exact modeling with speech recognition?

General Forums Blender and CG Discussions
1

i often have to model in exact dimensions and just had a thought. actually it could be quite cool if you just could say “move 3 cm up” or “move x 10.5” or “rotate z 37.5” or things like that.

probably it could be done with python quite easily. there are wrappers for the windows speech recognition engine. are there good open source alternatives?

has anyone experimented with something like that already? i have no clue how reliable speech recognition is nowadays… do you think it would be a good workflow or just stupidly cumbersome? :slight_smile:

2

Stupidly cumbersome.

The latency involved with speech recognition would make it way too slow to be useful.

But you are right, there are a number of python wrappers that could theoretically do it. I’ll believe it’s useful when I see it, so feel free to prove me wrong!

3

I was also thinking in something like this, my case would be different because in my situation would be handy due to the fact that I have pain in my hand when I work for a while in the computer so if I could save some mouse movements by just saying what I want to do that would be awesome, for example:

-Blender, render image
-Blender, go to the UV editor
-Blender, set the end frame to 248
…

4

Firstly, you know about numeric input in Blender?

https://www.blender.org/manual/3d_interaction/transform_control/numeric_input.html

Recently support for units was added too.
Try: G,X,=3cm

The current Python console is extensible, you can add your own console-type (without much trouble), as an addon even. And invent your own command syntax to perform any operations you like.

I should probably make some console template to show how trivial it is. This is a fun nerdy project - to define some command line to interact with a 3D scene.

The speech recognition part I’m not really convinced about being so useful. But you could always try invent some command line, then use speech recognition later on.

5

Blender, add cube at point 355,322,227 and scale to 1, 4, 6

insert keyframe, move to frame 36

rotate last added item 35 degrees on it’s x axis.

I could see how this would be very useful, especially once Vr becomes a tool instead of a toy.

(viewport oculus rift with 3d mouse glove?)

(doing 1 thing with hands while doing other things with speech?)

if it could do more then one task at once, on other threads ?

(run a second copy of blender with voice?)

I can see possibilities.

also what if your blending and your keyboard breaks?

what if you don’t have hands?

6

Yes, for you and me. I expect it’d be great for a lot of handicapped people though. Assuming it was fairly complete.

7

yes, i use it very often. but if possible i just use ctrl(+shift) increment snapping because for numeric input i sometimes have to remove my hand from the mouse.

i was just wondering if speech recognition could be faster. but i understand the point with latency… if it reacted really quickly i could imagine it as beeing fast but otherwise it could be quite annoying.

(doing 1 thing with hands while doing other things with speech?)

yes, that’s what i also thought. maybe some things could be done at the same time then. :slight_smile:

8

I’ve thought about this too. I would use it for operations that require going through a number of menus.

9

There is a very interesting talk where someone demonstrates how they wrote an interface for VIM which made use of acronyms and placeholder vocabulary that was phonetically unique (but invalid English) as a means for controlling the editor by speech. It seemed to prove rather effective If I recall.

10

You might have been joking but this is the best argument for this sort of input. People with disabilities (maybe reduced dexterity, not necessarily actual missing limbs) might want to use blender, but find it difficult due to the requirement of using lots of keyboard shortcuts.