There is no AI device on the market today that reliably induces Lucid Dreams AI. The best-known attempt, Prophetic’s Halo headband, took $100 refundable deposits starting in late 2023 for a device estimated at $1,500 to $2,000 and promised for spring 2025. It never shipped in that form.
In April 2026 the company announced two different products instead: Dual at $449, expected at the end of 2026, and Phase at $1,299, expected mid-to-late 2027. Anyone who reserved a Halo three years ago is still waiting, and the thing they reserved no longer exists as described.
What Happened to the Halo
Prophetic was incorporated in January 2023 and launched that June, founded by Eric Wollberg and Wesley Louis Berry III. The pitch was ambitious from the start. The company suggested the technology could effectively give you an extra six to eight hours a night to work toward your goals, and its CEO compared the device’s potential to what Meta and Apple were chasing with mixed reality.
The Halo was to combine EEG sensors with transcranial focused ultrasound, aiming stimulation at the dorsolateral prefrontal cortex, the region associated with the self-awareness that defines a Lucid Dreams AI. Industrial design came from Card79, the studio behind Neuralink’s N1 implant. Prophetic partnered with the Donders Institute at Radboud University in the Netherlands to build a dataset of EEG and fMRI recordings from lucid dreamers.
All of that is real work. None of it produced a shipping product on the announced timeline.
What Prophetic Is Actually Selling Now
| Dual | Phase | |
|---|---|---|
| Price | $449 | $1,299 |
| Expected shipping | End of 2026 | Mid-to-late 2027 |
| Ultrasound stimulation | Yes | Yes |
| EEG sensors | Not stated | Yes |
| Closed-loop adaptation | Not stated | Yes |
That EEG row matters more than the price difference. A closed-loop system reads your brain activity and times stimulation to your actual sleep stage. Without it, a device is stimulating on a schedule rather than in response to what your brain is doing, and REM timing varies substantially between people and between nights.
Coverage of the two devices does not clearly establish whether Dual can detect sleep stage on its own. Before paying $449, ask that question directly. It is the difference between a device that works with your sleep and one that guesses.
Both dates are expectations, not commitments. Given the Halo’s history, treat them accordingly.
Morpheus-1 Is Smaller Than You Think
Prophetic’s AI model is called Morpheus-1, and the framing invites you to picture something in the class of the models behind ChatGPT or Veo. It is not.
Morpheus-1 is a 103 million parameter transformer trained on eight GPUs over two days. Rather than generating text or images, it takes brain activity as input and outputs shaped ultrasonic patterns intended to interact with that brain state.
The architecture is genuinely interesting and the small size is not automatically a criticism, since this is a narrow task on a specialised signal rather than general language modelling. But two days on eight GPUs is a research prototype’s training budget. It is worth knowing when the marketing language leans on the word generative to borrow credibility from much larger systems.
What the Evidence Actually Shows
This is the section most coverage skips, and it is the one that should shape your decision.
Researchers at Northwestern University built a phone app delivering audio cues and tested it with 19 people over a week. Lucid dreams rose from an average of 0.74 in the previous week to 2.11 during app use, a large jump for this phenomenon. The study appeared in Consciousness and Cognition.
Then the lead researcher, cognitive neuroscientist Karen Konkoly, added the caveat that the marketing never repeats. It was not clear the sound cues caused the increase. Simply focusing on lucid dreaming for a week, or expecting results, might explain it on its own.
That is a researcher voluntarily undermining her own headline finding, which is what good science looks like and what product pages never do. Nineteen people over one week, with the causal mechanism unconfirmed, is where the consumer-facing evidence base currently sits.
On Prophetic specifically, Guy Leschziner, a professor of neurology and sleep medicine, told Live Science it was difficult to say whether Morpheus-1 achieves what the company claims. He acknowledged the roadmap targets the right brain circuits. Targeting the right circuits and reliably producing the intended state are different achievements.
The Apps You Can Actually Use Today
While the hardware slips, software exists and costs far less.
Lucidly runs on Apple Watch and iPhone, using neural networks to predict REM from the watch’s health readings and then delivering haptic or audio cues. It works alongside established induction techniques rather than replacing them, and you can adjust thresholds and alarm settings yourself.
What stands out is the honesty. The company states plainly that haptic feedback is not guaranteed to induce lucid dreaming, that prediction accuracy varies between people and between nights, and that a wrong prediction may deliver a cue during deep sleep or while you are nearly awake. That is a more accurate description of the state of the art than any headband marketing.
A consumer smartwatch cannot see brain activity. It infers sleep stage from heart rate and movement, which is a reasonable approximation and not the same thing as EEG.
The Trade-off Nobody Advertises
Here is the part that deserves genuine care.
The main induction techniques these tools support, including wake-back-to-bed, work partly by interrupting sleep. A peer-reviewed CHI paper on a lucid dream induction prototype notes directly that sleep disruption is associated with higher lucid dream frequency, and its authors speculated that disruption from their own app might raise the induction rate.
Read that carefully. Fragmenting your sleep is not a side effect of the method. In part, it is the method.
Waking yourself in the middle of the night, several nights a week, to chase a dream state has a cost, and that cost lands on the same sleep that governs mood, memory consolidation, and immune function. Occasional experimentation is one thing. A sustained routine built on deliberate night waking is another.
A few things worth knowing before you start. Lucid Dreams AI attempts are associated with sleep paralysis for some people, which is harmless but genuinely frightening the first time. If you already have a sleep disorder, take medication that affects sleep, or live with a condition where the boundary between sleeping and waking states matters clinically, talk to a doctor before deliberately disrupting your sleep architecture. This is one of those areas where a five-minute conversation with a professional is worth more than any product page.
For context on how to weigh AI health claims generally, our look at ChatGPT Health covers the same question from the clinical side, and our piece on AI dermatology tools examines what happens when consumer AI makes medical-adjacent promises.
If You’re Curious Anyway
Lucid dreaming is real. Roughly half the population has experienced one, though only about 20 percent of those have them monthly and around one percent weekly. Nothing here suggests the phenomenon is fake. The question is whether AI hardware currently adds anything to it.
A sensible order of operations:
- Start with a dream journal. Free, evidence-supported for recall, and it costs you nothing but a notebook. Most people who never get past this step were never going to get value from a $449 headband.
- Try a phone or watch app next. A few dollars a month tells you whether cued induction does anything for you personally.
- Set a limit on the disruption. Decide in advance how many nights a week you are willing to interrupt, and hold to it.
- Wait on the hardware. Let Dual ship, let independent reviewers test it against a control, and then decide.
On deposits specifically: the same caution applies here as with any pre-release purchase, which we covered when looking at tools still in alpha with unpublished pricing. Money handed over before a product exists is money you do not control.
Frequently Asked Questions
Can I buy an Lucid Dreams AI device right now?
Not a shipping one. Prophetic’s Dual is expected at the end of 2026 and Phase in mid-to-late 2027. Apps for phones and smartwatches are available today and cost a fraction of the price.
Does the Northwestern app study prove these work?
No. It found an increase across 19 people over one week, and the lead researcher stated it was unclear whether the audio cues caused it rather than simple expectation and focus.
What happened to the Halo?
It was never released as announced. The Halo prototype was superseded by the Dual and Phase product line announced in April 2026, at different prices and on later timelines.
The Practical Call
Keep your money for now. The science is early, the flagship device has slipped by more than two years and changed shape twice, and the only published consumer study comes with a caveat from its own author.
If lucid dreaming genuinely interests you, a notebook and a few weeks of consistent dream recall will teach you more about whether this is for you than any headband would. If a device eventually earns it, it will still be there. And whatever you try, protect your sleep first. The one thing every credible source in this field agrees on is that the technique works by taking sleep away from you, and that is a real price whatever the hardware costs.
Published: September 10, 2026
