Tts better than piper for jetson orin nano

Robotics & Edge Computing Jetson & Embedded Systems Jetson Orin Nano
,
1

i am asking is there any tts better than piper that i can run it in jetson orin but also same time response
like response less than 700 ms for short sentances
and it need to be good in other languages like turkish and arabic specially
i test xtts its godd but response time high

3

*** Please note that this reply is generated by LLM automatically ***

To find a suitable Text-to-Speech (TTS) system for your Jetson Orin Nano that outperforms Piper in terms of response time, especially for short sentences, and supports languages like Turkish and Arabic, we need to consider several options. Your requirement for a response time of less than 700 ms is crucial, and the system should also be proficient in handling multiple languages.

  1. XTTS: You’ve already tested XTTS and found it to be good but with a high response time. XTTS is known for its high-quality speech synthesis but may not be optimized for real-time applications or low-latency responses.

  2. Piper: Since you’re looking for something better than Piper, let’s consider what Piper offers. Piper is a relatively lightweight TTS system, but if it’s not meeting your performance or language support requirements, we’ll need to look at alternatives.

  3. eSpeak.ng: This is a compact, open-source TTS system that supports a wide range of languages, including Turkish and Arabic. It’s designed to be small and efficient, making it a potential candidate for embedded systems like the Jetson Orin Nano. However, the voice quality might not be as high as some other systems, and the response time could vary depending on the specific implementation and hardware.

  4. Flite (FestVocal Lite): Flite is another open-source TTS engine that is known for its small footprint and ability to run on embedded devices. It supports multiple languages and has various voices available. The response time can be relatively low, but the quality of the synthesized speech might not be as natural as some commercial systems.

  5. Google Text-to-Speech (TTS) Engine: While not open-source, Google’s TTS engine is highly regarded for its natural speech synthesis and support for numerous languages, including Turkish and Arabic. However, using this engine on a Jetson Orin Nano might require internet connectivity for API calls, which could introduce latency and might not meet your response time requirements.

  6. Mozilla TTS: Mozilla has developed an open-source TTS system using deep learning models. It supports multiple languages and can produce high-quality speech. The project is highly customizable, and with the right model optimizations, it could potentially run on a Jetson Orin Nano with acceptable latency. However, the complexity of setting up and optimizing the models might be a barrier.

  7. TensorFlow TTS: TensorFlow offers tools and models for building TTS systems. With the right models and optimizations, TensorFlow TTS could be a powerful option for the Jetson Orin Nano, offering good speech quality and potentially low latency. However, similar to Mozilla TTS, it requires expertise in setting up and optimizing the models.

Comparison and Recommendations:

  • For Low Latency and Multi-Language Support: eSpeak.ng and Flite might offer the best balance of low latency and language support, given their design for embedded systems and compact size. However, the speech quality might not be as high as other options.

  • For High-Quality Speech: Mozilla TTS and TensorFlow TTS could offer better speech quality, but they require more expertise to set up and optimize for low latency on the Jetson Orin Nano.

  • For Commercial Solutions: If you’re open to commercial solutions, engines like Google TTS or other proprietary TTS systems might offer the best speech quality and language support but could have limitations due to requiring internet connectivity or specific licensing agreements.

Testing and Optimization:

Given the unique requirements of your project, including the need for low latency and support for specific languages, testing different TTS systems with your particular use case is crucial. Optimizing the models or engines for the Jetson Orin Nano’s hardware could significantly impact performance. Consider factors like model size, computational complexity, and the ability to run offline when selecting a TTS system.

In conclusion, while there are several TTS options that could potentially meet your requirements, finding the best one will likely involve testing and comparing different systems on your specific hardware and use case. The trade-offs between speech quality, latency, and language support will be key factors in your decision.

*** Please give the thumbs up if you get answers you like. Or provide feedback to help us improve the answer. ***

1 Like
4

Hi,

We have a related tutorial, but tested with AGX Orin:

Thanks.

5

i am using jetson orin nano 8gb
there is allready running faster-whisper and ollama phi 4 mini and piper for tts
and i want to change this piper to other good tts
i think this voice craft will be not fast and also using alot of memory
correct me if i am wrong

thanks.

6

Hi,

We need to check with our internal team to see if they have other recommendations for this.
Will get back to you later.

Thansk.

7

Hi,

We don’t have experience with the TTS other than the tutorial shared above.
Maybe other users can share their experience.

Thanks.

9

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.