For my non-commercial curiosity, I'm interested in turning some of Lewis Carroll's work into machine-generated speech. When sending the output to an audio device, the
say command can do this even with very large amounts of input:
$ wc ~/Downloads/lewis-carroll.txt 7066 55439 311589 /Users/xxxx/Downloads/lewis-carroll.txt $ date; time say -f ~/Downloads/lewis-carroll.txt; date Wed Oct 3 00:24:38 EDT 2018 real 368m11.986s user 0m0.009s sys 0m0.011s Wed Oct 3 06:32:50 EDT 2018
When sending the output to a file, it appears to also work with very small amounts of input text:
$ date; head -c 310 ~/Downloads/lewis-carroll.txt | time say -o lewis-carroll.aac; date; ls -l lewis-carroll.aac Thu Oct 4 08:46:18 EDT 2018 0.17 real 0.08 user 0.02 sys Thu Oct 4 08:46:18 EDT 2018 -rw-r--r-- 1 xxxx staff 81426 Oct 4 08:46 lewis-carroll.aac
With any more input, it doesn't work:
$ date; time head -c 311 ~/Downloads/lewis-carroll.txt | say -o lewis-carroll.aac; date; ls -l lewis-carroll.aac Thu Oct 4 08:46:40 EDT 2018
^C real 0m30.243s user 0m0.090s sys 0m0.028s Thu Oct 4 08:47:11 EDT 2018 -rw-r--r-- 1 xxxx staff 80865 Oct 4 08:46 lewis-carroll.aac
That's only the beginning of the problems. Further attempts to run
say also hang, no matter how small the input or where the output is supposed to go (e.g.,
say Hello). Worse still, as soon as the first
say command hangs, Chrome starts beachballing. Happily, there's a straightforward workaround to return the system to normal operation:
$ pkill speechsynthesisd say
(Workaround found here.)
I can understand Apple wanting to put rate limits (or something) on speech synthesis, to prevent people from generating cheap audiobooks. (Which would be fine; that's not what I'm trying to do.) This would be a pretty awful way to implement rate limiting.
I don't understand this failing so badly that other software (such as Chrome) gets messed up.
I did something like this back in 2012 (on a couple of kilobytes of text) without running into anything like this. I don't have enough history to reproduce that.
Is there any way around this mess?
Run on: macOS 10.13.6 (17G65)
Like @ashley, I'm able to convert a big chunk of the dictionary to speech:
$ date; time head -c 2000 /usr/share/dict/words | say -o words.aac; date; ls -l words.aac Mon Oct 8 09:54:50 EDT 2018 real 0m2.552s user 0m0.555s sys 0m0.111s Mon Oct 8 09:54:53 EDT 2018 -rw-r--r-- 1 xxxx staff 543542 Oct 8 09:54 words.aac
Looking more carefully at my input, I discovered it was in DOS format (lines end with CR-LF) instead of macOS's native Unix format (lines end with LF). I made a copy in the latter format, removing six CR characters from the beginning of my file ... and now
say can handle six fewer characters before hanging:
$ date; head -c 304 ~/Downloads/lewis-carroll-lf.txt | time say -o lewis-carroll.aac; date; ls -l lewis-carroll.aac Mon Oct 8 09:49:51 EDT 2018 0.18 real 0.09 user 0.02 sys Mon Oct 8 09:49:52 EDT 2018 -rw-r--r-- 1 xxxx staff 81426 Oct 8 09:49 lewis-carroll.aac $ date; head -c 305 ~/Downloads/lewis-carroll-lf.txt | time say -o lewis-carroll.aac; date; ls -l lewis-carroll.aac Mon Oct 8 09:49:55 EDT 2018 Command terminated abnormally. 29.72 real 0.09 user 0.02 sys Mon Oct 8 09:50:25 EDT 2018 -rw-r--r-- 1 xxxx staff 80865 Oct 8 09:49 lewis-carroll.aac
(I'll add more info about my input at the end of this question.)
As to @ashley's other suggestions:
I've tried breaking down the input into smaller files and converting them individually. This is very helpful in exploring what I'm playing around with. I need to jump through many, many hoops to make this work. (I can document this further if it helps.)
I was hoping to do all this from the command line, without resorting to audio capture. It may end up being my best option for making one big audio file.
I can reproduce this problem with the Alex voice -- the default for me, and my preference at the moment -- but not the Daniel voice (though I get 79699 instead of 69867):
$ date; head -c 305 ~/Downloads/lewis-carroll-lf.txt | time say -v Daniel -o lewis-carroll.aac; date; ls -l lewis-carroll.aac Mon Oct 8 19:53:11 EDT 2018 0.68 real 0.08 user 0.03 sys Mon Oct 8 19:53:11 EDT 2018 -rw-r--r-- 1 xxxx staff 79699 Oct 8 19:53 lewis-carroll.aac $ date; head -c 305 ~/Downloads/lewis-carroll-lf.txt | time say -v Alex -o lewis-carroll.aac; date; ls -l lewis-carroll.aac Mon Oct 8 19:53:21 EDT 2018 Command terminated abnormally. 21.75 real 0.08 user 0.02 sys Mon Oct 8 19:53:43 EDT 2018 -rw-r--r-- 1 xxxx staff 80865 Oct 8 19:53 lewis-carroll.aac
This suggests an obvious workaround.... I'll try it in a little while.
Here's my current input:
$ head -n 11 ~/Downloads/lewis-carroll-lf.txt Alice was beginning to get very tired of sitting by her sister on the bank, and of having nothing to do: once or twice she had peeped into the book her sister was reading, but it had no pictures or conversations in it, 'and what is the use of a book,' thought Alice 'without pictures or conversations?' So she was considering in her own mind (as well as she could, for the hot day made her feel very sleepy and stupid), whether the pleasure of making a daisy-chain would be worth the trouble of getting up and picking the daisies, when suddenly a White Rabbit with pink eyes ran close by her. $ head -n 11 ~/Downloads/lewis-carroll-lf.txt | od -c 0000000 A l i c e w a s b e g i n n 0000020 i n g t o g e t v e r y 0000040 t i r e d o f s i t t i n g 0000060 b y h e r s i s t e r o 0000100 n t h e \n b a n k , a n d 0000120 o f h a v i n g n o t h i n 0000140 g t o d o : o n c e o r 0000160 t w i c e s h e h a d p 0000200 e e p e d i n t o t h e \n b 0000220 o o k h e r s i s t e r w 0000240 a s r e a d i n g , b u t 0000260 i t h a d n o p i c t u r 0000300 e s o r c o n v e r s a t i 0000320 o n s i n \n i t , ' a n d 0000340 w h a t i s t h e u s e 0000360 o f a b o o k , ' t h o u 0000400 g h t A l i c e ' w i t h o 0000420 u t p i c t u r e s o r \n c 0000440 o n v e r s a t i o n s ? ' \n \n 0000460 S o s h e w a s c o n s i 0000500 d e r i n g i n h e r o w 0000520 n m i n d ( a s w e l l 0000540 a s s h e c o u l d , f o 0000560 r t h e \n h o t d a y m a 0000600 d e h e r f e e l v e r y 0000620 s l e e p y a n d s t u p 0000640 i d ) , w h e t h e r t h e 0000660 p l e a s u r e \n o f m a k 0000700 i n g a d a i s y - c h a i 0000720 n w o u l d b e w o r t h 0000740 t h e t r o u b l e o f 0000760 g e t t i n g u p a n d \n p 0001000 i c k i n g t h e d a i s i 0001020 e s , w h e n s u d d e n l 0001040 y a W h i t e R a b b i t 0001060 w i t h p i n k e y e s 0001100 r a n \n c l o s e b y h e r 0001120 . \n 0001122