jbrew
commented
7 years ago Update on (1): I think it was coincidental that the issues arose just now. That error seems to have been the script's inability to handle an empty input, not anything to do with your changes.
On Sat, Oct 22, 2016 at 11:52 AM, Jamie Brew brew.jamie@gmail.com wrote:
Hi Matt. Thanks for taking an interest in this! I'm fairly new to github and am not a highly experienced coder, so you might have to be patient with me on things like testing, and conventions in general. I'm learning slowly.
A few things:
- I've received a couple of troubleshooting messages in the last couple days about an error that's cropping up in the user loop. Here's one such open issue https://github.com/jbrew/pt-voicebox/issues/16. Since this is the first time that error has shown up, I'm assuming that this is related to the modernization tweaks you made, and wondered if you had any insight into what might have happened.
- Re: splitting the generation and UI into different classes, I think that idea makes sense. Before you do that, though, I think you should check out this repo, called stereotype https://github.com/jbrew/stereotype, where I've been developing a tkinter interface for the program. I think it still has the same "interweaving" that you described, I think the one-touch GUI is more representative of how this thing is eventually going to work.
There are some important differences in the architecture in stereotype. Instead of voices, there are channels. Each channel contains just one corpus, and has an adjustable weight. If someone wanted to achieve the same effect that the voices in this version achieve, they'd be able to do that by saving a collection of weight settings, rather than creating an object that contains several corpora.
- I'd also like to ask your advice about the current repo arrangement. I'm keeping a separate repo for the GUI for two reasons: (1) It's out of the way and not highly publicized, which lets me tinker with it without the stress of troubleshooting requests. (2) I'm hoping to develop a version of the GUI commercially, so wanted to keep the open source voicebox repo available for people to play with, while developing the Apache-licensed version on the side. Is any of this contradictory? Should I just bite the bullet and sign up for a private repo if I want to do any such private development?
Let me know your thoughts on any of this, and thanks again for reaching out!
On Thu, Oct 20, 2016 at 6:19 PM, Matt Cooper notifications@github.com wrote:
This is a cool project :) It's big enough to do something interesting but small enough that I can wrap my head around most of it.
As it stands, Voicebox is both the editor UI and predictive text generator. I'd like to separate them so that perhaps additional frontends could be developed. Also, this should help with testability.
Opening an issue to discuss, since it would be a somewhat radical refactoring of Voicebox. @jbrew https://github.com/jbrew do you have any opposition to me attempting this split? I'm thinking I'll turn Voicebox into the owner of generator state and add a new editor module.
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/jbrew/pt-voicebox/issues/15, or mute the thread https://github.com/notifications/unsubscribe-auth/APtUnkGu3wZI7cYNtyWY4-uErNTOql49ks5q1_bvgaJpZM4KctMa .
vtbassmatt
This is a cool project :) It's big enough to do something interesting but small enough that I can wrap my head around most of it.
As it stands, Voicebox is both the editor UI and predictive text generator. I'd like to separate them so that perhaps additional frontends could be developed. Also, this should help with testability.
Opening an issue to discuss, since it would be a somewhat radical refactoring of Voicebox. @jbrew do you have any opposition to me attempting this split? I'm thinking I'll turn Voicebox into the owner of generator state and add a new editor module.