This basically is SAM-on-a-chip. Cool! (FYI, SAM stands for Software Automated Mouth and it's a software from Don't Ask Software, which is one of the most fun voice synthesis programs you can use)
Agree with you, this could make an awesome little debugging tool, rather than streaming print commands to a display, audio feedback of the same stream could be extremely useful and simply tagged onto the output stream of the device/module under test. Good job in on reverse engineering the data sheet. Impressive intonation and clarity for such a cheap and simple solution and presumably can be driven the same way via I2C and SPI bus. Subbed.
I suspect the intended application of this chip was as the core of a personal translator where arbitrary strings of text are used. Using MP3 or WAV files of standard or canned messages allow natural sounds like human voice to be used when more defined messages are output. This is the way the FrSky range of TX's do it and part of the reason the Amber sound pack is so popular (apart from her dead sexy voice) over sounds generated by TTS systems. creating the sound files is a big task hence why only a very few natural sound recordings exist for these radios.
Totaly agree. For higer grade transmitter this TTS generated voice sounds too artificial. But for amateur grade DIY transmitter it is easy solution - no WAV files, no sophisticated coding needed. Just send voice message via serial interface.
How difficult is it to have interrupts. It seems at this point it plays back but could you stack 10 phrases? Eg when you open the front door you get “welcome”, or when an LDR has sufficient light it would say “turning on lights”, then easily linked out to a relay for lighting. It would be helpful for someone of limited sight to have notification when mail has arrived (ok a blind person can’t read mail), however it could be medication that’s delivered by mail or perhaps a rain sensor so that that person can bring in the washing.?
From memory (it's been a while) for c syntax, you might need to escape a % character in the text by using another %. e.g. "%%" gives you one % character in the text. But I could be wrong.. :). But I guess it's probably doubtful their firmware is going to say that character correctly anyway. Interesting video, thanks.
My speech synthesis module returns 45(Incorrect command frame) while doing status inquiry.I am using arduino mega 2560. Please help as I am not able to figure out the fault
At 16:26 you ask why it does not work, i think cause the msg length is not good, it's 14 instead of 10, you added 4 bytes. Does this chip sing too, it lookt like the Emic2 and the tts-click board. when you feed it allophones like this the Emic2 or the tts-click-board sings. [AEEY][] Maybe you have to look for something like pnone on or arpaphone on, but the [..........] means do this as allophones. can you try this on your board Goodluck, sounds great
It's not what I would call simple. Try writing some software that does this yourself and you'll see just how simple it really is :) And not that the documentation should affect the price, but I would bet good money the Chinese documentation is perfectly fine - pretty sure I made it clear that I was using an automated translation to English. The PC soundcard comment is like comparing apples and oranges. This module is much smaller, lighter and low-power, the speech engine is entirely self contained and does not rely on the controller, and instructions can be given with just one signal wire. Please show me a PC soundcard with these features that costs less.
There are plenty od tts software C libraries. All you need is a separate faster mcu and dedicate it for one tts task. Most of soundcards support SAPI (speech api) Maybe not fully but have api features built in. I still think this should be in $5 range.
I still think you're comparing apples with oranges - nobody making a portable embedded project (eg. custom R/C transmitter) wants to bother with a big soundcard. With this you don't need libraries, and you don't need any separate faster mcu, or dedicate anything else for tts. Just show me a soundcard cheaper than $14 with the same features, that can be used with arduino UART on one wire and gets up and running with so few lines of code :)
It is just too expensive) The features are very basic. There was also a project for under $5 with very simple TTS using mp3 player ic. I'm not saying it is bad circuit but it is just too pricey for what normally people do.
Compare the XFS5152 to the Parallax EMIC2 - they are very similar devices, the EMIC is priced at $60 plus - and is very well documented. But it sounds much more robotic. The XFS5152 at under $20 is a good really good deal for a self-contained synthesizer - even with the slightly Chinese accent and lack of English documentation. The XFS5152 is a really good deal, and is in the price range that can be added to many projects and leave it in a project. I own 2 EMIC2 boards now that I've heard this, and see how it works, this will be on my list to for many many projects. Being able to buy three for the price of one EMIC2 seems like a great deal.
This basically is SAM-on-a-chip. Cool! (FYI, SAM stands for Software Automated Mouth and it's a software from Don't Ask Software, which is one of the most fun voice synthesis programs you can use)
Good to know about this module. I'm sure it will come in handy. Great vid, thanks!
Agree with you, this could make an awesome little debugging tool, rather than streaming print commands to a display, audio feedback of the same stream could be extremely useful and simply tagged onto the output stream of the device/module under test. Good job in on reverse engineering the data sheet. Impressive intonation and clarity for such a cheap and simple solution and presumably can be driven the same way via I2C and SPI bus. Subbed.
Pretty interesting device to play with!
Awesome! Thank you for sharing.
Reminded me of playing with a curragh micro speech for the zx spectrum.
thank you! I tested your sketch on two different modules and it worked fine, SYN6988 chip
Thanks for the feedback. Usually the only comments I get is when something doesn't work :D
Great video, thank you!
It's quite good for the price. I was surprised.
Hi you are awesome. I like your videos very much. Thank you
GREAT VIDEO, i mean vi'di'oh!
Cool. Now to hook it to the DIY transmitter for voice announcements :)
I suspect the intended application of this chip was as the core of a personal translator where arbitrary strings of text are used. Using MP3 or WAV files of standard or canned messages allow natural sounds like human voice to be used when more defined messages are output. This is the way the FrSky range of TX's do it and part of the reason the Amber sound pack is so popular (apart from her dead sexy voice) over sounds generated by TTS systems. creating the sound files is a big task hence why only a very few natural sound recordings exist for these radios.
Totaly agree. For higer grade transmitter this TTS generated voice sounds too artificial. But for amateur grade DIY transmitter it is easy solution - no WAV files, no sophisticated coding needed. Just send voice message via serial interface.
rfduino has a library for a module by iflytek that is very similar. XFS5051CE maybe it will work with the XFS5152CE.
3:07... You could just twist those output wires together to get some sort of noise rejection if it is an issue.
It's got exactly the same voice that Chinese version Android phones come with. That's often used with things like Baidu Maps.
Wishing I herd of this earlier also Do you have the original PDF?
How difficult is it to have interrupts. It seems at this point it plays back but could you stack 10 phrases? Eg when you open the front door you get “welcome”, or when an LDR has sufficient light it would say “turning on lights”, then easily linked out to a relay for lighting. It would be helpful for someone of limited sight to have notification when mail has arrived (ok a blind person can’t read mail), however it could be medication that’s delivered by mail or perhaps a rain sensor so that that person can bring in the washing.?
Very Good
I was indeed yelling at my screen ;)
I am really impressed. That China company should pay you being their application engineer!
From memory (it's been a while) for c syntax, you might need to escape a % character in the text by using another %. e.g. "%%" gives you one % character in the text. But I could be wrong.. :). But I guess it's probably doubtful their firmware is going to say that character correctly anyway. Interesting video, thanks.
Yes, that works. Thanks for the reminder
My speech synthesis module returns 45(Incorrect command frame) while doing status inquiry.I am using arduino mega 2560. Please help as I am not able to figure out the fault
I think you could use %d%% to get the battery percentage.
ah of course, how did I forget that :)
Cool
Improvement to above code so you can send txt strings longer than 253. chars also don't have to add the waitForSpech after each speak("bla bla")
void waitForSpeech(unsigned long timeout = 80000) {
unsigned long start = millis();
bool done = false;
while ( ! done && (millis() - start) < timeout ) {
while ( Serial2.available() ) {
if ( Serial2.read() == 0x4F ) {
done = true;
break;
}
}
}
}
void speak(char* msg) {
Serial.println(msg);
short meslength= strlen(msg)+2;
Serial2.write(0xFD);
Serial2.write((meslength >> 8) & 0xFF);
Serial2.write(meslength & 0xFF);
Serial2.write(0x01);
Serial2.write((byte)0x0);
Serial2.write(msg);
waitForSpeech();
}
The accent is a little bit Chinglish, but not bad. Thanks for the video!
Bona Fide it sounds like it was calibrated for east asian languages
At 16:26 you ask why it does not work, i think cause the msg length is not good, it's 14 instead of 10, you added 4 bytes. Does this chip sing too, it lookt like the Emic2 and the tts-click board. when you feed it allophones like this the Emic2 or the tts-click-board sings. [AEEY][]
Maybe you have to look for something like pnone on
or arpaphone on, but the [..........]
means do this as allophones.
can you try this on your board
Goodluck, sounds great
No, these boards can't sing. That's a DECtalk thing, like in the Emic2 and TTS-Click.
13:15 科大讯飞 is the brand name. So that's what it says.
When they make the pitch drop at the end of every word it makes it sound majorly depressed... big turn off.
yes i knew why it didn't work VALIDATE ME CHRIS.
Can it pronounce "Hernione Granger" correctly?
is that from McDonald's China?
For a simple IC with poor documentation this should be much cheaper. PC soundcard with such support cost less
It's not what I would call simple. Try writing some software that does this yourself and you'll see just how simple it really is :) And not that the documentation should affect the price, but I would bet good money the Chinese documentation is perfectly fine - pretty sure I made it clear that I was using an automated translation to English. The PC soundcard comment is like comparing apples and oranges. This module is much smaller, lighter and low-power, the speech engine is entirely self contained and does not rely on the controller, and instructions can be given with just one signal wire. Please show me a PC soundcard with these features that costs less.
There are plenty od tts software C libraries. All you need is a separate faster mcu and dedicate it for one tts task. Most of soundcards support SAPI (speech api) Maybe not fully but have api features built in.
I still think this should be in $5 range.
I still think you're comparing apples with oranges - nobody making a portable embedded project (eg. custom R/C transmitter) wants to bother with a big soundcard. With this you don't need libraries, and you don't need any separate faster mcu, or dedicate anything else for tts. Just show me a soundcard cheaper than $14 with the same features, that can be used with arduino UART on one wire and gets up and running with so few lines of code :)
It is just too expensive) The features are very basic. There was also a project for under $5 with very simple TTS using mp3 player ic. I'm not saying it is bad circuit but it is just too pricey for what normally people do.
Compare the XFS5152 to the Parallax EMIC2 - they are very similar devices, the EMIC is priced at $60 plus - and is very well documented. But it sounds much more robotic. The XFS5152 at under $20 is a good really good deal for a self-contained synthesizer - even with the slightly Chinese accent and lack of English documentation. The XFS5152 is a really good deal, and is in the price range that can be added to many projects and leave it in a project. I own 2 EMIC2 boards now that I've heard this, and see how it works, this will be on my list to for many many projects. Being able to buy three for the price of one EMIC2 seems like a great deal.