kind plea from visually impaired user

[JanuszChmiel]

Dear MR kaytat
Do you think, that there are some good Pulseaudio buffer values for Espeak speech engine, more precisely for Espeak-ng? FOrtunately, your professionally created advices on your block are still compatible. So if I run Pulseaudio on Termux but not Pulseaudio which have been prepared for Termux, but officially coded Pulseaudio for Arch Linux latest version, it is ossible to really stream sound from this server which is executed by using Termux and Proot.
But I AM facing big issues with long delay after Speech is being produced.
I AM using Orca screen reader and Speech-dispatcher speech server. The sad reality is, that Huawei have made something which causes that there is some delay after Android kernel driver will play some chunks of sound.
Even when I Am using PUlseaudio for Termux which uses Open sless. It is thanks to Huawei kernel upgrade. Version 7.0 of Android did not caused this long delay.

So I Am thinkink, if it would help if your Simple protocol player would stop to use Audio track and some advanced NDK related techniques to play audio with The lov audio latency would be used.

Thank you very much for your answer.
Reality is that your used technique is stable and do not produce issues.

[kaytat]

I have not tested this end to end with accessibility in mind and so I appreciate you bringing this to my attention.

Using something like the NDK or MediaCodec for low latency audio would require a re-write of the audio processing in the app so it's a big job. But it's something I could look at.

But I'm still unsure where the issue is. You mention a lot of Linux-specific utilities as well as Huawei/Android. I can't really tell if the issue is with the buffering on the Linux side or buffering on the Android-side.

Could you clarify that?

[JanuszChmiel]

I AM afraid, that I will have to buy new phone when I will save some money for it. Because as A paradox. My Huawei 10 Mate lite with Android 7.0 had better audio latency. I AM also using Pulseaudio for Termux. It emulate opensless audio device if The name is correct. But Android 7.0 had other issues related to file systém related operations when using Proot. May be that even Pokedata function did not work as good as it should for Proot needs. My phone is also suffering because of The file system filenames case partial insensitivity. As A result, kind MR Bednarsky had to invented special work around which allow Me to use Proot to work with files which uses also Big characters on their names. Several years ago I have also tried to use Termux and ArchLinux with Proot on old Blackview A5 device. Audio latency was perfect, kernel did not have issues with systém calls which are being used by Proot. But 1 GB of Ram is too small for running Orca, mate-session and Seamonkey. So may be, that my future new phone will be Blackview. I only have to hope, that Android kernel will work according to The needs of Proot and that Proot will not be forced to use special work arounds to allow easy file operations.

[JanuszChmiel]

I would like to express my deepest possible appreciation to you. Because GUI of your app is 100 % accessible. It is even possible to specify The buffer lenght by specifing its lenght in Milliseconds. So very well done. The solution will be to buy some Blackview phone. I have also noticed, That Espeak-ng or Espeak engine produces more than one Second delay even when executing it directly from Termux. So Proot routines are not plaiyng role on this problém. It is very probably some complex kernel driver problém. SO I must accept it, because rooting modern phones by using some apps is rather impossible. I have also found out, that it is not good idea to buy rassberry PI board. Sure, I could execute Whole Linux kernel, but I would like to have external microphone, or USB headset. Also if CPU is working too much, build in routineto prevent board overburning slow down CPU speed. So it is still much more better to use Android phone. Because Android kernel and other services are taking care about board temperature. And I think, that phones are constructed more safely to prevent some overburning. Sad reality is, that if I were sighted person, I could use Mate desktop and many apps including Libre office without issues at all. Thanks to Xvnc and Bvnc PRO for Android. My issue is, that audio apps are having some delay before audio subsystem get initialized. I have dreamed about Linux kernel based on normal Ubuntu Mate, which would run on my Huawei Mate Lite Because Ubuntu touch is not accessible at all for now. Sure, I would not expect touch navigation on screer for blind people. I would only use Phone as a mini computer. I could use Bluetooth external keyboard. I would even live without phone calls and SMS support. But unfortunately, I would have to many thousants of US DOllars to convince someone who would build a kernel with The corresponding device chips driver for Huawei 10 Mate Lite. The most sad reality is, that build in boot loader may do not support boot regular Linux kernel from Ubuntu Mate AArch64. So this is The reason, why I Am using Termux, Proot and Pulseaudio for Termux. There are officially developer Linux phones with regular Linux, but it is too expensive, 400 or 500 Dollars. And my goal is not to use some stripped down version of Linux on phone without Orca Speech-dispatcher and Speech engines. But I know, that I must only dream about it.

[kaytat]

I've added two new options.

1. Performance mode low latency
   This is specific to Android O but the documentation here suggests a low latency path for output

2. min AudioTrack buffer
   To minimize buffering, I've added a checkbox for "Min AudioTrack buffer" This option totally ignores the other buffer size option and tries to use the smallest allowable buffer size. However, this might be glitchy.

I have no yet published this to the play store.

[kaytat]

Android suggest using Oboe for low latency audio but that's going to be a bigger project.