| 00:03:25 | | Quit speachy (Quit: WeeChat 4.3.4) |
| 00:03:35 | | Join speachy [0] (~speachy@rockbox/developer/speachy) |
| 00:03:36 | Mode | "#rockbox +v speachy" by ChanServ (ChanServ@services.libera.chat) |
| 00:04:09 | speachy | _bilgus: Methinks there is a typo in your last commit |
| 00:04:11 | speachy | https://git.rockbox.org/cgit/rockbox.git/tree/firmware/export/rbpaths.h?id=89b4cd59ef1fcc1a83565dddde4127bb89e8e083#n72 |
| 00:05:13 | speachy | a stray ] in there. didn't cause a build error though, that case might not matter.. |
| 00:13:29 | | Quit othello7 (Ping timeout: 260 seconds) |
| 00:25:25 | _bilgus | indeed wow amazed it didn't barf |
| 00:26:04 | _bilgus | dconrad turns out theres already an event fired that should do what you want add_event(BUFFER_EVENT_BUFFER_LOW, cb_fn); |
| 00:31:26 | _bilgus | only problem is its not consistent enough from my limited testing it doesn't always fire esp with short tracks |
| 00:31:58 | _bilgus | or maybe it does and i'm just not seeing the splash but the other issue is it seems to trigger early |
| 00:32:11 | _bilgus | so maybe another event is in order |
| 01:00 |
| 01:42:58 | _bilgus | messing around with this for awhile its hard to get the end of the track using the buffering stuff because it we are looking at the data in the buffer not the actual fact it has run out |
| 01:56:09 | *** | Saving seen data "./dancer.seen" |
| 03:00 |
| 03:49:19 | | Quit SammysHP (Quit: *wuff*) |
| 03:56:10 | *** | Saving seen data "./dancer.seen" |
| 05:00 |
| 05:56:11 | *** | No seen item changed, no save performed. |
| 06:00 |
| 06:57:55 | | Quit paulk (Read error: Connection reset by peer) |
| 06:58:02 | | Join paulk [0] (~paulk@about/aquilenet/user/paulk) |
| 07:00 |
| 07:19:10 | | Join SammysHP [0] (~SammysHP@faol.sammyshp.de) |
| 07:27:55 | | Join bollerwagen [0] (~bollerwag@p5091c8f9.dip0.t-ipconnect.de) |
| 07:29:38 | bollerwagen | hi... I just wanted to leave some feedback on the wiki. The version at ArchLinux should be updated in the text. I don't want to create a user because of a version number. Thank you |
| 07:51:32 | speachy | bollerwagen: updated how? |
| 07:51:41 | speachy | (and where?) |
| 07:53:00 | bollerwagen | where: https://www.rockbox.org/wiki/RockboxUtility#Download how: 1.4.0 −−> 1.5.1-1 |
| 07:56:12 | *** | Saving seen data "./dancer.seen" |
| 07:59:29 | user890104 | also GetDeb is 404, Gentoo is updated to 1.5.1-r1 |
| 08:00 |
| 08:03:59 | | Join jacobk [0] (~quassel@2601:380:837f:3520:1113:1c75:8d3b:2220) |
| 08:08:08 | | Quit bollerwagen (Quit: Connection closed) |
| 08:22:58 | speachy | fixed the gentoo and arch version numbers. But as the page says, those are not supported by rockbox anyway, so <shrug>. |
| 08:45:39 | | Quit jacobk (Ping timeout: 276 seconds) |
| 09:00 |
| 09:02:36 | rb-bluebot | Build Server message: New build round started. Revision e8975cf85c, 337 builds, 10 clients. |
| 09:02:37 | rb-bluebot | Fix typo inadvertantly introduced in 89b4cd59ef1 by Solomon Peachy |
| 09:14:39 | rb-bluebot | Build Server message: Build round completed after 724 seconds. |
| 09:14:41 | rb-bluebot | Build Server message: Revision e8975cf85c result: All green |
| 09:29:03 | | Quit speachy (Quit: WeeChat 4.3.4) |
| 09:40:43 | user890104 | how do I debug why the target refuses to load a .voice file that I have generated using tools/configure, advanced, voice, then used espeak-ng? |
| 09:41:03 | user890104 | it plays the invalid voice file clip, and that's all |
| 09:41:17 | user890104 | (at least it works) |
| 09:49:32 | rb-bluebot | Build Server message: New build round started. Revision a872e38ccc, 337 builds, 10 clients. |
| 09:49:32 | rb-bluebot | voices: Use espeak, not piper, to build bulgarian voices. by Solomon Peachy |
| 09:56:14 | *** | Saving seen data "./dancer.seen" |
| 10:00 |
| 10:02:13 | rb-bluebot | Build Server message: Build round completed after 761 seconds. |
| 10:02:15 | rb-bluebot | Build Server message: Revision a872e38ccc result: All green |
| 10:03:14 | | Join speachy [0] (~speachy@rockbox/developer/speachy) |
| 10:03:15 | Mode | "#rockbox +v speachy" by ChanServ (ChanServ@services.libera.chat) |
| 10:04:12 | speachy | pro tip: Don't generate one language's voices using a different language TTS. the results are... not good. And ~10x larger than they should be, apparently.. |
| 10:09:49 | _bilgus | user890104, debug menu talk engine stats has an err code that corresponds to talk_status in tree.h |
| 10:11:11 | _bilgus | TALK_STATUS_OK = 0 TALK_STATUS_ERR_OOM = 1 TALK_STATUS_ERR_ALLOC = 2 TALK_STATUS_ERR_NOFILE = 3 TALK_STATUS_ERR_INCOMPATIBLE = 4 |
| 10:14:27 | user890104 | ERR (4), that's strange, both the build and voices are built using the same commit |
| 10:15:34 | speachy | is there a stale voice file lying around for another language (eg English) that might have been used before trying to switch to somethimg else? |
| 10:15:45 | speachy | (just a theory...) |
| 10:20:31 | user890104 | not at all, I built just the Bulgarian |
| 10:21:42 | user890104 | the texts look good so far, with some funny exceptions like USB HID translated to a correct but a completely made up acronym |
| 10:21:43 | speachy | I did a quick test just now with a sim build and Bulgarian voice (using the cmdline tools) which worked fine. sounded awful to my ears but it worked. |
| 10:30:02 | user890104 | I'll try to rebuild it, did you built it separately from the sim, or a sim build with the voice included? |
| 10:32:54 | speachy | the latter |
| 10:34:14 | speachy | I'd suggest grabbing the nightly voice build but that went quite wrong. Tonight's should be good though. |
| 10:36:07 | user890104 | well, my goal is to fix building the voice file locally, so I can modify the phrases that are not suitable and check the result |
| 10:37:03 | user890104 | and since rbutil does not work on my machine because of Qt/HTTPS issues, this is one of the possible ways to do it |
| 10:37:51 | speachy | I _think_ the rbutil ssl issue is due to mismatches between the openssl library it was compiled with versus what it is running with. |
| 10:38:10 | speachy | I'd expect that if you compile rbutil locally it will work okay. |
| 10:38:42 | speachy | but I try to avoid looking at anything C++. :D |
| 10:39:18 | speachy | (in all fariness, rbutil is pretty nice C++ as far as that goes. And Qt's APIs are quite sane) |
| 10:40:04 | user890104 | building firmware with logf caught some format warnings in apps/plugins/lastfm_scrobbler.c |
| 10:43:55 | _bilgus | you will get those everywhere |
| 10:44:08 | _bilgus | sizes change between sim and device |
| 10:44:19 | user890104 | things are breaking in unusual ways here. ROLO gave me a "bad checksum" error |
| 10:45:06 | user890104 | _bilgus: are those format strings supposed to be fixed, or numbers migrated to fixed-size integers? |
| 10:45:26 | _bilgus | printf doesn't have the expressiveness for that |
| 10:47:00 | _bilgus | atm you need ifdefs around logf between 32 and 64 bit builds that sounds terrible you might be able to do something automated but I doubt it will result in anything but heartache and pain |
| 10:47:49 | user890104 | well, if those integers are the same size on every env, their format could be expressed consistently, and the platform's compiler would take care of emulating bigger integer size than the CPU's native one |
| 10:48:24 | _bilgus | then when you go back to the device it shits the bed been here, patches welcome? |
| 10:48:55 | _bilgus | don't touch the ones making files you'll be sorry |
| 10:49:56 | _bilgus | seriously been down this rabbit hole several times you've been warned :) |
| 10:53:13 | _bilgus | you can cast the ones getting on your nerves but I wouldn't leave them that way |
| 10:53:31 | _bilgus | better to have the warning than a hidden silent bug |
| 10:55:37 | user890104 | i see, thanks for the heads up |
| 10:55:42 | _bilgus | anyway I won't tell you not too and it may event work but I have V. bad experiences with this |
| 10:55:48 | _bilgus | even* |
| 10:58:11 | _bilgus | all I can say is if you do you better do it in small chunks and test because when you get to the end and it dies you'll never figure it out and then you'll start over trying to be more careful then next thing you know you'll be ok with the warnings |
| 10:59:17 | _bilgus | RB therapy? |
| 11:00 |
| 11:01:59 | _bilgus | dconrad I tried for a while to find a place in the pcm buffer to get a ground truth on when the audio was idle it works but 100% of the time though and I think whats going on is there isn't actually that often that it happens in a way we can catch |
| 11:02:12 | _bilgus | Not 100% * sorry |
| 11:03:45 | _bilgus | it throws silent samples and I figured that was the perfect place but thats the buffer and we have no provision or way currently to look at the actual audio data in reference to the wallclock |
| 11:04:20 | _bilgus | so we see silent samples get added but we need to do some math to figure out when that will happen |
| 11:05:51 | _bilgus | I think you should be able to watch these silent samples get added and take the current buffer pos - the final and multiply by the frequency to figure out WHEN that silence will play thru the audio system |
| 11:08:19 | _bilgus | so we could send an event SYS_AUDIO_IDLE and the current_tick when it will be true or setup a timer to fire the event at that moment I think the former might be better so you don't miss it |
| 11:09:15 | _bilgus | I'll mess with it and see if I get something working |
| 11:20:13 | | Join dconrad_phone [0] (~dconrad_p@205.237.113.8) |
| 11:22:09 | dconrad_phone | _bilgus: that would be great! that would mean I wouldn't need to try to think through all the possibilities of start/stop/pause, though I think that would work alright as well |
| 11:22:52 | dconrad_phone | tying it to the start/stop/pause events seems to work pretty well also though |
| 11:24:09 | user890104 | the ROLO error was an incorrect target selected in configure |
| 11:24:09 | speachy | dconrad_phone: forgive me for asking the obvious, but can you detect the current jack being unplugged? |
| 11:25:47 | _bilgus | dconrad_phone, taht may be your best bet but i'll try to get you an event, it should be lass overhead (and hopefully code) |
| 11:25:54 | _bilgus | less |
| 11:26:49 | speachy | user890104: wrt warnings in logf, while we do mandate c99 support the format specifiers for those stdint types are awful to use in practice. |
| 11:28:13 | speachy | I _think_ what I usually do is for all unit32_t values that need to be printed, I use %l[udx] and cast the argument to [unsigned] long. |
| 11:28:23 | user890104 | speachy: I can confirm the Bulgarian voice clips sound awful, but the only TTS to produce reasonable sound I can think of is a website with a captcha on every string render |
| 11:28:27 | dconrad_phone | speachy, not with the current state of the patch, though I would like to keep that behavior for the headphones port. the line out port has a gotcha in that you could see a false unplug if you just happened to poll the detection during a large voltage swing |
| 11:29:29 | dconrad_phone | the headphones should be less susceptible I think due to the lower impedance but if nothing is plugged in or a high impedance device is plugged in it will behave the same |
| 11:29:54 | speachy | user890104: does the google tts option sound any better? I prefer to not use those for the nightly builds by default just in case we get blacklisted for making too many requests. |
| 11:30:21 | user890104 | I haven't tried, but I guess they will sound pretty good |
| 11:30:23 | dconrad_phone | well, I should clarify it detects an unplug only while nothing is playing |
| 11:30:31 | user890104 | the website I referred to is: https://freetools.textmagic.com/text-to-speech |
| 11:31:03 | speachy | user890104: I can generate a gtts bulgarian voice file for you if you'd like. let me know the target/type. |
| 11:31:22 | user890104 | I'm testing them on a physical ipod6g |
| 11:32:21 | user890104 | uhm, gtts is not available in the Ubuntu repos |
| 11:32:57 | user890104 | ah, it's python3-gtts |
| 11:33:06 | speachy | it's generating right now on my end. |
| 11:34:57 | speachy | (gtts is also pretty slow due to the round-trip and the need to transcode from the mp3 output to speex) |
| 11:35:43 | user890104 | figured out the latter, when it complained about missing ffmpeg :) |
| 11:39:45 | speachy | ~3/4 done.. |
| 11:40:14 | | Quit dconrad_phone (Quit: Connection closed) |
| 11:41:57 | speachy | user890104: https://www.shaftnet.org/~pizza/rb-ipod6g-bulg.zip |
| 11:42:37 | speachy | it's a complete build of master + the 64-bit storage patch, plus the bulgarin gtts voice |
| 11:42:48 | | Quit braewoods_ (Read error: Connection reset by peer) |
| 11:45:13 | speachy | 4.7MB voice file, about 2x larger than typical. |
| 11:50:35 | rb-bluebot | Build Server message: New build round started. Revision 780c05a64d, 337 builds, 10 clients. |
| 11:50:35 | rb-bluebot | Fix empty string in the Bulgarian translation by Vencislav Atanasov |
| 11:56:16 | *** | Saving seen data "./dancer.seen" |
| 11:56:48 | | Quit SammysHP (Quit: *wuff*) |
| 11:57:24 | | Join SammysHP [0] (~SammysHP@faol.sammyshp.de) |
| 12:00 |
| 12:02:34 | rb-bluebot | Build Server message: Build round completed after 719 seconds. |
| 12:02:36 | rb-bluebot | Build Server message: Revision 780c05a64d result: All green |
| 12:07:08 | user890104 | speachy: according to my girlfriend, espeak sounds like a misguided robot-foreigner, and gtts sounds like a mostly acceptable IVR menu |
| 12:07:32 | user890104 | gtts is a bit slow though |
| 12:08:05 | speachy | generation speed, or spoken "Words per minute" speed? |
| 12:08:17 | speachy | "misguided robot-foreigner", I like that. |
| 12:09:13 | user890104 | words per minute is a bit low, some 3-4 word clips sound way too lazy |
| 12:10:04 | speachy | unfortunately I don't think that can be tuned. |
| 12:10:32 | speachy | I take that back −− there's a "read more slowly" option. :D |
| 12:10:50 | user890104 | the intonation is all over the place, I don't expect it to match the sentence, as there's none most of the time, but it's very inconsistent |
| 12:11:21 | speachy | most TTS engines don't do well with very short phrases. |
| 12:11:55 | user890104 | but it's definetely a big leap forward compared to espeak, it's a pity that it can't be generated offline |
| 12:12:25 | speachy | there are other offline engines but they are a real PITA to deploy. that's one of the things I liked about piper. |
| 12:13:49 | user890104 | I didn't find any Bulgarian piper model, found the instructions to record an own one, but it looks like a lot of time to make it sound good, and NVidia GPUs in order to make it quicker |
| 12:14:11 | speachy | but language coverage falls off rapidly with others |
| 12:14:36 | speachy | there are patches to piper to allow use of ROcM, at least. :D |
| 12:14:44 | | Join braewoods [0] (~braewoods@user/braewoods) |
| 12:15:17 | user890104 | ok, so then my ETH miners can help generating the model |
| 12:15:21 | | Join IPG [0] (~InvoxiPla@2.122.6.66) |
| 12:15:28 | user890104 | (12 x RX580) |
| 12:15:37 | speachy | piper is pretty slow even on synthesis. When I pre-generated the voice pool on my beefy box here, the fans were howling for several hours...) |
| 12:16:40 | speachy | (dual Xeon E5-2690v4, total of 28 cores @2.6GHz..) |
| 12:17:20 | speachy | I guess that's not so impressive by modern standards but it's a _lot_ faster than the VPS that most of rockbox runs on... |
| 12:18:12 | speachy | user890104: in all seriousness, I'm sure the piper folks would welcome a Bulgarian voice model, even a relatively low-quality one. |
| 12:25:30 | user890104 | speachy: looks like the recording and model training are separate steps, so recording a bunch of phrases is definetely possible without beefy hardware to train the model on |
| 12:25:34 | | Part Demosthenex |
| 13:00 |
| 13:09:55 | | Join lebellium [0] (~lebellium@2a01cb0405d07f00d33a0da62bfc2206.ipv6.abo.wanadoo.fr) |
| 13:31:05 | | Join ac_laptop [0] (~ac_laptop@2a0c:e303:0:7300:e29d:31ff:fe2d:a258) |
| 13:34:08 | | Quit braewoods (Read error: Connection reset by peer) |
| 13:34:56 | | Join braewoods [0] (~braewoods@user/braewoods) |
| 13:35:19 | speachy | user890104: so on balance, do you recommand switching the nightly voice from espeak-ng to gtts? |
| 13:56:19 | *** | Saving seen data "./dancer.seen" |
| 14:00 |
| 14:27:43 | | Quit IPG (Read error: Connection reset by peer) |
| 14:58:36 | | Join advcomp2019 [0] (~advcomp20@user/advcomp2019) |
| 14:59:24 | | Quit advcomp2019__ (Ping timeout: 276 seconds) |
| 15:00 |
| 15:27:37 | | Join othello7 [0] (~Thunderbi@pool-100-36-176-164.washdc.fios.verizon.net) |
| 15:38:29 | | Quit ac_laptop (Ping timeout: 248 seconds) |
| 15:56:20 | *** | Saving seen data "./dancer.seen" |
| 16:00 |
| 16:25:23 | | Quit othello7 (Quit: othello7) |
| 16:31:38 | | Join othello7 [0] (~Thunderbi@pool-100-36-176-164.washdc.fios.verizon.net) |
| 16:52:48 | | Join ac_laptop [0] (~ac_laptop@2a0c:e303:0:7300:e29d:31ff:fe2d:a258) |
| 17:00 |
| 17:01:09 | | Quit ac_laptop (Ping timeout: 248 seconds) |
| 17:27:27 | | Quit lebellium (Quit: Leaving) |
| 17:29:17 | | Join Moriar [0] (~moriar@107-200-193-159.lightspeed.stlsmo.sbcglobal.net) |
| 17:56:24 | *** | Saving seen data "./dancer.seen" |
| 18:00 |
| 18:03:21 | | Quit braewoods (Ping timeout: 276 seconds) |
| 18:05:16 | | Quit sebagala (Read error: Connection reset by peer) |
| 18:05:40 | | Join sebagala [0] (~Burak@185.25.123.34) |
| 18:11:27 | | Quit sebagala (Remote host closed the connection) |
| 18:11:51 | | Join sebagala [0] (~Burak@185.25.123.34) |
| 19:00 |
| 19:48:30 | | Join massiveH [0] (~massiveH@2600:4040:a982:dc00:507:72c0:13b:d2cf) |
| 19:49:41 | | Join _bilgus_ [0] (~bilgus@syn-162-154-213-134.res.spectrum.com) |
| 19:50:38 | | Quit Bobathan_ (Quit: ZNC 1.8.2+deb2+b1 - https://znc.in) |
| 19:51:25 | | Join Bobathan [0] (~admin@syn-065-029-248-157.res.spectrum.com) |
| 19:52:53 | | Quit _bilgus (Ping timeout: 248 seconds) |
| 19:56:27 | *** | Saving seen data "./dancer.seen" |
| 20:00 |
| 20:04:10 | rb-bluebot | Build Server message: New build round started. Revision bcaeaa135a, 337 builds, 9 clients. |
| 20:04:10 | rb-bluebot | build-info: Add a trailing newline to the output by Solomon Peachy |
| 20:16:36 | rb-bluebot | Build Server message: Build round completed after 746 seconds. |
| 20:16:37 | rb-bluebot | Build Server message: Revision bcaeaa135a result: All green |
| 20:20:46 | munkis | speachy: uzziyah has an onboard Nvidia GPU if one is necessary for model training. |
| 20:36:36 | _bilgus_ | dconrad (logs) g#5840 |
| 20:36:39 | rb-bluebot | Gerrit review #5840 at https://gerrit.rockbox.org/r/c/rockbox/+/5840 : POC audio idle event by William Wilgus |
| 20:38:06 | _bilgus_ | let me know how that works for you I had a hard time with the time between tracks but putting it just before the codec load doesn' give you much time but it should be idle I hope |
| 20:38:57 | _bilgus_ | you can resgister your event w/e I put it in the audio stuff just because.. |
| 21:00 |
| 21:01:07 | | Join braewoods [0] (~braewoods@user/braewoods) |
| 21:04:52 | | Quit braewoods (Read error: Connection reset by peer) |
| 21:04:58 | | Join braewoods_ [0] (~braewoods@user/braewoods) |
| 21:08:12 | | Join jacobk [0] (~quassel@2601:380:837f:3520:1113:1c75:8d3b:2220) |
| 21:22:09 | user890104 | speachy: I sure do, but it could cause issues with rate-limiting, if these are logged somwhere then give it a go |
| 21:23:06 | | Quit braewoods_ (Remote host closed the connection) |
| 21:36:07 | | Quit speachy (Quit: WeeChat 4.3.4) |
| 21:36:19 | | Join speachy [0] (~speachy@hurricane.shaftnet.org) |
| 21:36:19 | | Quit speachy (Changing host) |
| 21:36:19 | | Join speachy [0] (~speachy@rockbox/developer/speachy) |
| 21:36:19 | Mode | "#rockbox +v speachy" by ChanServ (ChanServ@services.libera.chat) |
| 21:37:03 | speachy | _bilgus_: Got another regression for ya. With c0ac043c6d7b9177 you don't getsecondary volumes displayed at all. |
| 21:37:22 | speachy | reproducible on the x3 sim, just browse to files, and hit 'e'. nothing shows up. |
| 21:37:30 | speachy | revert that commit, do the same, and extdisk shows up. |
| 21:56:29 | *** | Saving seen data "./dancer.seen" |
| 22:00 |
| 22:25:52 | | Quit Moriar (Quit: Leaving.) |
| 22:33:45 | | Quit jacobk (Ping timeout: 276 seconds) |
| 22:56:06 | _bilgus_ | hmm I'll look at it |
| 22:57:39 | | Quit massiveH (Quit: Leaving) |
| 22:57:48 | _bilgus_ | thats.. pretty strange being that its gated on SHOW_PLUGINS |
| 23:00 |
| 23:02:37 | _bilgus_ | oh nm the other commit |
| 23:03:17 | _bilgus_ | ok I figured thi would be an issue but it never was in the devices since they already have the interfaces |
| 23:04:13 | _bilgus_ | should be able to just hardcode it in for the sim |
| 23:26:30 | _bilgus_ | oh nope turns out the ini function just wasn't getting run in the sim |
| 23:43:38 | rb-bluebot | Build Server message: New build round started. Revision 67db38e766, 337 builds, 9 clients. |
| 23:43:38 | rb-bluebot | [Bugfix] initialize the volume names for the sim by William Wilgus |
| 23:49:56 | speachy | ....heard the server fans spool up so I decided to check :D |
| 23:55:38 | rb-bluebot | Build Server message: Build round completed after 720 seconds. |
| 23:55:40 | rb-bluebot | Build Server message: Revision 67db38e766 result: 44 errors 8 warnings |
| 23:56:33 | *** | Saving seen data "./dancer.seen" |
| 23:57:34 | _bilgus_ | hmm |
| 23:58:46 | _bilgus_ | I bet I need ",,/apps/pathfuncs.h" |
icon identifies that the person is a core developer (has commit access).