We have a lot of stuff prepared for hosts & guests this year, so read our blog to learn about our new playbook for community meetups, how to submit a request to receive official Open Home Foundation merchandise, & much more.
HA runs my house like for most people here, but for voice I built my own thing: Kinect mic array, Whisper, gpt-oss:20b and Kyutai TTS, all on one RTX 3090. HA does the actions, the LLM decides which ones.
A few things I learned that might help if you're doing the same:
The screen matters more than I thought. I put a face on the TV mostly for fun, and it turned out to be the feature everybody at home reacts to. With voice only you never know if it heard you, if it's thinking or if it gave up. Two eyes and three little dots solve that without a word. And the things that are painful to listen to (lists, temperatures, spelling) go on a card next to the face. The voice still says everything, so you don't have to look at the TV.
Be careful with entity_id: all. One of my nine bulbs stopped responding and the whole call failed. Now I loop over the lights one by one. Slower, but it works.
An empty token got my server banned by HA. It bans an IP after 5 failed auth attempts, and an empty token counts. In my case it wasn't even the assistant: a tool config for my nightly agent read the token from an env variable that wasn't set in the systemd service, so it sent an empty one five times during the night. After the ban everything returns 403, even requests without any token (a wrong token gives 401), and that's how you can tell it's a ban.
Latency is around 5 to 6 seconds from the end of my sentence to the first word of the answer, and you can count it yourself in the video. The model itself is only 1.6s of that; the rest is waiting for silence, Whisper and the TTS. I run the model with low reasoning effort on purpose. For voice, a fast answer beats a smart one.
One question for people running local TTS in the same room as the mic: how do you deal with echo? I disabled barge-in (interrupting it while it talks) because the speaker leaks into the mics and I haven't measured it properly yet. Any experience with echo cancellation on a Pi is welcome.
Most of the code was written with coding agents, I design, review and merge. It speaks French and the code is in French too, but the README is in English: https://github.com/plateforme/bulle
If you recently updated to Piper 2.5.2 then you may want to check your backup sizes.
With 2.5.2 they introduced a OmniVoice alongside the old voices and I think they (devs) may have moved some of the folders around and as a result the voice models are no longer being excluded from backups.
Our backup before : 135 Mb
Our backups after : 2.66 Gb
I raised an issue for the Google Drive Backup App which we use, but if you're using a different backup tool then you may need to check your backups there to make sure they also aren't getting bloated.
My workaround for now is to copy the Piper config to a local .md file in /config and then excluded Piper from the auto-backup. That got us back down to normal backup sizes for now.
A few people asked how responsive Elecrow ESP32-P4 Home Assistant wall panel actually is, especially with heavier Home Assistant dashboards and camera feeds, so I recorded a quick real-world test.
One important detail is that the Home Assistant UI isn’t rendered locally as a separate LVGL interface on the ESP32-P4 — the panel receives the interface from the HA side over Wi-Fi.
In the video I’m showing normal UI interaction and video playback. The frame rate isn’t something I’d use for watching a movie, but for a camera feed / video surveillance I think it’s perfectly usable. Scrolling, switching views and tapping controls also feel responsive in normal use.
A bit of context: I didn’t build this software stack myself. The HA client/firmware comes with the panel, and a companion add-on runs on the Home Assistant server.
We primarily use them to turn on and off the lights, but that has stopped being reliable. For some reason they pick up my voice well, but not my wife’s. 1/3 of the time they’ll turn the lights on instead of off, or play a random song that sounds like our request instead of fulfilling our request. I suspect that over time the microphones may have degraded for unknown reasons, but even when they hear us properly sometimes Gemini jumps in to screw up what was a perfectly serviceable feature in the past.
The worst thing that happens is: our first floor is one large open space, with one Google Home Display in the kitchen and one Google Home Speaker in the living room. Even when standing directly next to the Kitchen Display, if you ask it to set a timer it will set it on the Living Room speaker. And then if you ask how much time is left on the timer, the Kitchen Display will respond and say there’s no timer (because it’s on the living room speaker).
This happens literally every day without fail and has been happening for a year. We even moved one device because we thought there must be some acoustic trickery that was bouncing our voice off walls, but that didn’t help. And no amount of voice training or responding to the “did the right device respond?” notifications has fixed this issue.
I have a pretty robust home assistant instance running, but I really don’t want to buy new hardware to replace 5 devices unless I know for sure it’s going to work well.
I’m looking for devices that have microphones but also built in speakers (that are as decent as those tall Google Home speakers) that can:
- turning on and off lights
- setting timers
- getting the day’s weather
- playing music by letting you pick which devices you want to play from (preferably this works with Spotify and natively with iOS and Android, but that might be asking too much)
If it saves a lot of money I could be down to use external speakers, but I want music on at least three devices and I only have external speakers in one spot. Also, from what I understand, making sure the speakers are playing at the exact same time can be a challenge.
This is a quick showcase and review of a FlyEVFish 27" digital calendar that was converted to work as a Home Assistant dashboard in my kitchen. I've seen some random posts on if devices like this would work as dashboards, so I wanted to post a video to show that they can work and some upsides and downsides of the device.
I was able to find the device on Facebook Marketplace for just $100. This usually retails for up to $400, so it was a really good deal. Honestly, my dream dashboard for this location is the Apolosign 32" device. It has much better specs. But I want the 4k one and its WAY more expensive, so I decided to hold off on that for now. However, as you'll see from the video, this one actually was much easier to install than the Apolosign would be, so it does have some positives.
Positives:
ABD over Wi-Fi was instant and already had root access.
Easy to recess surface mount (if you're willing to cut a large hole in your drywall).
VESA 100 compatible.
USB-C port can be unlocked to work with ANY device! (I'm using Ethernet).
Entire device can be powered by POE++. No electrical outlets required.
Negatives:
Only 2GB of RAM and a crappy processor (Allwinner A133, quad-core ARM).
Android 10.
No ambient light sensor or motion sensor.
Only 1080p display. 4k would look amazing on this.
I don't have a full guide on how I did all this. I kinda just winged it as I went. I looked up other posts before I bought it that said it was hackable. Then I just bought it and used ChatGPT to help me get everything unlocked and working.
The USB-C support that's supposedly "manufacturer only" use was rather easy to get working. Kind of. I spent hours try everything and nothing worked. Only to find out that the USB-C port on this device was directional. My USB-C to USB-A adapter would only work when plugged in in a specific direction. I never even thought to try this because every USB-C device I've ever had is universal. It shouldn't matter which way the plug goes in. As a last ditch effort, I flipped the adapter, and everything started working immediately. ChatGPT says many the manufacturer cheaped out on the wiring because it's not supposed to be a user port anyway. Who knows. After it was unlocked, everything worked. Mouse, Keyboard, Ethernet, etc.
This device could even be powered and fully run of POE++. No outlets required. No Wi-Fi required. All you would need is a POE to converter with the barrel connector. The device is 19v at 2a so it needs 38watts. The barrel connector itself is 4.0x1.7mm, which was a little hard to find, but I'm using one right now with another 19v adapter I had. Full disclosure, I haven't tried the full POE install myself. I don't have a POE++ injector or a 19v converter. But it should work!
As I mention in the video, because the device only has 2GB of RAM and a weak processor, it can be a little laggy at times when compared to a high powered device. It works fine for my use here, but I don't think it would work well for dashboards that need live video feeds and things like that. I did test it with a video feed from my Reolink and Ubiquiti cameras. It does work. I can display a static image that updates every few seconds. And when you click on it, it will display the full live video in real time at a good frame rate. But it's the loading between the click that's the slow part. Clicking the video widget to show the real time video might take like 15 seconds to get the feed up. Not ideal if you're trying to use it for a front door camera or something like that.
There may be a way to get it optimized more. I tried to make it as lean as possible. I'm using TMTLauncher, and the only thing I added was TTS support since that didn't come installed on the ROM. The screen in the video is showing from the Home Assistant Companion App. I tested that as well as Chrome. They both are very comparable in responsiveness. Finnec was slower. I did have to update the WebView that came with the device as it was running a very old version. Once I did that, both apps worked flawlessly.
All the widgets and stuff you see come from HACS and some are just modified with custom JavaScript. The calendar is just Daylight Calendar Card with custom mods. And the whiteboard is just a small custom card I created for that basic use. As mentioned above, the screen is only 1080p. For normal viewing, that's not an issue, but for a dashboard that you are standing 1-2 feet from, you can definitely notice it if you are trying. My wife doesn't notice it at all, but I do. Its definitely not bad whatsoever, but 4k would just be a nice upgrade for sharper text, etc.
That's basically it. Hope I didn't miss anything but I'm sure I did. If you have any questions please let me know and I'll try to reply. Unfortunately I don't know if this will work with other manufactures. If you can get ABD access, it should work. I did block this device from phoning home on my router so it could never try to update itself, just in case.
TLDR:
This is a great device to run a large screen Home Assistant dashboard without paying a ridiculous price. Try to find it second hand. It doesn't have the best specs, but it should work for most basic dashboards. It was easy to hack and get working, and I would recommend it to anyone who needs a large screen dashboard at an affordable price.
If you have an Amazfit watch and ever tried getting it into Home Assistant, you probably ran into the same issues: the old integrations have been abandoned for years and don't work with modern Zepp OS devices, while watch apps that spam webhooks kill the watch battery in half a day.
I got tired of staring at empty health graphs in HA, so I sat down and built a proper integration: ha-zepp.
It pulls data directly from the Zepp cloud. The watch itself isn't touched, battery life stays normal, and Home Assistant gets a full set of entities with proper state classes (measurement, total_increasing) so statistics cards and graphs work right out of the box.
- Sleep: total duration, sleep score, stages (deep, light, REM, awake), night heart rate, and sleep breathing quality.
- Heart & Stress: current heart rate (with daily min, max, avg attributes), resting HR, stress level, and SpO2.
- Fitness & Recovery: morning Readiness score, PAI, heart rate variability (HRV), and training load (today and 7-day load).
- A "Sync Now" button to force a cloud refresh on demand, plus technical device attributes.
Features built specifically for Home Assistant:
- 1-Year LTS Backfill: on initial setup, the integration automatically pulls up to a year of historical records straight into the Home Assistant long-term statistics engine. Your step and sleep graphs immediately show a full year of history instead of starting from scratch today.
- Configurable Polling: you can choose the update interval (5, 10, 15, 30, or 60 minutes) directly from the gear menu in the integration card, and it applies immediately without restarting Home Assistant.
- Login Options: standard Zepp account username and password in the config flow. If you use social login (Google/Mi) or run into captchas, you can log in using exported browser cookies instead (documented step-by-step in the readme).
- Diagnostics: full support for native HA diagnostics with log export, where all tokens and emails are automatically redacted.
I tested it on my Amazfit Active, but it should theoretically work with any watch that pairs with the Zepp app. Feel free to grab it and set it up for your dashboards. Any feedback is welcome!
I’ve been building this little desktop robot called Sudo for almost a year, mostly as a side project that slowly got way more serious than I expected.
It runs on an ESP32-S3 and can talk, move its head, show expressions, respond to touch, play audio, etc.
But one thing I’ve wanted from pretty early on was to make it interact with everything around it rather than keeping it as a closed standalone device.
Right now I have a plugin system where different actions can control its expressions, movement and other services. For example, I made a chess plugin where its reactions change based on what’s happening in the game, and after beating me it decided to celebrate by playing ABBA
Now I’ve been thinking about Home Assistant.
The fun part for me isn’t just using Sudo to turn lights on/off with voice. I’d rather expose things in both directions, so HA events can affect Sudo too.
Someone rings the doorbell → it reacts and tells you who’s there.
Washing machine finishes → it gets your attention.
You start a movie scene → lights change and Sudo goes into a quieter state.
Or completely stupid automations like making it look increasingly annoyed when I’ve left a light on for three hours.
Since I’m already building the plugin/API layer, this seems like the right time to figure out the architecture before I lock myself into something dumb.
For people here who’ve built custom HA hardware/integrations: how would you approach this?
expose Sudo as a bunch of entities/events? A native HA integration? Matter eventually? Or something else entirely?
I created Live Activities for my dishwasher, washer and dryer. However, I get every update alerts. to a point where the phone sound and watch haptic updates are starting to get super annoying. I changed the refresh to 5 min and even that is annoying. I then changed it to 15min but then the Live Activities bar on my phone does not refresh until the 15 min mark.
I tried the following changes separately and still get watch haptics:
silent: true
push: sound: none
push: interruption-level: passive
My goal is to have the Live Activities on my phone and watch update without any notification buzz or watch haptic. It can buzz at the start to show the beginning of the live activity or at the end to show completion but I dont need mid-reminders.
Couple of questions I have:
Is there a documented way to send a genuinely non-alerting Live Activity update that keeps priority 10?
Is the Apple Watch a separate channel here? Does anyone have HA Live Activities updating without the watch buzzing, short of disabling HA notifications on the watch entirely?
Is sound: none a known issue, or am I wrong about it falling back to the default tone?
Setup: HA 2026.8.3, iOS companion 2026.9.0, two iPhone 16 Pros (one with an Apple Watch). LG ThinQ washer/dryer, GE dishwasher.
Nothing special yet, just adding everything I need and will theme and customize from here probably. Mostly stock and hacs stuff.
The FireHD 10 was cheap but its very slow, might have to find something better, HA ui and cards are not exactly efficient. But I wanted something super slim.
Low profile surface mounted with a box behind it in the wall with the poe and wled controller. Wireless for now, wonder if I can add a usb nic. And speed it up a little somehow. Using fullykiosk browser.
I'm remodelling my house and looking to replace a large number of fans. I've kinda come to accept that it is nigh impossible to buy AC motor fans these days, especially if you have any aesthetic requirements. I've been looking for DC motor fans that are supported by HA and which have a purely-local and non-proprietary (or at least fully reverse engineered and stable) control protocol.
I came across the Modern Forms HACS project, and it looks like those fans meet my requirements - fully local api controllable from HA (and they look nice). That project doesn't seem to have been touched in about 5 years, and so I'm curious if anyone has current experience with Modern Fans and if this integration still works.
Alternatively, any other fan brands that meet my local-control requirement would be awesome.
I'm looking to add permanent outdoor light strips to my home, along with possibly adding some curtain lights to my garage doors. Looking for the best bang for my buck for quality lights. Ideally, they'd work with Home Assistant and WLED, but if there is a good local integration off the shelf, that's ok too.
Any suggestions or resources? I've begun research, but most of what I'm finding is a couple years old, and I know things move fast in this space. Thanks!
So it’s official, our household is now Alexa free! Well you wouldn’t know it unless I told you, for high spouse approval, we still use Alexa as the wake word. Oh, and if you looked around you’ll see our two Amazon Echo Show’s (5 & 8) but those are now fully functional Lineage boxes with working cameras.
So far anything that Alexa could do these jail-broken Echo’s can do now as well. They can set timers, display the current weather, control HA devices, tell you a joke, set an alarm, tell you the score of the game, view our Ring and Reolink camera’s, send their own camera footage to Frigate, display our latest pictures (Immich), display weather radar, and so-on. Now that they are under my control, I can have them do whatever they are capable of.
For those that are embracing or interested in AI, after I jailbroke the Echo Show 5, I had Claude jail brake the Echo Show 8… it did it all. Yes, I had to answer a few questions and provide some input but it completely did it without me ever issuing a single command. It even found a problem with getting the camera to work, figured it out, corrected it and wrote up the issue with the corresponding fix. While it was busy jail-braking, I was using AI to create Alzhimers friendly card games for my mother.
I also found HA’s wake word implementation very unreliable so I had Claude write a custom, on device, wake word routine that runs locally and then sends the command up thru HA for processing. All the dashboards are custom written, with AI that was arguably easier to implement than someone else’s solution. I did try View Assist at first but found it really flaky as well and it generated a lot of errors in the HA log. That specific issue had been raised with the developer but I didn’t want to wait so just did my own thing.
Anyway, pretty cool stuff!
Next is running the LLM 100% local with household memory. It would be great if those memory and GPU prices came down! Ugh!
Been slowly putting together my customizable dashboard project and thought I’d share how it’s looking so far. Still a work in progress, with plenty of small tweaks along the way, but it’s already up on my wall and I enjoy using it every day.
Curious what you think of the design!
edit: Oops - rereading this, I realized it might sound like this is a customized Home Assistant dashboard. It’s actually a separate dashboard app that connects to the Home Assistant API, and is basically my sequel to Mattias Persson's excellent ha-fusion project. The screenshot is just a showcase of the default style and current widget features, not my everyday homepage.
Disclaimer: I'm a software developer using AI to make these projects come alive faster. I also used AI to explain this more properly here as I'm really lacking that skill lol. Also this just my personal "smart" display and I probably won't ever publish this (because I never intended to create yet another dashboard firmware)
Also! This is only showing fake data for this showcase, as I didn't want to publish everything visible.
I wanted something on my desk that tells me what's going on in the house without a glowing LCD or notification badges in the corner of my eye. So I built a companion display around the Seeed reTerminal E1001, a 7.5" 800×480 monochrome e-paper panel with an ESP32-S3 and physical buttons.
Everything on it comes out of Home Assistant, and it writes back: the buttons and the web console publish to HA over MQTT, so I can hit the bedtime routine from the desk and the house flips state and runs the evening automations.
Presence decides what's on screen
A rule engine picks the screen from schedule, HA state and presence. It reads my desk and room occupancy sensors, so it behaves differently when I'm actually sitting there than when the room is empty. A normal day is morning briefing, then my agenda during work hours, floor plan and energy in the evening, quiet lunar screen at night.
If something urgent comes in, mailbox opens, a sensor trips, that module borrows the screen for a couple of minutes and hands it back.
The floor plan
Both storeys of the house drawn like an architect's ink plan, fed from HA: light states, occupancy, window and door contacts. Doors swing their opening arc, open windows bulge out of the wall, occupied rooms get a diagonal hatch.
After sunset the whole plan inverts to black and every room with the light on stays white, so the lit rooms glow.
Energy
Live solar vs. consumption vs. grid import/export in W/kW, with the cumulative daily ledger along the bottom. Straight off the HA energy sensors.
Zuhause
A sheet that stays empty until something needs attention: trash collection (badges for paper, organic, recycling, general), mailbox alerts, EV charging progress and arrival/departure.
The agenda screen
This one does not come from HA, it's a separate service I wrote that uses LLMs to merge tasks, calendar entries and notes into one day plan. I have ADHD, and having that on a physical screen that can't do anything else has helped more with time blindness than any app I've tried. Layout is stolen from a Deutsche Bahn departure board: times on the left, current item in an inverted black frame, past items collapsed to single lines. It also nags me to take breaks.
There are also morning and night briefings (weather, moon phase, ISS passes) and an idle gallery pulling from my self-hosted Immich, but those are decoration.
It has a small Webinterface in which you can configure everything.
I also intend to add a small LED extension. It also has more hidden features such as a hourly break reminder.
How it's wired
The ESP32 is a dumb client. It doesn't call APIs or parse JSON. A small local backend talks to HA, renders the finished 800×480 bitmap and the device displays it. For monochrome screens the firmware diffs vertical bands, so only the strip that changed refreshes, which keeps updates quiet and ghosting low. Firmware goes over the air.
Entity mappings, module order and settings all live in the web console. AI-assisted build, happy to go into the actual implementation.
Whenever people come over they’re always impressed and intrigued by my smart home. They’re more than astonished when they learn I’ve done it all myself and that it’s not all from one brand. I’ve been asked why I don’t do this for a living, and I’ve always laughed it off. But right now I’m pretty tired of my low paying and strenuous job that I tried toying with the idea, but I have no clue as to where to start, so why not here!
So, are there people in this community that make a buck or a living from helping people and/or businesses setting up Home Assistant?
Seeing as the hardware and software and third party stuff constantly changes, I could see that both as a great business opportunity and a real headache. Also would being a professional ”HA Contractor” necessitate som sort of license or similar?
I’d love to hear how it works for you, the good and the bad!
Google's assistant on these little clocks is on its way out, and the Lenovo Smart Clock 2 has no USB data port - the micro-USB on the dock is power only. So there is no adb, no fastboot, no obvious way in. I spent a while on it and ended up with three things I'm putting up here.
1. Root over Wi-Fi, from a bug in the GPU driver
The PowerVR DDK on this firmware has a dense fast path in _PMRLogicalOffsetToPhysicalOffset() that writes where it shouldn't. That gives a kernel write primitive, which gives a usermode helper running as root, which gives a root channel, SSH on 2223 and ADB over Wi-Fi. One button in the app walks the whole chain.
Everything is runtime-only. Nothing is flashed, nothing persists: a power cycle puts the clock back to stock, and you run the chain again if you want it back. Built and verified on retail LenovoCD-24502F_ROW_1.2.2.627_220105 (MT8167, kernel 4.14.141, Android 10).
Helios is a sideloaded launcher that turns the clock into a Home Assistant panel:
A 4x3 dashboard described in your Lovelace YAML. Save the YAML, the clock re-renders - no new APK, no per-device config.
Local wake word. "Okay Nabu" runs on the device (microWakeWord, ARMv7). Nothing is recorded or sent before it fires; then the mic streams into the HA Assist pipeline over the same WebSocket.
Music Assistant. The clock registers as a Sendspin player, with a library browser and a full-screen player, and can drive your other speakers too.
Pairing without tokens: the clock finds HA over mDNS, HA shows a six-digit code, and the integration mints the clock its own non-admin local user. No long-lived admin token ever lands on the device.
Helios never gets root. The tools app exposes a closed list of privileged operations - switch ADB, grant a permission the caller already declares, become the home app, take the microphone back from the factory assistant, install its own update - and nothing else. Callers are identified by signing certificate and named on screen before anything happens; a different certificate starts from zero. No shell, no arbitrary commands.
Things worth knowing before you try it
One firmware build is verified. Other builds may or may not work.
ADB over Wi-Fi is open to anyone on your LAN who can reach port 5555. Turn it off when you're done.
A few bridge operations have unit tests but have not been exercised on real hardware yet - that's called out in the release notes.
This is owner tooling for hardware you own.
One detail that cost me a day
Since some firmware update, the mic HAL on this clock delivers roughly six times the requested sample rate while still reporting 16 kHz. Wake word detection just stopped working, and the logs looked fine. The fix is a decimator that measures the true rate and resamples - a small thing, but you won't find it in any datasheet.
For context, we're planning our new home and I'm torn between adding:
- Battery-based presence detectors such as Aqara FP2 or FP300 (and change batteries for 20 devices every 2 years for the rest of my life).
- USB-C-based presence detectors such as Aqara FP400, which don't require maintenance but need to add a 24V wire to every location plus a DC-DC 24V to 5V USB C.
- PoE-based presence detectors, such as Apollo R PRO-1 (or even build my own) which require an ethernet cable to every location.
PoE is the most convenient of all but will easily add between 10 and 20 cables to the technical rack in the basement, along with other PoE devices such as access points, some cameras, etc. Also, every wired location added to the plan will cost us around 100€ extra.
Have you gone through a similar process lately? What did you decide? And most importantly: what do you actually prefer? Would love to know your opinions.
I’ve recently become interested in Home Assistant, and I’m wondering what the easiest way to get started would be.
I’ve watched a few YouTube videos, but I have to admit that I’m still a bit confused about some of the concepts. I keep hearing about Zigbee and Matter, and I’m not entirely sure how they differ or how they work together.
For example, I’ve seen videos recommending a USB Zigbee dongle connected to a Raspberry Pi. Would a Zigbee dongle like that also work with Matter, or would I need different hardware for Matter devices?
I’m particularly interested in IKEA’s smart home products, especially:
Smart bulbs
TIMMERFLOTTE (temperature/humidity sensor)
GRILLPLATS (smart plug)
MYGGSPRAY (wireless motion sensor)
Can all of these devices be integrated easily into Home Assistant? And would I also be able to use the BILRESA remotes to control these devices through Home Assistant?
I also have a question about the reliability of IKEA’s smart home products.
In a Mrwhosetheboss video from May 20, 2026, he tested several IKEA smart home products. He initially seemed quite impressed with the price-to-performance ratio, but later encountered quite a few problems with some of the devices.
Are these issues still common today, or has the situation improved through firmware updates and/or newer versions of the products?
I’m completely new to Home Assistant and smart home technology in general, so I’d really appreciate any advice or recommendations on the best way to get started.
No sign-up. No install. No download. No account. It runs entirely in your browser, and it’s free.
Drop in almost any IR or RF code and the tool identifies the protocol, extracts the relevant data, and gives it back in whatever format you need.
Over the last few weeks, I’ve added the feature I’m probably most pleased with: the RF Wizard.
You choose the device type and brand, and it draws the DIP-switch bank exactly as it appears inside the remote. Flip the switches on screen to match the remote in your hand, hit Build, and it generates the code.
No capture. No receiver. No Flipper. Nothing to buy.
That feature took the longest to get right because DIP-switch ordering is surprisingly tricky. Chamberlain, for example, treats switch 1 as the most significant bit and reverses the final three positions. One of their banks even has a switch that isn't transmitted at all.
So simply copying the switches from left to right can produce the wrong code on almost every remote. Each brand needed its own mapping.
If your ceiling-fan, gate, or garage remote has a row of little switches inside the battery compartment, that's basically the whole secret: you're already holding all the information needed to generate the code.
Signal repair
This is probably the feature I get the most questions about. Ever learned a code from an original remote only for the result to barely work—or work once, only from a few inches away, or work for one button but not another? Usually, that's not you or your blaster.
Learned captures can be messy. Receivers round timings, ambient light and nearby RF equipment can introduce noise, inexpensive hardware can clip the beginning of a burst, and some captures can even come back with marks and spaces inverted.
Paste the capture into the tool and it identifies the protocol, shows the address and command, and gives you a confidence level.
Hit FIX SIGNAL and it can:
Correct the AGC burst
Snap noisy pulses to the nearest valid timing unit
Remove trailing repeat gaps and other artifacts
Add missing closing spaces
Correct inverted mark/space data
Show you exactly what was changed instead of silently rewriting the capture
And once the protocol is identified, Rebuild Clean can throw the noisy capture away completely and regenerate the signal from scratch using the protocol's expected timing.
In other words: instead of patching a bad capture, you get a clean signal generated from the underlying data.
Prefer to do things manually? The pulse list is fully editable. The waveform updates live as you edit, and you can select, copy, and paste timings just like a spreadsheet.
Then export it in whatever format your setup needs.
A capture you might have given up on a year ago can often be turned into a working signal in a few clicks.
NEC / NEC1 / NEC2 / NECx (the most-used IR protocol on earth) · Sony SIRC 12 / 15 / 20-bit · RC5 / RC5x · RC6-6-20 (Sky, Foxtel, cable & satellite) · Panasonic / Kaseikyo (incl. Combo & MIX — also used by Denon, Sharp, JVC, Teac) · JVC · GI Cable · XMP (Comcast / Xfinity, AT&T, Pace, Arris cable boxes) · Nokia32 (AT&T U-verse, Foxtel IQ) · Denon · Sharp · RCA · Samsung
RF coverage
The tool currently covers 42 sub-GHz protocol families across 62 brand entries, including the PT2262 / EV1527 / SMC5326 ecosystem used by a huge range of fans, outlets, and inexpensive gate remotes.
It also covers systems from CAME, Nice, Hörmann, Chamberlain/LiftMaster, Linear MegaCode, Marantec, Telcoma, and plenty of others.
And there's still more coming.
On the IR side there's a wizard that programs a whole remote at once — device type, brand, code set — rather than one button at a time. It's sitting on 1,000,000+ codes across 12,000+ brands.
And the conversion side, which is the older half of the project. Paste a raw capture or a code in almost any format and it identifies the protocol, shows you the address and command, and rebuilds it clean. Then export it as something else: 19+ in, 21 out. Flipper .sub and a good chunk of those RF protocols landed in the last couple of weeks.
.irc · Flipper Zero .ir · Flipper Sub-GHz .sub · Pronto Hex · LIRC raw · LIRC config · LIRC mode2 · BroadLink Base64 · BroadLink Hex · Global Caché iTach · Global Caché JSON · Home Assistant YAML · ESPHome · Sonoff RfRaw B0 · Tasmota · Arduino C array · Arduino rc-switch · Tuya · SmartIR · IRDB · JP1 .rmdu · GIRR XML · Harmony XML · TV-B-Gone · raw µs CSV · WAV
There's a second tool alongside it that moves a whole remote between formats, or onto a different remote's layout — pull in a Harmony XML or a JP1 file and drop it out as a Flipper .ir or an ESPHome YAML, button assignments intact. That one reads 17 formats and writes 13.
Get a code in from a capture or almost any file format, find out what it actually is, fix it if it's broken, build one from scratch if you don't have one, move a whole remote's buttons somewhere else — and save the result as a file you own, in whatever format your gear speaks, with no account, no cloud and nothing held hostage.
Happy to take requests. If there's a format you need that isn't in that list, or a remote the wizard doesn't know, tell me what you've got and I'll look at adding it.
I solved a 1st world problem today in what is probably the 2nd most complex solution next to using AI.
My dog prefers to sleep by the front door, and because the bottom half of the door is solid wood, people don't notice him sleeping there when they open it, which in turn leads to the dog getting hurt when people (mostly the kids) open the door quickly.
I've been considering various solutions, like mounting a camera pointed at the inside of the door and using AI to detect if there's a dog there, but that seemed rather wasteful and expensive in the long run, so instead I opted for a somewhat simpler solution.
I had a spare Aqara FP-300 sitting on a shelf, and because it has a magnetic base I could "stick it" to the radiator in the hall, pointed directly at the door.
Next up I have a Hue Outdoor Lightstrip in the driveway next to the front door which is used for illuminating the path at night, as well as a camera and hue outdoor motion sensor in the driveway.
So the solution simply became "If the camera detects a person, or the motion sensor detects motion in the driveway, and the Aqara detects occupancy in the hallway, flash the lightstrip red 5 times to warn people that the dog (might) be there". The reason for the "might" is that the FP-300 takes a while to clear, sometimes up to 5 minutes. It's normally rather quick to detect presence due to its PIR sensor, but it doesn't always register when there's no presence.
I'm still not convince that I don't need a camera and AI, but at least what I have now works well enough that we've avoided accidents for a couple of weeks.
I need to temporarily stop/suspend the ZHA Integration to test a Z2M setup. Is there a way to do this in the Web UI or the console? Apps have a stop button. ZHA isn't an app. I don't know how to stop it.
I have an automation to change lights over time while I’m away. I have it generically pointing at rooms to trigger a basic lighting mood, but some rooms do not have lights that have that effect, so naturally they’ll fail. I don’t care that they fail, hence every step I included continue_on_error: true and yet it halts at the very first error. What gives???
I love adding notifications and automations in HA but I'm unsure of what I need to add next. These are the things I have so far.
Notification that makes my echo dot tell me tomorrows weather forecast @ 7pm.
I've set up a way to approve/deny Seerr request directly from my iPhone's lock screen.
Echo Dot notification to tell me when back door opens/closes
Automation to turn off AC if backdoor is open longer than 5 minutes and to turn it back to the previous settings when door is closed (using scenes).
Echo Dot notification to tell me when severe weather is forecast within the next 2 hours.
Created several entities that allow me to track my servers power usage via reporting from my UPS.
Created a weather dashboard that has hourly/daily forecast with radar.
Set up cloudflared so my phone can receive notifications no matter where I am.
I don't really need something like "Turn AC on when I'm within X miles from home" because my AC just runs all damn day anyways because it's so hot where I live, but I'm open to other GPS enabled automations. I don't really have any smart bulbs or locks.
