from the main menu, “Show Movie Properties”, click on the “Video Track”, then on “Visual Settings”, and then change “Transparency” to “Straight Alpha”. Then, enjoy.

And here are two frames, directly out of the 720p stream, without any post-processing adjustments either.

Now look me in the eye and tell me that you need a better HD camera than this, if you’re just a video art enthusiast. In my opinion, this is the cheapest camera on the market today that has the ALL of the MOST NEEDED video features to do art! It has exposure compensation and locking, and it has color control, so it doesn’t make your footage look like cheap video. Only thing missing is 24p support (in addition to its 30p), but if you’re just doing visual video art without audio, then you can always slow-down your video 20% to 24p to give it the movie look (this trick has worked wonders for me).

Auto-focus works well. Manual focusing is not present as it is in the SX200 IS, but that’s easily fixed by focusing at a temporary object in front of the lens at the desired spot. Only real disadvantage of the SD780 IS is that it’s not good under very low light. Thankfully, I don’t plan to use this camera indoors.

UPDATE: One more video!

UPDATE 2: A 30-sec excerpt for a music video I shot with this cam.

Snapshot from the Canon SX200 IS

Today I shot a new music video for a band from Santa Cruz, named Old Arc. Conditions were such that I had to use my trusty, small Canon SX200 IS for the video, and as you can see in the frame directly out of the camera, 720p quality is more than adequate for such a small camera (even if this was shot at the long end, where lens quality usually suffers, and even if the shot was handheld). The video was shot in 30p and then was slow-down’ed to 24p via Cineform NeoHD. Will start editing it tonight. I’m excited.

Is CHDK worth it for video?

And the short answer is… NO.

I installed the CHDK firmware on my Canon SX200 IS (1.00c) tonight, and made some video tests. I shot high-detailed objects with the normal firmware (default video quality is 75%), and with the CHDK firmware at video quality “99%” (and with all its video quality-related options set to ON).

The result was that the 99% quality setting would produce over 50 mbps bitrate videos (instead of the default 24 mbps). This is good. Problem is, after inspecting the footage, there was only about 2% visual quality gain. I tried with a small moving object too, same result. You can view the test images below yourselves (saved as full quality, with decoder at full quality too).

Test image 1, default 75
Test image 1, CHDK 99

Test image 2, default 75
Test image 2, CHDK 99

The other problem was that with any quality over 85%, the SD card (or was it the DSP/CPU of the camera?!?) could not keep up. Sometimes the video would bail out after 1 second of recording, sometimes after 6-7 seconds. Only below 85% it was stable with a Class 6 SDHC card. And this made that 2% of gain, disappear too.

Here’s an actual JPEG image, captured in AUTO mode, and then resized to 720p and re-saved again with only 75% quality. Even after such harsh treatment, that image is zillions of times better than any still from the videos I captured.

If the CHDK developers did not actually have any bug in their video tweaks, it means that the Canon digicam videos lose tons of quality DURING capture/resizing, internally, before the footage even reaches the encoder! So that would be a weak DSP they use, rather than a weak encoder.

Finally, the histogram, and zebras did NOT work on video mode with the CHDK firmware! Obviously, the video mode gets very little love from the CHDK developers. Which is sad. I mean, one other big feature that filmmakers would love is additional 24p/25p support. But this feature is not even on their todo list!

So, unless you’re interested in CHDK for its photography features, don’t touch it for video. It offers nothing of note to us videographers.

UPDATE: I shot one more test, under better lighting conditions this time, during daytime. As you can see, the quality gain was still minuscule.

Test image 3, default 75
Test image 3, CHDK 99

Interview with Cineform’s David Newman

Cineform is one of the most popular intermediate codecs today, helping out not only filmmaking professionals, but also video enthusiasts. I’m very happy today to introduce you to David Newman, CineForm’s CTO. David is also the company’s compression & image processing engineer.

1. What is taking Cineform apart from other intermediate codecs, e.g. Avid’s DNxHD, R3D, Lagarith, or Apple’s ProRES?

David Newman: Workflow. 9 years ago when we created the CineForm codec, it was not our ultimate goal, in fact we didn’t intend to build a codec at all. We set out to make post-production workflows, completely software-based, rather than the hardware RT solutions of the day. We found that existing codecs where too slow, with quality issues hampering our workflow goals, so we set out to build something new. As this new codec was not to be a final format, had to be very easy convert into and out of, without accelerator hardware, supporting a huge range of deep pixel formats and operate with as many tools and platforms as possible. This flexibility has resulted in CineForm compression used throughout the Hollywood post industry as mezzanine format, a high quality fast archive for film and TV finishing.

In more recent years we do have some competition in the mezzanine/intermediate format space, each addressing some of the reasons that brought the CineForm codec into being, yet none of codecs listed are as flexible or as widely supported. However data in and data out is not all we are interested in (it never was). Over the last four years our developments efforts involve more of the creative, than the technical part of image development. This started with RAW image compression, which could not be displayed without a debayer/demosaic operation, which itself is more of an art form than hard science, and we felt this should be under user control. This expanded to include, white balance, curves, primary color correction, secondaries through film emulation LUTs, image re-framing and burn-in metadata — all as part of the decoder. The more recent addition of FirstLight allowed the filmmaker full non-destructive control over how the image is developed dependently from their NLE, yet remain fully compatible with all their tools. Tool independent color correction is one the ways CineForm is reinventing the workflow through a codec.

2. So far, there’s been only one camera that would record directly to Cineform (to my knowledge). Why not more prosumer or pro cameras? Why is it difficult to persuade camera manufacturers?

David Newman: We are software company with only 8 people, mostly engineers. The camera guys generally need an FPGA or ASIC implementation for lower power usage and long battery life. While the CineForm codec is designed with potential hardware implement in mind, and will have excellent low power characteristic, it will be the responsibility of a licensing partner to do the hardware port. In some ways we’re a victim on our software successes and Intel’s every increasing performance, hardware vendors from Silicon Imaging, Wafian, CineDeck and 1Beyond, all simply use a mobile Intel CPU for live CineForm encoding, which constantly get faster and cheaper — pushing back the need for a hardware port.

3. Any plans to resurrect the HDMI-recording device? Would that make sense today?

David Newman: It would totally make sense today — more so than ever — CineDeck is hugely popular, and a pro-sumer version would be a far larger market. That original proposal was just to tease out hardware partner(s), we did have one for a while, but they didn’t survive the market change. We are still looking for the partner to do it, it is getting easier to build such a device.

4. You are also a filmmaker. How important do you feel that “visual quality of footage” is? Do you believe that source footage quality is paramount in creating an astounding piece, or it’s all in the hands of the filmmaker to create magic via whatever means he/she has in his/her disposal? Basically, “is it the camera, or the camera-man?”

David Newman: The camera-man. We develop workflow-based compression, which means we understand that quality at the expense of everything else does not yield the best results — i.e. uncompressed rarely if ever benefits the user. While our Filmscan-2 quality mode is indistinguishable from uncompressed — I don’t use it in my own projects, as perfect source does not make a movie. There is a balance between the volume of data, speed of retrieval, and image flexibility (how much you can push it in post.) I’ve shot a Canon 7D directly to CineDeck, to bypass the camera’s H.264 compression, yes it was better in post, but my shoot was less flexible (this was an on-stage event, so mobility was not an issue). On other projects I’ve shot to in-camera compression and posted in CineForm. So I choice the acquisition format to fit the project needs, so while quality is important, it is not the driving factor.

5. Cineform is a mature codec. However, as an engineer, I’m sure you have more ideas on how to enrich the product. Could you share with us some of these ideas?

David Newman: The codec core is very mature, changing very little on the encode side in the last 4-5 years, yet the progress has not stopped. The codec version is now at version 6.5.1, which just added uncompressed profiles for RAW (existing), YUV and RGB. This helps with battery life and compute loads on mobile acquisition devices where not all frames are compressed (as you can’t tell uncompressed from compress in FS2, this can happen without the filmmaker every being aware.) This also helps with those doubting the compression quality to test and compare for themselves — so it’s both of technical and marketing value. We will be adding many more image development features for 2D and 3D workflows, some I’m really excited to use in my own projects. We are currently working on an idea so outside the box, we don’t know yet the limits to how it will used; it takes today’s already powerful metadata engine to another level.

6. Do you think that 3D movies will be the norm soon, or is it yet another Hollywood effort to resurrect 3D unsuccessfully, as it has happened at least 5 times so far in the last 80 years?

David Newman: It will not become the norm, but it will settle to an ongoing percentage of the feature film market, likely around 10%. This technology is now to the point where is it not the limiting factor, as it was in all the previously incarnations. Now that I’ve seen a good number on 3D features, I’m starting to miss it at some that aren’t — Iron Man 2 should have been 3D, but I’m happy that Inception was not.

7.Any plans for a Linux decoder via the Gstreamer framework?

David Newman: We have a Linux version of the encoder and decoder SDK that third parties have started to implement. This SDK would make a GStreamer implementation straight forward. We are not sure whether CineForm or a third party will do it first. It is up for grabs. Care to do some Linux coding?

8. Are you personally a supporter of WebM or h.264 towards becoming HTML5’s default video codec?

David Newman: Decoder licensing should be made free for all but the highest volume distributors.

9. How do you feel about the recent explosion of videography and filmmaking by amateur enthusiasts in the last 2-3 years online? Do you see this as an artistic and social revolution, or merely sign o’ the times?

David Newman: I believe it’s the impact of Canon 5DMk2 on enthusiasts, more than any other market or social force. That camera, along with the 7D, T2i, GH1, etc. has allowed part-time/hobbyist filmmakers to make images that look like the high-end projects which inspired them. This is what I got into this business to help create. As a hobbyist filmmaker myself for 20+ years, I could never afford the high-end tools, so I started building my own, working to make the results look as good — that just got a lot easier.

10. And the 10 million dollars question: would Cineform be a good codec for applications like remote medical procedures, or… spying satellites? Do you eye markets apart the entertainment industry?

David Newman: We are in some of these markets, we have users at NASA, on extreme deep ocean ROVs, we used on terrain imaging, and many other applications well beyond our design goals. All these project require, fast, high-quality compression processing, and our easy to use SDK certainly helps.

An update on my garden

A few weeks ago I wrote a blog post about starting a vegetable garden on my patio. Two and a half months later, the plants have gotten rich to look at. Here’s just one side of my garden, with the runner beans and my cherry tomato looking prominently large:

Today also marked the first day where I harvested some amaranth. Unfortunately, my patio only has sun for 3 hours a day, so this has kept back most of my plants from growing faster. The amaranth should have been considerably bigger by now…

From all my plants, the ones that didn’t do well was my spearmint (I think it contracted the same disease the tree in our yard has), two of the three peppers, while my zucchinis are making lots and lots of flowers… but no actual zucchinis (no bees are left in the Bay Area?!?).

Tonight I’m grilling on our electrical grill these pork chops, which as you can see here I have marinated in olive oil, lemon juice, and fresh, cut out right from my garden: oregano, thyme, sage, basil, parsley, and spearmint. I can tell you, these chops are coming out so flavorful, that I can’t even describe in words!

Click below to view more pictures of my garden! 🙂

Read the rest of this entry »

Market Segmentation: one of the evils of the world

Today, Canon announced their two highest-end HD-capable small digicams, the SD4500 IS and the S95.

The S95 is the superior still camera of the two, with a large CCD sensor and a very fast stabilized lens (for what it is). Surely, a great buy for stills. Regarding video though, the S95 has the same capabilities as their much cheaper digicams have: 720p at 24 mbps bitrate, h.264 MOV, color control, exposure compensation, the as importantly exposure locking, and manual focus via the “wheel”. The only difference between the S95 and the other cheaper Canon HD digicams in terms of video is that the S95 does 24p instead of 30p.

And then there’s the CMOS-based SD4500 IS. The SD4500 IS has an embarrassingly small sensor for its price/range (same as in their $100 digicams), and a not-so-fast lens. But, it does 1080/24p. At 43 mbps no less. And 720/30p at 30 mbps. Plus all the other video features mentioned above.

So here we have the perfect small still digicam that leaves a lot to be desired in terms of video, and we have another small still digicam that comparatively sucks at stills, but it does video better.

Perfect market segmentation for Canon! This way they can sell both of them, to people who are mostly interested in video, or mostly interested in stills. But absolutely none of the two lives up to the expectations. This pisses me off so much.

The only hope I have right now regarding “small” digicams is for the upcoming Canon G12 to have SELECTABLE resolution AND frame rate: 1080p/720p at 30p/25p/24p each (and why not, 720p @ 50p/60p).

Because if they screw this up too, these guys don’t know what they are doing. Why not give their market what they need, instead of creating 40 products with 99% the same feature-set, confusing people? Or segmentize their products so much, so no one gets to be happy with what their bought?

I’m not asking for freaking manual control here. That would indeed be somewhat high-end at this point for the price range of these cams. I’m just asking something that is most important for video to be able to do something useful with it: frame rate selection first and foremost, and then resolution. For their higher-end digicams, these features should have been there as standard.

UPDATE: Well, I’m now even more pissed off at Canon. C|Net just leaked by accident the details of the new G12, and it also only does 720p.

Another worthless “high-end” digicam in terms of video — at a day and age when Canon has the capability of adding the resolution and frame rates needed without much trouble.

Updated video

I re-cut the following video that I shot last year, in order to make it slightly shorter, and to make it a bit more interesting in some specific places that I had since identified as… boring. The reason for this update was so I can submit the video to this year’s Free Culture ShowCase competition by Ubuntu — a competition that I won first place last year.

I really like this video, I feel that it’s one of my most underrated ones. Each time I play that video on our PS3/TV, JBQ tells me “It’s obvious that this video means something to you.” He personally can’t connect with it, but he can somehow notice that this video represents me.

Indeed, the video reminds me of myself when I was a kid, living in the city of Preveza, and being care-free. That was late ’70s/early ’80s (before we moved back to my father’s village because of financial and family problems). When my mother visited me here in the USA last June, she also mentioned how the place in the video (and in real life too) also reminded her of Preveza.

So basically, I’d say that this is possibly one of my few videos — if not the only one — where some of my early life experiences, frame of mind, and feelings can be seen visually. The rest of my videos are pretty impersonal (not phony, but trying a bit too hard), while this one comes straight from my nostalgic heart.

“This Is Still My Life” by Hello Monster

Official video for the Bay Area band Hello Monster, for their song “This Is Still My Life”.

I shot this back in May, with a Canon 5D Mark II, a 35mm & a 50mm lens. We shot this on a Sunday, over a few hours (no budget). It’s my first video where I incorporate some dialog.

Greek Spinach Pie

In Greek rural places, the well known “spanakopita” (“spinach pie”) doesn’t have that much spinach in it. This is why we actually tend to call it “lahanopita” (meaning “leafy vegetable pie”), instead of “spanakopita”. Usually, villagers would put in it any green-leaf vegetable, wild or not, spinach-related or not. But of course, in the last 25 years or so, the Greek standard of living has gone high, and this recipe has gotten more civilized. Finding spinach is not so hard anymore. So, this is how we do this pie at the Epirus department of Greece.

Ingredients (for 4-5)
* 3 cups fresh spinach
* 1/2 cup fresh sorrel (λάπατα, optional), or 1 cup of kale
* 1 cup fresh Swiss chard
* 2 leeks
* 4 scallions (or 1/2 of a big onion)
* 1/4 cup fresh parsley
* 1/4 cup of mint
* 1/3 cup rice
* 150 gr feta cheese
* salt to taste
* enough flour
* olive oil
* 1 egg

Execution
1. Make the dough/phyllo, as shown in this video by my mother. In the video you will find both the recipe for the dough, plus the technique on how to shape the phyllo. Please note that for this recipe right here, you will need to use half the suggested dough ingredients. Finally, after making the dough, let it rest for an hour before making the phyllo with it. TIP: You can also make μπλατζάρα, which is the same pie, but without any phyllo.

2. Wash the leaves with cold water. In a big bowl cut in small pieces the onion, parsley, mint, leeks, spinach, sorrel/kale, and Swiss chard. Add some salt, and about 1/3 cup olive oil. Using your fingers, work the mixture, until they almost start lose shape (they start to look wilted). If there’s juice coming out of it (especially if you’re using thawed spinach), you must discard it.

3. Boil some water in a pan, and boil the rice for 3-4 minutes. If you’re using brown rice, boil for 6-7 minutes. Drain the water away, add the rice in the vegetable mixture.

4. Crumble the feta cheese using your fingers, and add it to the mixture. Lightly mix all ingredients again. If the ingredients don’t look as in my first picture, you can optionally stir an egg in it. The egg helps the ingredients to get “glued” better together. Pre-heat the oven at 400 F (200 C).

5. Oil your baking dish. Place one of the phyllos on the bottom of the dish, add the mixture on top of it, even it out using your fingers or a fork, and then secure the whole thing with the second phyllo. Spread some more olive oil on the top of the phyllo. Using a fork make 3-4 holes on the fyllo, so it can “breath” while baking.

6. Place it in the hot oven, and bake for 40-45 minutes, or until the phyllo has started to get a nice brown color. At this point remove the dish from the oven, and let it rest for 2 minutes. Then, turn the pie upside down (technique is shown in the video linked above), and let it rest again for another 10 minutes. After that, it’s ready to eat, hot or cold.


We had this pie tonight along some fried/battered catfish, and coleslaw salad.

Tip: If you would like a bit more kick with this pie, lift the top phyllo, and pour a bit of lemon juice in the filling. It’s much better than just adding salt, as lemon goes better with it.

Chillwave iMix

This is my first Chillwave iMix, published on iTunes today. If you’re curious about the Chillwave sound, this is the place to start: