VVonder

I didn't see this thread, happy to answer any questions. I have a gender from the game database for most NPCs, it picks a base voice at random and applies a random twist to basic knobs (like speaking rate). Then the only trick is making sure that character always gets those same parameters. You _can_ modify them to sound better/different. (and some comments unrelated to Steampunkette) It's hilarious that people are worried about SAG. There is no money here. There never will be. The voices are either built into Windows where Microsoft paid a few shovels of cash to voice actors to develop them or other commercial companies (amazon, google, elevenlabs, openai, etc..) did similarly. As much as I would have enjoyed sampling a certain someone for Aeon I didn't. Do you want human voices? Not a problem. sidekick responds to chat log messages by playing wav files, the AI bit only comes in if the wav file doesn't exist. Go ahead and record all the dialog you want. I'd even be happy to try and make that process easier. The only serious challenge are the dialogs that include the names of your character or the team name of a taskforce. Sidekick is entirely client side. That means no one gets voices unless they choose to. It is based on reading the chat stream, so it induces no delays in gameplay; totally non-blocking. For most of the cutscenes the timing is fairly decent. It also does things like tell you which badge you just got and which level you just reached. Things the vanilla UI doesn't do a very good job with. If you don't like it that is totally up to you. It is opt in, and as far as I'm aware that is not changing. I enjoy playing with voices. In my opinion it makes the game a richer experience. All you have to do is not install it and you'll never hear a syllable of it (or do install it, odds of it not working correctly are quite a bit higher than 0%) Again, happy to answer any questions. The code is all on github. It has been a fun project to build and while I do hope other people enjoy it I made it for myself. I write software for a living but it is backend linux stuff, this multi-process, multi-threaded windows app in python is a nice stretch outside my comfort zone. Even if no one uses it my goals have already been accomplished.

Ok folks, new version is available and it has the most amazing new feature that I'm sure you're all going to love. It actually installs. https://github.com/jason-kane/coh_npc_voices/releases/tag/v4.1.1 Seriously. I setup a clean windows 10 in vmware and tried to install and it blew up in my face which is horribly embarrassing and the code with the error has been there for a _long_ time... so sorry to anyone that tried to install but it failed to work so in a fit of rage you went red side and started killing snake people. Details including the full installation walkthrough are at https://github.com/jason-kane/coh_npc_voices/issues/15#issuecomment-3091475834 It isn't untested, but neither is it -well- tested. So.. have fun out there.

I finally got the latest version (4.0.0) packaged up. This version should be much easier for anyone interested to get working. It's best with a few TTS services configured, but even with just windows voices it's a surprisingly compelling experience. Warning: after you've gotten used to this, playing without the extra audio layer feels a little hollow. The primary technical feature is multiple, simultaneous voices. When a bunch of different people say things at the same time, it's a little muddled and confusing, but no one talks over themselves and the narration generally keeps up with the action. The installation is a zip you can un-compress wherever you want it. Download links and more detailed installation/configuration instructions are here: https://github.com/jason-kane/coh_npc_voices/blob/main/README.md#installation

I messed around with this a while ago: The idea is to watch the logs, grab the stuff that works as spoken audio, translate it, then speak it. It also displays the translated chat stream in a text window. That is step one. Step two is whenever _you_ type something, it sees that, translates it to english and injects it into the chat stream. So when you talk to your team-members they see both what you actually typed and the english translation. Step three is to input the non-english text in the text window and inject only the english translated text. Don't even need to type it, sprinkle some whisper speech-to-text and you can just talk in your native language and read/hear in your native language. The problem, and it may be well within the capability of the Homecoming team to resolve, is that injecting text with it being glitchy is really awkward. I'm stuffing the keyboard buffer, and it _mostly_ works but the timing is hell. If you are typing a message or in combat or just traveling.. throwing a bunch of characters as if the player typed them junks it up. You can minimize the impact with macros, but it gets complicated. If the homecoming team added a reliable way to send chat text, maybe a localhost port? That would be make a client side integration reasonably viable. Note to French speakers: yes, the translations in this clip are really horrible. It is a proof of concept, dropping in a better translation engine/service is the easy part.

New release is available: https://github.com/jason-kane/coh_npc_voices/releases/tag/v3.2 No big features -- mostly incremental improvements. Oh, it will try and send a /tell to itself, it isn't nefarious. It's the best way I have to determine your archetype, level, primary and secondary -- it sends a tell to yourself then catches the $variable expanded version in the chat log. I'm going to try and stick to adding polish and stability to the current features for the next release or two.

The generic approach would look a little like Vortex (https://www.nexusmods.com/about/vortex/); each game would need just enough code to locate the right logs, tail them properly, differentiate the things that should be spoken from the things that can be ignored and choose voices for each speaker. I fiddled with reading from CoH memory instead or in additions to logs; I want to get current xp, arch/primary/secondary etc.. but I haven't gotten that working. I can send key presses to the game, I've only barely scratched the surface of the options that opens up.

Demo of the latest version: I'm on the fence right now reg. the "You have defeated" messages. It's amusing enough I'll at least make it an option.

Thanks; I'm making steady progress. Latest new stuff is https://github.com/jason-kane/coh_npc_voices/pull/12 I rewrote all (well, most) of the widgets to use customtkinter instead of vanilla tk/ttk. I added Azure and OpenAI voice support. Some simple quality of life things; like when you level up it says congrats and tells you what level you've just reached. Some more utility things, like the damage table, it's for the current session: (^ I've fixed the accuracy numbers and 'special' cosmetics); this is to provide you some hard numbers to decide on the best slotting based on your play style. I need to give the latest build two or three more game sessions to make sure it is stable and I'll cut a new release.

In Yin TF: 2024-07-05 18:32:12 [Caption] <scale 1.75><color white><bgcolor FireBrick>Oh no... I didn't check to see who ELSE was in there with Alderman! I'm so sorry, TaskForceName! (should be the replace-with-taskforce-name variable)

And in French, with less stuttered/jumpy but unfortunately very blurry video https://github.com/jason-kane/coh_npc_voices/releases/tag/v3.0 updated with new windows installer

This again? This is the text to speech system, but with a translation layer. You select a language (Chinese, French, German, Japanese, Korean, Spanish), and it will translate every [NPC], [TEAM] and Tell message and speak it out loud. It's a bit of a parlor trick, but it may really open the game up to people with weak English.

With apologies for raising a long dead thread from its grave; a good percentage of related searches land in this thread. I've got this more or less done; computer generated text-to-speech for all NPCs. It's all client side as an agent reading the chat log in real time. It doesn't block (obviously) so game play never waits for it to finish talking. The current version supports Windows TTS (the not great thing that comes with windows), Google text-to-speech, Amazon Polly and ElevenLabs. The theory is if you setup a bunch of different services and save the resulting sound files locally then your usage will stay down in/near the free tiers. Curiously enough the timing for cut-scenes works out really well in most cases. Shockingly better than would be expected. I also decided to 'speak' player messages in the team channel. That feels pretty good. Some in-game events are audibly highlighted -- badge awards, players joining/leaving the team. Demo vid so you can hear; it's gotten better since I made this clip a few days ago. It now does a much better job making voices more distinct. I have made some minimal efforts at a windows installer: https://github.com/jason-kane/coh_npc_voices/releases/tag/v2.0.0 If it doesn't work for some reason please let me know. If you have previous installed it you will probably have to delete the "voices.db" file since I haven't bothered to implement a proper migration.

New demo; much better voice quality and some visibility of the UI. It can now use (all at once) ElevenLabs, Amazon Polly, Google Text-to-Speech and Windows TTS. Yeah, you can't actually hear me talking for a lot of this because I didn't balance the volume very well. Bonus. I still have some bugs to squish before I cut another windows installable release, maybe this weekend. But.. it's pretty slick now and once you're used to it the game feels a little hollow without it.

Thanks. Right now it has Windows TTS, Google TTS and ElevenLabs, the next easy one is Amazon Polly. I'm interested in getting Coqui working too. That would give us another local option but I'm a little worried about resource requirements. Should be a fun experiment. The next challenge is that having a bunch of perfectly smooth and clear human voices is great, but many of the foes in-game are very much non-human. An easy example are the snakes in COV. They sound terrible because the text-to-speech is trying to pronounce ssspeach like thisss and gets it all wrong. Most TTS engines support something called SSML which has a phoneme tag that I'll try and manipulate to get the right sounds.

Totally agree; Windows Text for normal human voices is pretty bad. There is support for both google text-to-speech (better) and ElevenLabs (much better) but it is hard to setup in the current release. The next will be much better (maybe this weekend); I've added a configuration panel for setting the elevenlabs api key (they allow for a large amount of text-to-speech each month on a free account). You will also be able to set the default engine to use for unknown voices both for NPC and players. For NPCs it is easy because they are repetitive and easy to cache to disk. So a high quality voice is easy to justify. For players the cache is rarely useful and a little creepy, so I disabled it. I'm cheap-as-hell, so I'm making it easy to use Windows TTS for players (because it is free and fast) while using the slivers of pennies high quality services for NPCs. One case where Windows TTS is more than fine -- Clockwork. If anything it is not robotic enough, so I'm wrapping it in vocoder effects. It's pretty good. Oh, I can attach.. a work in progress, but here are "Prototype Oscillator" and a Vahzilok "Reaper"; I'm reasonably happy with both these. 95c4f_Pieceremov.mp3 b7fc7_Itsalwayse.mp3