Hot glue and wire Stylish and robust I built some glasses with LCDs as lenses for a Christmas fancy dress party, exhibiting craftsmanship that is... errr... second to none o.0
The project was hastily thrown together in the evenings of the week running up to the party, so if ever anything can be classed a 'quick and dirty' hack, this is it.
The LCDs came from scrapped phones, and are 101*40 pixel graphic displays, which are controlled over a (poorly implemented) i2c bus. The glasses run off a couple of old camera LiPo batteries and an Attiny2313. Front of Display Rear of Display (Pin 1 on right) The display used is an 8-pin affair, a Seiko S-4548ACG (I have recently written a linux kernel driver for this display. Find out about that here). This display doesn't exist anywhere on the internet (except in some obscure patent application), however I was fortunate enough to obtain a datasheet and get it working.
The display pinout is (pin1 is on the right when looking at the back of the display):

Pin	Symbol	Function	Comment
1	VLCD	VLCD	VDD-8V - VDD-5V
2	VSS	Logic Ground	0V
3	VDD	VDD	2.4V - 5.5V
4	SCL	I2C Clock	~100kHz (though I managed more)
5	SDA	I2C Data	Write address 0x74
6	OSC1	RC Oscillator input	Connect 470k across OSC2/OSC1
7	OSC2	RC Oscillator output
8	VDD	VDD	Tie 3 & 8 together externally

The display's VRAM is split up into 5 pages (rows) of 101 bytes, so that each bit in each byte represents a pixel.
The I2C transaction looks something like this:

S	0	1	1	1	0	1	0	0	1	0	1	1	0	0	1	1	1	1	*	0	0	0	0	0	0	1	1	.	.	.	P
	I2C Address (0x74)								A	Page 0b000 - 0b100					Cmd			A		Col Address 0b0000000 - 0b1100100							A

You may notice that all the 'ACK' (A) bits are written as 1, apparently there is an error in the silicon which means the LCD will not generate the correct ACK signal, so that cycle should be ignored (and ACK assumed). The "..." is the display data - here you write as many bytes as you want, which will go into the VRAM sequentially starting at the location specified by Page and Col Address. When the end of a page is reached (i.e you get to column 0x65), the next write will go to the first column of the next page and so on.
The bits in the command field 'Cmd' have the following meanings:

MSB	All pixels ON(1)/OFF(0)
-	Power Save ON(1)/OFF(0)
LSB	Display ON(1)/OFF(0)

Power Save has the highest priority and select all pixels has second highest.
"Aha! That should be a '1'!" Let it snow! As much as I would have loved to implement a lovely, neat, general-purpose, hardware-accelerated I2C stack... I didn't.
This being my first attempt at I2C on an AVR, I didn't have a ready-to-use I2C library (and though I did look online, I didn't like any of the implementations I saw). Given the tight time constraint, I wrote code to bit-bang the communications with the display.
As the screens are write-only, I only bothered to implement writing, and was amazed when I managed to get the screen to do something (after tracking down the odd erroneous bit on my oscilloscope).
Once I could write to the screen, I had to decide what I was going to actually show on it. I broke out the GIMP and drew a little snowman, as well as 5 'pages' of snow (101*8 pixels), and used the ever helpful "C header file" output to save the monochrome bitmaps into something I could use. I had to do some trivial modification to get the images into the 1 bit per pixel format I needed, and then got to work getting them onto the actual screen.
By writing each 'page' of snow to each page of the screen, then incrementing the page number I started on, I could get a nice falling snow effect, and then I overlayed the snowman and sat back and admired my work.

Power Supply I should be a fashion designer The glasses were simply a piece of cardboard, cut to shape and reinforced with wire. I pulled the reflective backing off the displays and put them where the lenses should be, added a couple of LEDs and used copious (there really was a lot of it) amounts of hot glue to hold it all together. The power source was two 3.7 V Lithium camera batteries in series, giving me about 8 V when fully charged. This was split using an op-amp buffer to give me a virtual ground and the positive and negative supplies I needed to run the AVR and LCDs.
A few wires and a rubber band later, and my creation was finished.