That would be much appreciated- I suspect others will also be looking to program in some simple audio commands too… the current example file for “Yes” and “No” audio commands runs well and if whoever created that could offer guidance on how this was done using the HE core - thanks!
Hi Richard, I was given this link for how to do the pre/post-processing: firmware-alif-csolution/app/edge-impulse/inference/ei_run_audio_impulse.cpp at main · edgeimpulse/firmware-alif-csolution · GitHub
It looks like this will take some time to figure out what is happening. I think I’ll be able to use claude code to quickly get to the bottom of things here.
However, if you haven’t use the AI auto coders, I would suggest given them a try like Claude Code. Me and Ibrahim have limited bandwidth to help with everything, so, if you want to go faster and not wait on us give these tools a shot. We have a lot of customers using the system with them.
When I show the version of my build under Help> it shows IDE 4.8.6, however, I get errors which indicates I’m actually on 4.8.1.
(The version string at the bottom of your error shows this: OpenMV v4.8.1; MicroPython v1.26.0‑77; OpenMV‑AE3 …).
Claude indicates “that’s why every audio.init() form fails — the API we’re trying to use was added or stabilized after v4.8.1 and is only partly stubbed in that build. So at this moment, you’re on a cut of the firmware that exposes the audio names but doesn’t actually allow them to be configured or used directly.”
Sound correct?
further debuggin: Claude tells me that “this confirms that in build 999dd9965b, audio.init() is present, but it is not wired to accept any arguments at all.
There’s an option to install and flash the latest firmware from the IDE. Here’s the library the implements microspeech: openmv/scripts/libraries/ml/ml-core/ml/apps.py at master · openmv/openmv · GitHub
Note, microspeech uses a small model to do the audio samples pre-processing, and another one to do the KWS. I don’t know how EI’s model works.
The audio module for Alif port is here: openmv/ports/alif/modules/py_audio.c at master · openmv/openmv · GitHub
It’s very simple with just 2-3 functions. audio.init() accepts a bunch of args, like channels, sample rate etc.. The defaults should be fine. audio.start() accepts a callback that gets invoked with a samples buffer.
It would be easier if you’re using any agent to checkout the repo to get the latest code.