Hi John, another thing, tiles-shell isn't working for me. Is there something else I need to do before running `sudo chsh -s /usr/bin/tiles-shell $USER`? By not working, I mean I'm not able to do anything in tiles after running this command as far as I know. I tried using suck+1/2/3, suck+space, suck+m, and suck+alt+\ but nothing happened. I also tried replacing suck with alt but no luck. After booting in recovery mode as root, running `sudo chsh -s /bin/bash <my user name>`, and rebooting all is working again. Maybe this is related to another problem I've had since installing tiles (discussed below). Text in tiles-terminal is disappearing (ex. text getting cut off at the edge of the previous border before it was re sized), text appearing (some of the text from last vi file that was opened after quitting, past commands, etc.), etc. when windows get re sized in any way (this is unpredictable behavior; the text doesn't get messed up every time after resizing, but often). This makes it difficult or impossible to read the output of commands and difficult or impossible to read what you just typed (sometimes it will show that you typed something but you really didn't; but if you begin to type a command with the command/text on the same line, only the command you typed will register after hitting enter). This also interferes with backward search as you would expect. A possible suggestion I just thought of: sometimes it's hard to read the coordinates to click with your mouse. A command to change the font color may be helpful. Perhaps yellow would be the next best alternative. I'm still in the early thoughts of how to make this better. Maybe a command to change the font or the style as well such as bold, italic, etc. Not a big deal though. Hope you're doing well man. If I don't hear from you soon I hope you have a nice holiday season. Not religious myself, but I still have a good time with family and friends :).
Hi sublime, just a quick reply but thank you and I hope you've had a nice time too :) Oooof yes I should have said you need to add /usr/bin/tiles-shell to the /etc/shells files before you run chsh. That said, I really thought it just complained if you hadn't. Well done with the recovery shell but if it happens again, you also could fix it logged in as another user (e.g. root) at the virtual console. Text getting cut off when resizing is normal urxvt behavior but it is annoying and I've been looking at alternative terminals I could get working with keyboard selection, for other reasons too. Yeah you're right about the coordinates. You could have a look at `coord --help` and see if you can get something you prefer with what's there like: coord --color purple --font-size 11 As a hack, you could override the command with a script at /usr/local/bin/coord like: #!/bin/sh exec /usr/bin/coord --color purple --font-size 11 "$@" Because /usr/local/bin has priority over /usr/bin. (Or use your own sxhkdrc but I wouldn't for that alone.) You could still run the original with `/usr/bin/coord`. I'll probably increase the font size to 11. Also I did pretty big boring changes to numen's implementation before I was away. My `numen --version` is: 0.4-7-g28f57b63b506
It's been good, thank you John. tiles-shell still isn't working after adding /usr/bin/tiles-shell to /etc/shells. Do you have any other ideas to get it working? Glad to know it's not just me seeing text get cut off when resizing with urxvt. My current workaround is to just keep my terminals at half screen or full screen, resize as little as possible and utilize more desktops and workspaces. Font size of about eleven or twelve looks just about right. But I already have a python file at /usr/local/bin/coord with the following shebang line: #!/usr/bin/env python3 How should I override the command in this case? If increasing the font isn't enough then I may change the color with commands using sxhkdrc. Another feature that would be nice if it's not too difficult to implement would be having the alternative transcripts (brought up with saying 'change' or 'transcripts') first letter capitalized if you just said 'transcribe' and the letter i capitalized. Then you wouldn't have to manually edit the changed transcription. Maybe there's a workaround for this currently?
Aye hi sublime :) tiles-shell is a configured zsh so make sure you have zsh and zsh-autosuggestions installed. See if you get any errors when you just run the command: tiles-shell Oh yeah I must have installed coord to /usr/bin rather than /usr/local/bin. You could do the same: sudo mv /usr/local/bin/coord /usr/bin I've been working on numen to support multi-word phrases, and then to add phrases like "dash scribe" to-transcribe-like-this, and I'll try to improve "change" at the same time.
I've created a branch for numen called multiword if you'd like to test the new changes with me.
Hey John, I didn't get any errors when running tiles-shell. I just didn't have zsh and zsh-autosuggestions installed. That being said, I think I might put tiles-shell on the back burner for now. Some stuff I read online seemed to suggest that I should know more about shells before trying zsh. I probably could go either way but I do know some bash already and I think it may be beneficial to learn and relearn more bash first before trying other shells so I can better compare them and then find what works for me best. bash is the only shell I kind of know currently and I found some really useful syntax for searching history and auto completing past commands with a safeguard so I don't blindly execute commands. I read zsh is most like ksh (I have no idea what that is other than a shell) but borrows the best from many shells. If in your opinion it's similar to bash in many ways then I might try tiles-shell sooner. But as of now I think I'm starting to get settled in to my setup and I will be able to spend most of my time figuring out what to do with my life in a month or less. After that I can start preparing for whichever path I choose. I think I know some fields I want to go in but a lot of that is based on information 3+ years ago so I'm going to retrace my steps and combine it with new information. Thinking ai, security, or web dev but there's so much in those fields. I may change my mind now that I'm 3 years older and after trying more projects, jobs, etc. in those fields. The coordinates are easier to read now after using the mv command on the coord script like you said. 11 is perfect for the font size. 12 makes the coordinates on the right a bit too crowded I think. I tried the multiword branch but I wasn't able to use 'change' or 'transcripts'. When I would attempt to I would get this error: /bin/sh: 1: /usr/libexec/numen/phrases/transcripts: not found That would be cool to use multi-word phrases such as dash scribe. Looking forward to the changes :)
Hey sublime, Good the shell works, and I wouldn't worry too much about bash vs zsh, they're the same idea. They're both POSIX shells, which means they implement the unix spec and can execute #!/bin/sh scripts like all the other POSIX shells, so you're learning *real* shell syntax. What you learn with zsh applies to bash and sh. The main benefit of zsh is the interactive use, nicer tab completion etc., at least once configured like with tiles-shell. Glad the coordinates are better and I agree about the crowding. Make sure you ran `sudo ./install-numen` or goblins will sour your milk. I know nothing about jobs and that, I'm hoping to get somewhere writing free software. qutebrowser and its success with donations gives me hope. Besides numen, I'd like to write a game and the software to make it possible to create and play it handsfree, but I might need to think ahead more.
Hey John, Appreciate you explaining more about zsh. I feel more comfortable using it now that you've cleared things up :) Those pesky goblins are still souring my milk somehow. Still getting the same error with 'change' and 'transcripts' after reinstalling. I don't know much about jobs in tech myself. It will be my first when I land one. I would also like to write free software or at least open source or something along those lines. Bottom line, I don't want to take advantage of people, however I can achieve that. I'm not a philosopher, professor of ethics, security specialist, or economist but transparency is the biggest thing I think. At least then people can make up their minds for themselves if everything is written in clear and simple language that anyone can understand quickly. I think free and open source models are the future and one that will eventually be expected by end users once they understand what's actually going on. Donations will likely be a big part of that for a long time. Especially if governments continue to not lead the way. If you make a game, I would love to try it. I used to play games quite a bit but mostly gave them up years ago. I became pretty busy and my RSI didn't help. Going to try some voice recognition games eventually. I have a few picked out. Don't give up hope, it sounds like you're on the path to salvation.
Thanks sublime, heading to bed but I'd missed a file and "change" should work now for realsies after pulling and reinstalling.
Change is working now, thanks! Will let you know how it's going after some time.
Hi sublime, hope it's going all right. I just wanted to say the multiword branch now uses a new version of dotool, so if you pull numen, run: sudo ./install-dotool.sh; sudo ./install-numen.sh I added keyhold to dotool for setting how long to wait between press and release (whereas keydelay sets the delay after each key) and I've tweaked the values while testing it with the only program I know that was missing keys. And completely guff but I put numen on my phone after messing with it big time: https://fosstodon.org/@geb/109654796143028017
It's going good, thanks John. I'm enjoying the multiword branch so far. Haven't ran into any issues with it. Just pulled and reinstalled. Thank you for the heads up. I made a small script to update all of my git repos which is really handy. That's very cool you have numen and what looked like tiles or a modified version of tiles on your phone now too. Eventually I will get a new phone and likely a raspberry pi. I'll be looking for a FOSS solution too. I had a phone running Replicant OS from technoethical but had a lot of trouble with it. This was about 5 years ago though so the FOSS phones have probably come a long way since.
Whoops, after rereading your post I see you're running sxmo environment, not tiles or a modified version of tiles. Will take a look at it. Looks to provide similar functionality to tiles in your video.
Glad it's going good, sublime. That phone test took a lot of goofing and my phone is barely supported by the os and doesn't even function as a phone or have audio or wifi. It was me trying postmarketos for the first time. They have the good list of device support here: https://wiki.postmarketos.org/wiki/Devices, but you'll see barely any of them are really ready for everyday use. Getting numen and tiles packaged for it would be cool though.
Interesting, thanks for sharing. I will keep postmarketos in mind as an option when looking for a new phone. The biggest issues I had with Replicant OS and technoethical's phone was battery life (ex. when unplugged it would lose 20% while I'm sleeping with nothing running on it) and calls often not coming through (not good when you're looking for a job haha). Support wasn't very good either but you kind of know what you're getting into with linux phones being new and a business with limited staff. Again, this was about 6 years ago so hopefully things have changed.
(resending without the image attachment because it made the mailing list grumpy) Hi sublime, a bit back I installed ubuntu on a wonky laptop and was able to replicate issues you mentioned. I got the same display manager infinite login loop problem. I haven't worked out how to stop gdm barging in but switching the display manager to lightdm is a workaround. With lightdm, Suspend/Hibernate should work now too (turned out zzz is a void linux util). Currently you need to type in your password which is a bit annoying. Also I was a nugget about Dragonfly, you were right. Using it or copying some of its kaldi internals (maybe with vosk instead) could be a possibility for the future, but it would be a lot of work with a lot of uncertainty and a couple downsides from the start. I still think it's best to stick with vosk for the foreseeable but we are pushing its features to the limit and it's lacking in some places. I've been tweeking multiword a lot and I'll hopefully have polished it and merged it into master next week. Messing with an art program too, partly for the game, but still extremely extremely prototypey. I'm also going to look into this noise control for realtime stuff: https://github.com/chaosparrot/parrot.py the videos are very impressive.
Hey John, good to hear from you. I might have a look at lightdm sometime. I appreciate you getting back to me about it. But with slock setup to run on suspend (when I shut the laptop lid) and the ease of running `slock` in the terminal I'm sitting pretty comfy right now. :) Do you know when the new model of Vosk is planned to release? I don't know much about any of the projects other than surface level stuff. But I combed through old links and came across some new links that may interest you: - Question 8 explains the difference between Caster and Dragonfly which is probably a good place to start. The next link below helps explain this answer. https://caster.readthedocs.io/en/latest/readthedocs/meta/Caster_FAQ/ - "The short version is: CCR allows you to speak sequential commands without pauses between them, greatly speeding up your ability to use commands in general." https://caster.readthedocs.io/en/latest/readthedocs/Rule_Construction/Advanced_Caster_Rules/CCR/ - There's also the licenses for the non original code at the bottom of this page. https://github.com/dictation-toolbox/Caster/blob/master/LICENSE - https://github.com/dictation-toolbox/Caster/blob/master/requirements-mac-linux.txt - A site called handsfreecoding.org has some posts about Caster and Dragonfly. I'm not sure which posts are applicable anymore but there's a posts page and you can search the website for these terms. - Dragonfly has eye tracking and OCR (Optical Character Recognition is the process that converts an image of text into a machine-readable text format.) capabilities. The gifs at this link and the video included are interesting: https://handsfreecoding.org/2022/11/27/gaze-ocr-talon-support-and-10-new-features/ Looking forward to the new multiword changes and your upcoming game. What kind of game are you making to integrate the art program into? It does seem like the multiword branch is working better than the master branch. Ex. I didn't have to adjust the typedelay once you added the 'keyhold' and 'typehold' variables. I haven't been getting any gobbledygook (misspellings and typos) since changing typedelay on master and switching to the multiword branch. Also seems like I'm experiencing less random typing after saying scribe but before I say anything else. Haven't heard of parrot.py, but I agree the videos were impressive. From looking the project over in summary it looks to have the potential to be very useful.
Another reason it may seem the multiword branch is working better may be the arm and pop filter I bought for about $30. Or I could just be getting lucky lately, but it does seem like the accuracy has been a bit better too. Still just using the Blue Yeti Nano but I noticed my voice was noticeably clearer with the arm and pop filter addition (as opposed to having the Blue Yeti Nano on the desk in front of me and slightly to the side and all other mics I tried w/out an arm or pop filter but those mics were double the price). I have the mic set up directly in front of me at about where my neck and chest meet so I can see all of the monitor. There's about 2-3 in. of space between the pop filter and mic. Although, I've read you can have that space as much as 6 in. but you would have to adjust your mic gain. I used these commands to test my voice with different mics, different positions, and different levels in alsamixer: arecord --duration=5 --format=dat test-mic.wav aplay test-mic.wav I've saved all of my .wav files so I can hear how much clearer my voice has been getting over time. I still haven't wrote udev rules but I might eventually. I'd say it's worth a shot to pick up an arm and pop filter and see what you think if there's more to be desired from the accuracy. Or maybe check back once I've tried more.
Hi sublime, thanks for links and confirming all's good. I've gone ahead and merged the new changes into the master branch, so `git switch master`, pull and reinstall. I don't know when about the model but this is the vosk developer's blog: https://alphacephei.com/nsh and says he wants to catch up with other models here: https://alphacephei.com/nsh/2022/10/22/whisper.html I'd love it if numen got some love on sites like handsfreecoding, I sent him an email way way back and before I rewrote numen but I didn't get a reply. The game's to be a 2D physics platform (a bit like LittleBigPlanet if you know it) where the levels are just text that could come from files, the web or scripts. I don't play games now but I like the idea and don't think there are many free software games. Glad you're liking your Blue Yeti and that's all good info. I could do with mine being a bit more Nano!
Hi John, no problem at all. Back on master; thanks for the heads up. And interesting links, thank you for sharing. I'm still new to the the speech recognition scene but I think you're on your way up to more recognition. I certainly appreciate what you're doing. :) Not familiar with LittleBigPlanet but I will give your game a try once you release it. I don't think I know of any free software games either so it'll be refreshing to see yours. For the price, I think the Nano has been treating me very well! I'll let you know if I find out anything more that may help with your setup.