lusca
GrapheneOSWe've built our own text-to-speech system with an initial English language model we trained ourselves with fully open source data. It will be added to our App Store soon and then included in GrapheneOS as a default enabled TTS backend once some more improvements are made to it.
We're going to build our own speech-to-text implementation to go along with this too. We're starting with an English model for both but we can add other languages which have high quality training data available. English and Mandarin have by far the most training data available.
Existing implementations of text-to-speech and speech-to-text didn't meet our functionality or usability requirements. We want at least very high quality, low latency and robust implementations of both for English included in the OS. It will help make GrapheneOS more accessible.
Our full time developer working on this already built their own Transcribro app for on-device speech-to-text available in the Accrescent app store. For GrapheneOS itself, we want actual open source implementations of these features rather than OpenAI's phony open source though.
Whisper is actually closed source. Open weights is another way of saying permissively licensed closed source. Our implementation of both text-to-speech and speech-to-text will be actual open source which means people can actually fork it and add/change/remove training data, etc.
Ted's Tech Tips@GrapheneOS You guys are the best 🙌
Avi 
🟣@GrapheneOS i could help with spanish and esperanto models if needed
d@nny disc@ mc²@GrapheneOS the "largeness" of language models is precisely a measure of the difficulty to reproduce them. this methodology has some similarities to something i proposed to huggingface a few years back in a cover letter. no surprise to see they were not interested in reproducibility or the scientific method
@GrapheneOS i have also been trying to find similarly motivated people to collaborate with on a research project to reproduce the fawkes facial recognition poisoner upon a mobile device (ideally as an asynchronous but fully local image postprocessing technique) cc @xyhhx @bunnyhero
@GrapheneOS @xyhhx @bunnyhero i have been putting it off repeatedly but the fawkes paper itself is very high quality and imo intended to be reproduced. if there are resources your team has developed or considered regarding modern hardware on mobile phones for statistical training and inference (fawkes especially requires a training step with local user input iirc) it would be tremendously helpful for our goals here.
@GrapheneOS @xyhhx @bunnyhero we obviously expect reduced efficacy vs the SANDlab implementation with GPU acceleration but the math and the code are both very approachable and since its publication we have seen phones add specific "NPU" chips for matmul/etc and this would be a fun way to subvert the utility of "AI" ubiquitization to embed panoptic surveillance
King_of_Ooo@GrapheneOS I replied to one of your posts a couple months ago when yall asked about TTS, suggesting Piper TTS models (https://github.com/OHF-Voice/piper1-gpl). There are def some quality (English) and performant models, though I haven't dug into whether they are truly open source (aka open dataset) or just open weights.
Either way, I am very excited to see more projects by gOS and more quality options in the TTS & STT spaces. People with disabilities deserve equal access to technology, and anything that brings us closer to a world were that is possible is a good thing.
@GrapheneOS i was really impressed with the efficacy and UI of transcribro. no surprise to hear that was the mark of a grapheneos app
luna the doggie 
@GrapheneOS highly interested in seeing high quality open source TTS/STT, great work!
DeTietsch@GrapheneOS wow ! this is great ! Good Work GOS-Team 👍
640KB@GrapheneOS ++ Thanks
Tim Chambers@GrapheneOS Fascinating is the text to speech and vice versa model and code you’re working on platform specific?
@tchambers It's not really platform specific. It currently runs on the CPU but we plan to add TPU support for Tensor and NPU support for Snapdragon in the future. It's made for GrapheneOS and we're not interested in doing any significant work on use outside of GrapheneOS. It will be possible to install it from our App Store on other Android 16+ operating systems but it's not our focus. We're focused on making GrapheneOS better and haven't gotten much out of making stuff available elsewhere.
Rebel Zhang@GrapheneOS What about SherpaTTS? It is of high-quality, support a wide range of languages, and is libre software.
@rebel1725 It's too slow which results in poor usability especially for blind users depending on TalkBack where it adds a lot of latency to every request. We already have an all around better implementation.
Rodirik@GrapheneOS For German there is "Thorsten Voice" with an Open Dataset.
MalMen@GrapheneOS will this enable speach commands on android auto?
@GrapheneOS please, please add German to that list
Tim Makarios@GrapheneOS
This is great news! I'll be interested to find out how well your English speech-to-text model copes with non-rhotic accents. (Think of a posh British "received pronunciation" accent, or anyone from southeast England, or Australia or New Zealand.) Currently, if I want something like Dicio to understand me, I have to put on an American accent, which my wife says is "creepy".
This is great news! I'll be interested to find out how well your English speech-to-text model copes with non-rhotic accents. (Think of a posh British "received pronunciation" accent, or anyone from southeast England, or Australia or New Zealand.) Currently, if I want something like Dicio to understand me, I have to put on an American accent, which my wife says is "creepy".
Lou 🏳️⚧️@GrapheneOS
I have no idea how easy or hard it is to do, but please keep in mind that some of us speak two or more languages and don't want the STT to translate to the phone's language on top of transcribing (the last STT I tried did that, I don't want to update my phone's settings every time I switch chat 🙄).
I have no idea how easy or hard it is to do, but please keep in mind that some of us speak two or more languages and don't want the STT to translate to the phone's language on top of transcribing (the last STT I tried did that, I don't want to update my phone's settings every time I switch chat 🙄).
Phillip Plays@GrapheneOS very very cool!
| 🇺🇦 Kishka 🏳️⚧️ ~ Cat [She/Her]@GrapheneOS Will this eventually support more than English?
Chief Twat@GrapheneOS You guys rock. Period.
@GrapheneOS Fantastic! Will this allow us to use voice input for Android Auto without downloading Google's TTS engine? If not, are there plans to do so? Thanks!
Adam Wysokiński
Terminal TiltThis is great news! Thank you!
@GrapheneOS very good news 👍. Thank you!!!
Breizh@GrapheneOS Well, some times ago, open-source TTS was pretty lacking, but now Kaldi / Sherpa is pretty good, did you check it? If yes, what was the problem with it?
@breizh It wasn't quite good enough and has very high latency which makes it unsuitable for use with TalkBack. We're making this because existing options including Sherpa don't meet our requirements. Otherwise, we could have forked those. It made more sense to make our own instead which we'll be able to continue improving long term. It's similar to our network location and geocoding implementations where we want things done a particular way focused on high quality in all areas we care about.
@GrapheneOS Well, it’s in two parts, the models and the app itself. But I guess even using the models in your own app still had problems.
And personnally I find it better than the Google one, but maybe it’s only some models (since I’m only using french ones…).
Wassil@GrapheneOS happy to help with French
Marlow@GrapheneOS I am very excited to not have to use an external tool to do this anymore.
Cat 🐈🥗 (D.Burch) 
@GrapheneOS How good is this model at meowing?
I am Jack's Found 404
@GrapheneOS Perfect timing! Since december 2025 google text to speach is not working anymore without Exploit protection compatibility mode.
Bart Groeneveld@GrapheneOS made my day! Each time I think GrapheneOS is already amazing, you make it even better! Thanks a lot for your dedication!
Grant Joseph@GrapheneOS Fantastic work as always! Any news on the replacement keyboard plans I've seen mentioned from time to time?
Peter Vágner@GrapheneOS I'm interested in having working open-source #TTS for Slovak language in the future. I my-self am not proficient enough to create my own high quality engine but I think I still might be helpfull. During all these years I've been contributing to eSpeak-ng slovak language support and recently with friends @ondrosik and @asael we are contributing to #RHVoice.
Thus I will be watching the development of your TTS engine to find out if it can be trained using a homelab setup and if adding support for the language is something I might be able to help with.
Thank you for this news and all the hard work you've put into it.
Thus I will be watching the development of your TTS engine to find out if it can be trained using a homelab setup and if adding support for the language is something I might be able to help with.
Thank you for this news and all the hard work you've put into it.
Marcial 🇨🇷🇻🇪@GrapheneOS For those who do not have neither GitHub accounts or crypto, do y’all have plans to set up funding channels through the likes of LiberaPay, OpenCollective, etc?
@moshimotsu We have a bunch of donations options including local bank transfers with very low fees via Wise or sending money via Wise itself which has no fee if it's a matching currency. We already have a lot of options and adding more results in higher costs for managing it and handling accounting. We have to spend a huge amount of money on accounting and also auditing. The more complex the finances, the more money we'll need to spend on that. Why not use bank transfers or Wise?
@moshimotsu GitHub Sponsors provides credit card donations with 0% fees for donations from individuals which isn't something we can get elsewhere. The fees elsewhere are often over 5% when currency conversion is taken into account. We don't want to move more of the donations to high fee platforms. We have PayPal as a donation method but our donate page tries to make sure people understand the high fees and that it's better to donate another way such as GitHub Sponsors, Wise or bank transfers.
@GrapheneOS The processing fee problem is super valid. If people can donate that way, more power to them, but in an ideal world, less techy users who might not have GitHub would still be able to donate through a channel that’s familiar to them.
It might be worth doing a poll on here, just to gauge general sentiment? “To those who don’t donate but would like to, what’s the biggest barrier?” and see if not having those platforms is causing a significant figure to avoid donating.
@GrapheneOS Truth be told, I’ve never heard of Wise until this very moment, and I wouldn’t be surprised if many others haven’t either. But really the reason I ask is because, for many, these are very recognizable channels through which supporting a project “just makes sense.” When onboarding friction is lower, more people usually donate because they can just use a familiar system. I only asked because I felt having those might lead to more donos, rather than personal need!
@moshimotsu LiberaPay isn't used much based on https://en.liberapay.com/explore/recipients. GrapheneOS likely receives more monthly donations than the total amount going through LiberaPay based on past numbers they've given on the total going through it. We were setting up Open Collective at one point prior to having our non-profit but ended up just using our own non-profit instead of needing a fiscal host to handle it. We're on Benevity already because some people can only donate through that but we prefer Wise.
@GrapheneOS Does this follow the Android TTS API? Or does it have it's own?
@draeand It implements the Android text-to-speech API for use as an Android text-to-speech backend.
@GrapheneOS Oh very good, I won't need to add another Android backend to Prism then lol