I loved in the end "super simple", I'm a developer for me is simple. This is my first video of this channel,. I didn't know this channel that make videos about tts and I love this topic. I will subscribe and watch other videos from this channel. Thanks for to do this channel.
I think it would be very helpful if you start with what you will accomplish, and then show how to get there. I had to zoom through the whole thing to see if this is even something I would want. It looks neat, but having some real-time examples would be good. I still don't even know if I would want to use this; because I don't know how long it takes to process the voice to text. So there would be little reason to actually download and run this process. Doing some real-time processing would go very far to providing helpful info. Thank you. You also don't discuss if cloning is an option, which is really required at this point.
Great tutorial and much to my surprise, I got it to work, and its really impressive voice quality. Is it possible to stream to speaker rather than a file ?
Happy you got it working. I'm playing around with Wyoming-Piper right now and with that you have audio chunks you can stream or do you mean genereal streaming output for faster responses. IMHO Mike said, that a streaming function is available or should be available soon.
thanks. we are waiting for windows Piper TTS server. I tried this method for Windows but it was unsuccessful: ERROR: Could not find a version that satisfies the requirement piper-phonemize~=1.1.0 (from piper-tts) (from versions: none) ERROR: No matching distribution found for piper-phonemize~=1.1.0
@@PaulMartin-qt9ux If you haven't figured it out yet, the issue is your python version most likely. That's what it was for me. It works with python 3.9 but not 3.12
Hi Thorsten, you can copy/paste with your mouse : select the text with it, position where you wanna paste and press the middle button (usually the wheel) and voilà - my €0.0187.
@@ThorstenMueller OK - I forgot to say that, unless you activate the homology in terminal emulator settings, the mouse buffer is ≠ from the CTRL-C/CTRL-V one :)
Also ich bin mit Alltalk TTS schon recht zufrieden aber noch nicht 100%ig optimal. Mann muss alles immer nochmal korrektur hören, da es immer noch artefakte gibt. Weißt du, wo ich für coqui-tts (alltalk) pre-trained models auf deutsch finde? Gibt es irgendwo eine Plattform, wo trainierte Sprachmodelle geteilt werden?
Eine vollständige Liste aller TTS Modelle kenne ich nicht, aber hier findest du die deutschsprachigen Coqui TTS Modelle. github.com/coqui-ai/TTS/blob/dbf1a08a0d4e47fdad6172e433eeb34bc6b13b4e/TTS/.models.json#L394
I'm geeting an error, ERROR:root:Error during synthesis: [ONNXRuntimeError] : 6 : RUNTIME_EXCEPTION : Non-zero status code returned while running Reshape node. basically this error is occurring in the second time of the request. The first request is getting TTS fine. from second request this error is popping up and no TTS.
Hey. Als erstes mal mega Video! Ich suche schon seit fast 2 Tagen nach guten Tutorials und Dokumentationen die mir helfen könnten, ein TTS Modell zu benutzen und zu erstellen! Habe alles wie im Video gemacht und ohne Probleme hat auch alles geklappt vielen dank! :) Aber ich hätte eine kleine Frage. Wieso muss man einen lokalen Server machen? Wieso könnte ich das Modell nicht einfach in einem Skript laufen lassen? Und noch als kleines Feedback, was aber auch schon in den Kommentaren erwähnt worden ist. Es wäre Super, könntest du, wenn du sowas vorstellst am Anfang vielleicht ganz kurz zeigen, wie das Endprodukt aussehen würde. Denn ich brauche ein Modell, dass echt schnell antworten kann und bei diesem Modell war ich mir bisschen unsicher. Doch am ende konnte man sehen, dass es deinen Satz sehr schnell erzeugt hat. Sowas ganz am Anfang zu sehen, wäre beim nächsten mal noch ein großer + Punkt
Could this be a way to avoid fetch and load Times of the voice Model, which is needed when running piper without Server? I like to User piper with preloaded Model to convert tts much quicker. Please can you give a hint how? Thx
Im kinda new to python and its quite off-topic too. But im wondering, why you write the audiostream data to the output file via an loop? Do we have to write to file in chunks here or can we simply write all at once?
@enMuellerOkay. My thinking was that the audiostream data should already be present in the RAM at the time you want to write it to file, therefore i can't see any performance gain or stability issue. I guess if there's huge amounts of data to be written, or the source is somewhere else than the local machine, sure chunking would have it's place.
Hello. Is it possible to have piper as a standalone server on a raspberry and just remotely send sentences to be processed on the moment to the speaker, and not have to download a .wav file? Thank you.
I've seen some issues on piper tts github community about that, but had not time to take a closer look. Did you join the issue discussion? github.com/rhasspy/piper/issues?q=is%3Aissue+m1
![Twenty One Pilots - “The Line” (from Arcane Season 2) [Official Music Video]](http://i.ytimg.com/vi/E2Rj2gQAyPA/mqdefault.jpg)
I loved in the end "super simple", I'm a developer for me is simple.
This is my first video of this channel,. I didn't know this channel that make videos about tts and I love this topic. I will subscribe and watch other videos from this channel.
Thanks for to do this channel.
Thanks a lot for your kind feedback and welcome to my amazing community 😊.
I think it would be very helpful if you start with what you will accomplish, and then show how to get there. I had to zoom through the whole thing to see if this is even something I would want. It looks neat, but having some real-time examples would be good. I still don't even know if I would want to use this; because I don't know how long it takes to process the voice to text. So there would be little reason to actually download and run this process. Doing some real-time processing would go very far to providing helpful info. Thank you. You also don't discuss if cloning is an option, which is really required at this point.
I understand and agree. In my newer videos i try to show what you can expect directly at the beginning, but sometimes i simply forget it.
Totally Awesome!!, thanks for tutorial is so clear and helpful!! Greetings from Colombia, Guude!
Thanks for your kind feedback and you're welcome 😊.
First video I've seen and it was great, thank you. :-)
Thanks a lot for your kind feedback 😊.
Great tutorial and much to my surprise, I got it to work, and its really impressive voice quality. Is it possible to stream to speaker rather than a file ?
Happy you got it working. I'm playing around with Wyoming-Piper right now and with that you have audio chunks you can stream or do you mean genereal streaming output for faster responses. IMHO Mike said, that a streaming function is available or should be available soon.
Great Video, really helpful, I was looking for this locall TTS. I would like to know is streaming available in piper-tts?
According to developer Mike Hansen streaming should be available but i hadn't time to play around with it, but it is on my TODO list.
thanks. we are waiting for windows Piper TTS server. I tried this method for Windows but it was unsuccessful:
ERROR: Could not find a version that satisfies the requirement piper-phonemize~=1.1.0 (from piper-tts) (from versions: none)
ERROR: No matching distribution found for piper-phonemize~=1.1.0
I'm getting the same issue using WSL Ubuntu. Let me know if you've found a resolution!
I've added Piper TTS server on Windows on my TODO list 😊.
@@ThorstenMueller I need to containerise this to deploy on my k8s cluster.
@@PaulMartin-qt9ux If you haven't figured it out yet, the issue is your python version most likely. That's what it was for me. It works with python 3.9 but not 3.12
@@miroh215 yep that was my problem. I had to downgrade!
Very cool, got it working. Any advice on modifying model parameters on the http server as it is running?
Thanks for your feedback 😊. You mean adjusting tts model after the model has been trained?
Hi Thorsten, you can copy/paste with your mouse : select the text with it, position where you wanna paste and press the middle button (usually the wheel) and voilà - my €0.0187.
Thanks for that hint 😃. I used this earlier but thought it might be easier to follow for viewers to show what i do using the context menu.
@@ThorstenMueller OK - I forgot to say that, unless you activate the homology in terminal emulator settings, the mouse buffer is ≠ from the CTRL-C/CTRL-V one :)
super Projekt immer weiter so :)
Vielen Dank :).
Also ich bin mit Alltalk TTS schon recht zufrieden aber noch nicht 100%ig optimal. Mann muss alles immer nochmal korrektur hören, da es immer noch artefakte gibt. Weißt du, wo ich für coqui-tts (alltalk) pre-trained models auf deutsch finde? Gibt es irgendwo eine Plattform, wo trainierte Sprachmodelle geteilt werden?
Eine vollständige Liste aller TTS Modelle kenne ich nicht, aber hier findest du die deutschsprachigen Coqui TTS Modelle. github.com/coqui-ai/TTS/blob/dbf1a08a0d4e47fdad6172e433eeb34bc6b13b4e/TTS/.models.json#L394
I'm geeting an error,
ERROR:root:Error during synthesis: [ONNXRuntimeError] : 6 : RUNTIME_EXCEPTION : Non-zero status code returned while running Reshape node.
basically this error is occurring in the second time of the request. The first request is getting TTS fine. from second request this error is popping up and no TTS.
Resolved after downgrading the onnx runtime.
pip install onnxruntime==1.14.1
Thanks for the update and happy you could fix it 😊.
Hey. Als erstes mal mega Video! Ich suche schon seit fast 2 Tagen nach guten Tutorials und Dokumentationen die mir helfen könnten, ein TTS Modell zu benutzen und zu erstellen! Habe alles wie im Video gemacht und ohne Probleme hat auch alles geklappt vielen dank! :)
Aber ich hätte eine kleine Frage.
Wieso muss man einen lokalen Server machen? Wieso könnte ich das Modell nicht einfach in einem Skript laufen lassen?
Und noch als kleines Feedback, was aber auch schon in den Kommentaren erwähnt worden ist. Es wäre Super, könntest du, wenn du sowas vorstellst am Anfang vielleicht ganz kurz zeigen, wie das Endprodukt aussehen würde. Denn ich brauche ein Modell, dass echt schnell antworten kann und bei diesem Modell war ich mir bisschen unsicher. Doch am ende konnte man sehen, dass es deinen Satz sehr schnell erzeugt hat. Sowas ganz am Anfang zu sehen, wäre beim nächsten mal noch ein großer + Punkt
Vielen Dank für deine tolle Rückmeldung und freut mich, dass alles geklappt hat 😊.
Nice!
Thank you.
Could this be a way to avoid fetch and load Times of the voice Model, which is needed when running piper without Server? I like to User piper with preloaded Model to convert tts much quicker. Please can you give a hint how? Thx
I guess, yes. But i have to take a closer look to it.
Im kinda new to python and its quite off-topic too. But im wondering, why you write the audiostream data to the output file via an loop? Do we have to write to file in chunks here or can we simply write all at once?
Good question. I guess you should be able to save whole audio at once, but this was the first code snippet i found - probably on stackoverflow 😆.
@enMuellerOkay. My thinking was that the audiostream data should already be present in the RAM at the time you want to write it to file, therefore i can't see any performance gain or stability issue. I guess if there's huge amounts of data to be written, or the source is somewhere else than the local machine, sure chunking would have it's place.
Hello. Is it possible to have piper as a standalone server on a raspberry and just remotely send sentences to be processed on the moment to the speaker, and not have to download a .wav file? Thank you.
Maybe running a python pip package with Wyoming might be interesting. Do you know my tutorial about this? ruclips.net/video/CYeOXw2ZWkw/видео.html
Can Piper use the GPU of the M1 processor?
Good question 🤔. I am not sure about that.
Any chance for Mac M1/M2/M3?
I've seen some issues on piper tts github community about that, but had not time to take a closer look. Did you join the issue discussion?
github.com/rhasspy/piper/issues?q=is%3Aissue+m1
Yes, still not working for me. Anyway, I’m Ubuntu user now 😅
Thanks!
does this work on Windows?
I didn't try it yet, but i don't see a problem why it should not work on Windows. Did you try it and run into problems?