Hey, thanks for the video. I'm encountering an error though around 2:33, if you've any suggestions please let me know. stream.py: error: argument --model_name: expected one argument Changing the command to : python stream.py --model tiny But seeing this error now: ERROR: Failed to initialized SDL: dsp: No such audio device I've got a headset with microphone attached to the Pi 5 via USB port. Is it because I need an external soundcard/other hardware like in your video? Any ideas what the issue could be?
@Hardwareai5 сағат бұрын
Yes, there is an ongoing issue, which I am working on fixing: github.com/AIWintermuteAI/whispercpp/issues/88
@ameetkarn5 күн бұрын
This is too good....I think this should fit in directly with one of my project. Do you have any recommendation for real time TTS ?
@Hardwareai4 күн бұрын
Hopefully! I used espeak before for other projects... it is pretty horrible by modern standards, but does its job. For this example I used piper TTS - much better quality, but not as fast as espeak.
@sleetible6 күн бұрын
Does the new Hailo AI module offer any improvement on running any of these LLMs? I know it speeds up the vision side of things but I haven't seen anyone use it for LLMs yet.
@sleetible6 күн бұрын
Oh, and would it speed up Whisper at all? Allowing a larger model to be run?
@Hardwareai4 күн бұрын
re: LLMs, not really. The question has been asked many times in different locations, here is one of the replies www.reddit.com/r/LocalLLaMA/comments/1d7shcr/comment/l71q04c re: Whisper: given that it as transformer as well, Hailo are not geared towards this type of NN. but I remember seeing a paper about modifying BERT to be run with Google Coral USB, so... your mileage may wary, but it is going to be very far from plug-n-play
@glikoz10 күн бұрын
Please advise the hardware setup for offline RAG, TTS, STT
@Hardwareai7 күн бұрын
Hard to estimate without knowing the details?
@bluest152411 күн бұрын
You did great, thank you. I personally would prefer just to hear your voice though, without it being degraded and filtered the way you did. Your voice is good.
@Hardwareai8 күн бұрын
Hello! Thanks for the feedback - it is not actually even my voice though, it is generated with Bark TTS to resemble the voice of HK-47 from Star Wars Knights of the Old Republic :) I made a video about it here kzbin.info/www/bejne/mGbKamWBrMdqrNk Nevertheless, I'll be re-assessing Robotics Bi-weekly format in near future.
@bluest15247 күн бұрын
@@Hardwareai lol, okay. Gotcha. I can understand why, I see how it fits the subject matter. I'm an audio professional and work with vocoding, talkbox, voice manipulation too, isn't it a blast?
@amorpheuses162717 күн бұрын
Got one off of aliexpress - looks nice. You should probably power down using the settings menu power down command (as recommended) - ask me how I know.
@Hardwareai7 күн бұрын
As recommended. I'll guess - you bricked it :)
@amorpheuses16277 күн бұрын
@@Hardwareai Didn't brick it - but destroyed the OS on the sdcard. Had to reflash.
@jackwarner544522 күн бұрын
I'm trying to make an AI voice assistant and would be completely lost without your videos. Thanks so much!
@Hardwareai20 күн бұрын
Glad I could help!
@markantinozzi497023 күн бұрын
I'm going to try to install it.
@Hardwareai20 күн бұрын
There is a known issue at the moment: github.com/AIWintermuteAI/whispercpp/issues/88#issuecomment-2171120795 I'll be fixing it once I get back from traveling, beginning of July.
@markantinozzi497012 күн бұрын
@@Hardwareai Certainly let me know. I'm excited to try it. Happy traveling!! Be safe! >M<
@exploring-electronic24 күн бұрын
Leaving the comment here hoping you'd make a video about Cozmars 🎉
@Hardwareai20 күн бұрын
Noted!
@bystander8526 күн бұрын
I've been trying to find a way to make end of speech flag to be more intelligent than just detecting a pause. I find it common that I may have a mental blank, or misspeak, and the delay in my speech incorrectly flags end of speech. It would be interesting if STT systems can continue listening after a pause if it detects an incomplete sentence. Any ideas?
@Hardwareai20 күн бұрын
That's a hard one. I don't think this one is solved even in commercial STT engines - e.g. google assistant or siri. That would require understanding on sentence context. We might be getting somewhere with multi-modal models, such as GPT4o, but I don't think there is anything available to be run on Raspberry Pi format computer. Also, as a shortcut, perhaps it would be possible to either run a classifier or modify whisper model to output probability of sentence being finished... It's just an idea though, finding out how well will it work is another thing entirely.
@DavidPsurny27 күн бұрын
What is the full command line at 2:35 ?
@Hardwareai20 күн бұрын
python stream.py --model_name tiny.en-q5_1
@americo999928 күн бұрын
so this can't be used for ML , DL or Stable difusion , can it?
@Hardwareai20 күн бұрын
Well, there is more than one board in the video. Which one are you asking about?
@azhuransmx126Ай бұрын
Jetson Nano 0.476TOPS 😑 Jetson Xavier 25-40TOPs (Interersting) Jetson Orin 200TOPs (Nice but expensive) Jetson Thor 0.8PETAFLOPS (A little monster, start controling humanoids 2024) Next Gen over Thor 4PETAFLOPS (Robots take over the Industry 2026) Human Brain 12PETAFLOPS Next Gen Over Next Gen 20PFLOPS (Robots take over the World 2028)
@Hardwareai20 күн бұрын
Make Earth Great Again 2028 :) or MEGA
@HardwareaiАй бұрын
Hey, guys! I'm traveling in June, so I made some special episodes of Robotics Bi-weekly in advance for you - this is first one of them. Be good to my robotic co-host Dim-E and leave your feedback below :)
@NoidoDevАй бұрын
Finally a good overview video. Most of them are just trashy clickbait.
@HardwareaiАй бұрын
I enjoyed making it :) glad you enjoyed watching it. Trashy clickbait gets the clicks lol
@electric_sand8 күн бұрын
@@Hardwareai When it comes to "technical" content, core subscribers and consistent viewers who are into the field may not appreciate clickbait or bizarre videos. A sufficiently glamorous title and video may only attract the masses once in a while. May be wrong, but that's how I see it.
@Hardwareai8 күн бұрын
@@electric_sand well, I hope I did good enough job for my core subscribers and consistent viewers :)
@electric_sand8 күн бұрын
@@Hardwareai The video was nice. Thank you.
@levbereggelezoАй бұрын
thanks for the information, I will be waiting for new news releases
@HardwareaiАй бұрын
Always welcome
@dad2979Ай бұрын
I have been following you for a long time and you have done such a fantastic job of crafting your style and keeping your content relevant. Great job Dmitry!
@HardwareaiАй бұрын
Thank you for leaving this comment! I'm still refining my style to tell the truth. One of the things I was successful recently (I think) is keeping my videos more to the point, with good flow of information. Now it looks to me I was blabbering way too much in my older videos at times. I cut a lot of stuff now on post-processing if I feel the video is overloaded. I plan to make some more storytelling-oriented robotics content next half-year, stay tuned and see how it goes.
@gabrieloscarwolhein2721Ай бұрын
How much is the energy consumption?
@HardwareaiАй бұрын
With the screen and camera: idle around 1W, under load (NN inference) around 1.5W. Numbers are provided by Sipeed.
@FUBBAАй бұрын
I got a bittle X a while back and I have been loving it! Wondered if I could enhance the servos to metal and the angular option is awesome.
@HardwareaiАй бұрын
Perhaps contact Petoi and ask them if it's possible to simply purchase upgraded servos? The control board and everything else is the same.
@mandelafoggie9359Ай бұрын
😮
@HardwareaiАй бұрын
😊
@HardwareaiАй бұрын
Origin story of the narrator robot ---> kzbin.info/www/bejne/mGbKamWBrMdqrNk Even despite the voice is AI generated, I write the script and edit the videos myself - and it does take quite a bit of time. Leave some feedback below!
@levbereggelezoАй бұрын
Thanks for the information, it's very interesting
@HardwareaiАй бұрын
Glad you think so!
@BogdanMnikovАй бұрын
TIL I learned about the robo-snails 😅
@HardwareaiАй бұрын
Unlike birds, these are real!
@torstenaltmann62Ай бұрын
It threw me an error at "python3 -m build -w": PermissionError: [Errno 13] Permission denied: 'src/whispercpp/__about__.py' ERROR Backend subprocess exited when trying to invoke get_requires_for_build_wheel
@HardwareaiАй бұрын
Can you post this error with detailed steps preceding it and some environment info (OS, architecture) to the Github issues and tag me there?
@ClericHereticАй бұрын
Good info. Thanks.
@HardwareaiАй бұрын
Glad it was helpful!
@nhannguyenthanh46Ай бұрын
I orderred it from China. Maybe I will get it in this weekend. My idea is a compact camera with I/O + Ethernet Port + HDMI ( I see LicheeRV-Nano-E, however to be confused by Sipeed's comparison table ). this is my starting with Sipeed's product. Hope it will be good. Thanks for your quick review. Looking forward your next review.
@HardwareaiАй бұрын
Nice! LicheeRV-Nano is based on the same chip, but won't be getting same level of software support, e.g. MaixVision IDE. Theoretically you can develop the same capabilities from scratch using SOPHGO resources.
@nuanda82Ай бұрын
Nice!
@HardwareaiАй бұрын
Thanks! I agree.
@cipanmandulАй бұрын
Thank you! I almost buying Sipeed Maix-II dock until I found this video.
@HardwareaiАй бұрын
MaixCam is absolutely better choice at the moment. Maix-II was nice at the time, but then the chip shortage happened. Current LTS product for Sipeed is MaixCam as they told me.
@shakhizatnurgaliyev9355Ай бұрын
Хороший выпуск, Дима!keep it up!
@HardwareaiАй бұрын
Спасибо! I will!
@levbereggelezoАй бұрын
thanks for the information, we are waiting for the second part
@HardwareaiАй бұрын
You are welcome!
@BogdanMnikovАй бұрын
Keep it going! Looking forward for the second part
@HardwareaiАй бұрын
Coming out in two weeks!
@AntonMaltsevАй бұрын
Thank you for the video! I'll wait for the next one. I had a terrible experience with the Milk-V DUO board (I explain it a bit more in my video: kzbin.info/www/bejne/mpysh3eFmcR5l5Y ). It has the same SG2002 inside. It's super interesting to see how Seeed Studio worked with the same problems. Is there a good toolchain that works out of the box? Is there good accessibility to running a native model, etc.? But in terms of price/speed/ability to run, fp16 SG2002 looks great.
@shakhizatnurgaliyev9355Ай бұрын
I know you man!!😆🤣
@HardwareaiАй бұрын
Yeah, from what I see the software experience is much better with this one. Sipeed had spent plenty of time (maybe even too much? I will talk about why I think it might be too much in the next video) working on their (C/C++ and Python) SDKs and even homegrown IDE, so many things just work, at least for getting started example 100%. Btw, it is Sipeed board, not Seeed Studio. Seeed probably resells it as well, or at least they used to resell Sipeed boards, but it is a different company, a bit smaller in size.
@AntonMaltsevАй бұрын
@@Hardwareaiyeah, sometimes I get the two companies mixed up:) I'll be looking forward to the next video!
@jlbciriaco31422 ай бұрын
@hardwareai what raspberri pi are you sing?
@HardwareaiАй бұрын
I normally sing raspberri pi tenor, but I can do raspberri pi falsetto as well for comic effect xD Okay, I guess you asked what raspberry pi was I using, not singing. For this video it was Raspberry Pi 4. There is another newer video where I was using Raspberry Pi 5 as well, kzbin.info/www/bejne/aaqvd4qmgLCVm5o
@TomanswerAi2 ай бұрын
Very cool guide. Thank you.
@HardwareaiАй бұрын
Glad you enjoyed it!
@pauldolton91182 ай бұрын
These videos are great please keep the content coming I'm really enjoying watching them
@Hardwareai2 ай бұрын
Thanks, will do!
@levbereggelezo2 ай бұрын
Very good
@Hardwareai2 ай бұрын
Agreed!
@exploring-electronic2 ай бұрын
Thanks for fixing the sound!
@Hardwareai2 ай бұрын
No problem!
@Hardwareai2 ай бұрын
This is an updated version of the recent video - I fixed the sound, so hopefully you will enjoy it more :)
@BogdanMnikov2 ай бұрын
Thermonator will also work for duck hunting 🦆
@Hardwareai2 ай бұрын
And cooking!
@ptsckts61232 ай бұрын
hello, same benchmark results in 5925.774ms computation time on my RPI 5 currently, should I do anything differently? the audio file i've used is 10 secs, same JFK speech
@Hardwareai2 ай бұрын
One thing I could have improved about my little benchmark script is multiple measurements. First run is always the slowest. Is 5925 ms. for the first run or even for later concurrent runs as well?
@ptsckts61232 ай бұрын
@@Hardwareai Ooh that was it, now I get ~600ms. Thanks! Also I got 1.218 sec computation for a 145 seconds talk, I don't know how it works but segmentation takes much longer
@NextGenSellPOS2 ай бұрын
this tutorial needs a tutorial
@Hardwareai2 ай бұрын
Does it though?
@bens44462 ай бұрын
I had heard about faster whisper on other channels but thought it couldn't work on an SBC because it uses GPU which an SBC doesn't have. I have no idea how you did this. Thanks!
@Hardwareai2 ай бұрын
Interesting. No, it certainly can run on CPU - I made a follow-up on this video, explaining more about faster-whisper specifically, you can find it on my channel.
@domesticatedviking2 ай бұрын
Hey, just wanted to say I really appreciated your last two videos. Will you please be my sensei? Thank you!!
@Hardwareai2 ай бұрын
I appreciate your appreciation! xD I'd say that I'm already a sensei of sorts... You always can support me on Patreon for some extras, but otherwise simply stay tuned for more videos!
@levbereggelezo2 ай бұрын
Thx
@Hardwareai2 ай бұрын
No problem!
@levbereggelezo2 ай бұрын
Thx
@Hardwareai2 ай бұрын
Appreciate it!
@levbereggelezo2 ай бұрын
Very good
@Hardwareai2 ай бұрын
Glad you enjoyed the video!
@Hardwareai2 ай бұрын
Support my work on making tutorials and guides on Patreon! www.patreon.com/hardware_ai
@Hardwareai2 ай бұрын
The follow-up video is also live on KZbin - find it in my channel. Support my work on making tutorials and guides on Patreon! www.patreon.com/hardware_ai