A thin recorder is easy to carry around. But the real evaluation of a work tool begins after the recording stops. Can the conversation be turned into text, the necessary parts edited, and everything moved to existing documents or a PC? For an AI recorder to settle into everyday work, the way information exits the device matters just as much as its dimensions.
The NWO JP50, introduced on Makuake, is a smartphone-attachable AI voice recorder that claims a thickness of 3.0mm and a weight of 28g. In addition to transcription and summarization via the dedicated DOWAY app, it explains multi-format file output and syncing to a PC. Laid against the Meydo S1 concept, the boundary between moving information and actually advancing work with it becomes visible.
Figure 1 Official listed images of the NWO JP50 and the Meydo S1 concept. Not to actual relative scale. The S1 is at the stage of product concept and user-needs validation.
Splitting roles between the thin body and the app
The operator is NWO JAPAN LLC. The FAQ states that recording can be started and stopped on the body alone, while checking transcriptions and mind maps requires a smartphone with DOWAY installed. Compatibility requirements are iOS 15 or later and Android 8 or later.[1]
What to consider in this configuration is the difference between being able to record and being able to use AI features. The small body serves as the entry point for capturing audio; organizing and viewing are handed off to the app. When deciding on a purchase, it is worth treating portability together with whether your existing devices can perform the necessary processing and verification as a single environment.
The official materials explain that the body may warm up during long continuous recording or transfer, and that pickup performance is constrained in noisy outdoor settings or when the microphone is covered by clothing. The headline feature of thinness comes with usage conditions of its own. Ease of carrying and recording at the same quality everywhere are two different things.
Figure 2 The controls shown in the official video. It illustrates mode selection for different uses. It is not a figure guaranteeing recording regardless of calling method or device combination.
Input conditions differ between meetings and calls
The body has NOTE and CALL modes, described as switching with a double press of the button. NOTE uses dual MEMS microphones and claims pickup of up to about 7m. CALL is described as using bone-conduction principles, but it has not been confirmed that results are identical across all phones, cases, and calling apps.[1]
Picking up voices traveling through the air in a meeting room and recording a call on a device are different input conditions. Making the modes explicit is also an operation that forces users to recognize that difference. To judge practicality, the first question is whether you can verify the content — including the other party's side — in the calls or meetings you actually want to record.
Transcription and translation are described as supporting 80 languages. Industry-term settings, summary templates, and mind maps are also provided. However, the FAQ notes that complete transcription is difficult with technical jargon, strong dialects, or simultaneous speech. Judging work accuracy from the number of languages or model names alone is premature.
Figure 3 Transcription and summarization shown in the official video. Condensing content and correctly preserving the speaker's intent are two things to check separately.
Output formats become the contact point with existing work
The NWO JP50's stated outputs are TXT, SRT, DOCX, and PDF. Re-editing text, using it as subtitles, sharing it as a document — even for the same recording, the required format changes with the next task. Having format options is the condition for not locking information inside the dedicated app.[1]
Wireless syncing to a PC, saving recording data, and link-based sharing are also introduced. But syncing means, first of all, making information usable somewhere else. Data moving to a PC does not automatically advance the work of selecting a quote, checking the recipient, and sending an email.
Even if meeting minutes can be output as DOCX, editing to match internal formats and deciding distribution lists remain. A share URL does not replace the management of who gets what. Beyond whether files can be exported, what matters is how seamlessly they can be handed to the next step of your own work.
Figure 4 Official example of handling records on a PC. It explains data syncing; it is not evidence that AI autonomously operates other applications on the PC.
Second-year conditions deserve the same weight as thinness
The cost does not end with the hardware. In the official post-second-year pricing chart, the free plan's transcription is capped at 300 minutes per month, and some template limitations apply. The paid unlimited plan is listed at ¥5,000 per year on iOS and ¥4,520 per year on Android. The displayed figures of about ¥417 and ¥377 per month are monthly conversions of the annual amounts.[1]
On the cloud, the operator explains that it runs on Microsoft Azure, uses AES-256, and does not secondarily use data for AI training. These should be referenced as the provider's statements; this article has not audited third-party certifications or actual operations. For internal use, they must be checked against your own requirements, including storage locations and sharing settings.
Crowdfunding purchases ended on September 26, 2026. The returns included a limited tier at ¥18,800 including tax and a planned general retail price of ¥29,800, with shipping planned by the end of December 2026. These are not figures representing current sales conditions or completed deliveries, and the FAQ notes the possibility of future specification changes on the app side.
Figure 5 Mode selection and recording, organizing via DOWAY, verification and output, separated out. This is an arrangement of official materials, not a measurement of performance or processing time.
What Meydo S1 envisions lies beyond the handoff of information
The Meydo S1 materials describe a flow that not only organizes audio and images but connects what the user has verified to subsequent operations. Because the S1 is at the stage of product concept and user-needs validation, its performance cannot be compared with the NWO JP50 as a finished product. What can be laid out is the design difference in where each decides work ends.[2]
When a record made on the NWO JP50 is synced to a PC, the information having moved is one stopping point. In the S1 concept, combining USB or KVM with AI screen understanding to connect to operations in permitted apps is under consideration. This is not an explanation that the two products will integrate; it is a comparison showing that moving files and executing work are separate design problems.
Figure 6 An S1 meeting scenario based on provided materials. It does not show integration that automatically imports NWO JP50 data, nor does it show implemented features.
Delegating operations requires units of confirmation
Take a task like “send the meeting minutes to participants.” Beyond generating the document, there is choosing the version, specifying recipients, and checking attachments. The S1's screen plus the AI, confirm, and cancel physical buttons are envisioned as the contact points where the user decides on these next actions. Compared with simply passing data along, it becomes necessary to make the target of approval explicit.[2]
Long-term memory is envisioned as storing information the user has approved, and FoneClaw is described as an option handling the syncing of records, memory, permissions, and task states. The more work continues, the more management is needed to know which instructions are in effect and what has already been executed. How far these can be realized remains to be verified.
Figure 7 User verification and feedback of execution results as envisioned by the S1. It does not represent operation confirmed through hardware testing.
Therefore, when evaluating the S1 concept, the question is not only the big one of whether it can handle arbitrary apps, but also whether it can be stopped when the wrong target is selected, whether the permitted scope is visible, and whether failure is detectable. At this stage, no confirmed information on weight, battery life, pricing, shipping timing, or compatibility can be added.
Figure 8 Comparing the output and syncing described by the NWO JP50 with the verification-to-execution connection drawn by the S1. It does not indicate a performance ranking or the final specifications of the S1.
A recorder's value is decided by the next step
The point of reading the NWO JP50 is its structure: a thin body as the entry point, with information connected onward to the app and multiple output destinations. The S1 concept widens the field of view to the act of using that information. Is the burden the recording during transit, the document editing after recording, or the approval and sending? Pinning down which step still carries the burden is the starting point for choosing an AI recorder.
