Skype works by changing the speed of the discussion (without distorting the voice) when there are communication problems, filling in the gap until the next packets come in.

But (a) they could do this with video too, and (b) the essence of my idea is, in addition to the stretching of the voice and video, to gve a degraded alternative, something like an edged detection outline only, of the images for the video, and a porosodian version of the discussion, I'll explain immediately what that is, when things get real slow.

Years ago at IBM Haifa Research we were able to reconstruct Ariel Sharon's voice from a short sample of his voice, pulling out "the tonal way he talked" along with "the sound of his voice", and to artificially construct his voice saying funny things that he would never say in real life. Everybody used to laugh when they heard him say: "We don't need the territories, we only want peace!"

The information will be passed at the beginning of the conversation, and the phonetics will be passed in short packets BEFORE the rest of the information, as a backup.
-- pashute, Feb 14 2014

Voice to MIDI [pashute, Feb 15 2014]

I used to wonder, while doodling along with sports page photographs of athletes, if predictive algorithms like these could come from the lines that it takes to predict 3d motion from 2d photographs, you know and then create a 3d space in which words or topics are located and then predict out with the same formulas. Doodling along with pictures of soccer players for instance is one of those border line machine/human things that i would figure a machine could learn how to do from watching a human, and the algorithms could be statistically essenced out of that. By "doodling along" I mean using a line to balance between following the lines of the body, predicted motion and making mistakes. The hard part would be fitting conversations into that same kind of space but I think if you worked archetypically at a number of different levels you could do it. It would probably take a long time to perfect.
-- JesusHChrist, Feb 14 2014


Had an idea for Aerial Sharon Drones. Obviously the invention would be clear, as in transparent airborne defence balloons, that intercept oncoming missiles while simultaniously accumulate cloud cover using solar energy and humidification apparatus. It holds a large sandwich and fans exhaust fumes with seperate propellors.
-- rcarty, Feb 14 2014


[cheese crust] its not predictive, only allows for more time stalling and deceiving the listener during the bad parts of the connection. It works the same as Skype works, but with a the possibility for even more time without breaking the discussion, or for completing a conversation under worse conditions (such as an emergency).

[bigsleep] Some clarifications, in order to save me from the fishbone fate: I worked at IBM research Haifa from 1993-1999, wrote part of the programs that ran the Thinkpad, and now Lenovo that I'm writing this on. The program for "prosode" synthesis was written around 1994 or 5. Sharon at the time was removed from politics since 1983 (!) by court decree, never to be a minister of defense (or commander of the military) again, for his part in approving the Christian Lebanese Marones attack and massacre in the Palestinian Muslim "refugee camps" in Lebanon of Sabra and Shatila (by then towns), and giving them (never acknowledged) assistance.

Who would have thought that he would become Prime Minister some day, voted in by the smallest vote ever, and that the courts would allow it. And then that, after provocatively walking on the temple mount in Jerusalem, that he would dismantle settlements forcefully, at the same time becoming the hero of the Israeli media, and most of the left.

About the Mossad, that's not me, its my mother in law, in the pictures from Dubai. I thought she looked familiar, but only when you mentioned it, I finally realized where I saw that face.

And about the prosodies, lookup wikipedia Prosody (linguistics) - its not my entry there.

While googling for old colleagues who worked on that, and made it, I came across an anno that I wrote in 2011 on the idea Voice to MIDI where I give some details.
-- pashute, Feb 15 2014



random, halfbakery