When a technical lead at a toy factory first comes across an AI toy module, the word they most often get stuck on is "dual stack". What exactly is a dual stack? And what is a command module? In the years we have been running the TY / LX / NT product families, this is the question we are asked more than any other. We are laying the architecture out plainly today so that you are not led astray by terminology when you choose, and so that you can follow what a solution provider is really saying when you talk to one.
What a dual stack actually is: local SDK + cloud API
Start with a definition. When we say "dual stack", we mean that one and the same module carries bothlocal SDK andcloud API capabilities at the same time: the local SDK covers wake-up, playback and the offline fallback for basic interaction, while the cloud API covers LLM conversation, character memory and content updates - the parts that need compute and continuous operation. Building on the Tuya T5E / Espressif ESP32S3 / PY32 platforms, Nablai packages three routes into the TY / LX / NT product families: TY (Tuya T5E) gets you overseas fast, LX (Espressif ESP32S3) gives you full brand sovereignty and ships with the Nablai app and the AMS console, and NT (PY32) is a low-cost way to test the market.
One common misconception has to be cleared up here: TY002 and LX002 both sit on the same "local SDK + cloud API" dual-stack architecture, and the difference lies only inecosystem ownership and brand sovereignty - TY runs the full Tuya stack out of the box, while LX runs Espressif plus the Nablai private cloud and leaves brand autonomy completely intact -it is absolutely not a case of treating "API vs SDK" as two parallel product lines. NanoToy (PY32) is the true pure-UART command module: no AI semantics, no cloud, no SDK. It only accepts commands over a serial port to perform fixed actions and playback. Settle the platform and the architecture first, and your selection will not go wrong at the very first step.
TY002: the full Tuya stack, the fast lane overseas
The TY Series is built on the Tuya T5E, and its defining characteristic is "works out of the box". Tuya's years of IoT plumbing mean that TY002 arrives at project kick-off already carrying 60+ languages of TTS/ASR, GDPR / COPPA / CCPA pre-compliance, and global certification and onboarding resources. For small and mid-sized toy factories with thin hardware and software capability, who want to ship a smart toy quickly, or whose customers sit mainly overseas, TY is the lowest-risk choice. Its dual stack is hosted centrally by the Tuya platform, so how far you can open it up is bounded by that platform - what you get in return is a launch measured in months.
LX002: Espressif plus a private cloud, full brand sovereignty
The LX Series is built on the Espressif ESP32S3 and paired with Nablai's own AI cloud platform, the Nablai app, the WeChat Mini Program and the AMS vendor management platform. LX002 is a dual stack too, but the cloud side of it belongs to the brand owner: a standalone account system means the toy factory owns 100% of its users, content and data, and the AMS console brings device management, user operations, content operations and data dashboards into one place. For a toy company moving from "selling hardware" to "selling hardware + content + service", and wanting to keep users inside its own system, LX is the route that gives genuine sovereignty. It also leaves the most room for secondary development and customization.
NT (NanoToy): pure UART command module, a low-cost way to test the market
The NT Series is built on the PY32 and is the lightest of the three lines. It carries no AI semantics, does not connect to the cloud and ships no SDK; it simply takes commands over a UART serial port to drive motion, lighting and fixed voice playback. For customers who are cautious about AI voice interaction, sensitive on cost, and only want to test the smart toy market for now, NT pushes the barrier to entry as low as it goes. It is not on the same axis as a dual-stack module - one answers "can it hold a conversation", the other answers "can it move, light up and make a sound".
How we help customers choose
Selection is not a comparison of specifications, it is a comparison of business structures. We usually open with three questions: where are your customers, domestic or overseas? Do you want a user system of your own? Are you trying to launch fast or to operate for the long term? Overseas first with limited in-house capability → TY. Build your own users and content, want brand sovereignty → LX. Just want to validate a channel cheaply → NT. The three routes map onto three kinds of business, and none of them is a substitute for another. Shenzhen Nablai Intelligent Technology Co., Ltd. puts all three into a single selection matrix so that toy factories of every size can turn an idea into a smart toy that keeps evolving, at their own pace - rather than meeting for the first time, on the production line, the traps that could have been removed well in advance.
About Nablai
Nablai is the AI smart toy module brand of Shenzhen Nablai Intelligent Technology Co., Ltd., and the application area we focus on today is AI smart toys. We supply AI smart toy modules together with the companion software, and our goal is to give toy companies a product-level solution for bringing AI into their toys.
We have built three product families:
- TY Series: built on Tuya T5E, 60+ languages, the fast lane into global markets
- LX Series: built on Espressif ESP32S3, full brand ownership, ships with the Nablai app / WeChat Mini Program / AMS, deep customization
- NT Series: built on PY32, a low-cost way to test the smart toy market
- Website: https://www.nablai.com.cn
- Phone: 15917703502
- Slogan: Driven by gradients, optimized by intelligence. Powered by Nabla, Optimized by AI.
Giving every toy a soul is not a slogan. It means every conversation being a little more natural, every act of perception a little more accurate, every upgrade a little more reliable. What we do is put that capability within easy reach of a toy factory.
A lot of customers ask us whether it is worth it. Our answer is: look at total cost of ownership before you look at unit price. Hardware is a one-off purchase, but software and the cloud run continuously, conversation calls are billed on active connections, and OTA and accounts carry operational overhead. Only when the first year of cloud spend is written into the project budget do the numbers actually add up.
We run three product families to match three kinds of need - going overseas, sovereignty, and testing the water - not to box customers into one model, but to let a toy factory choose at its own pace. Helping customers turn an idea into a smart toy that keeps evolving is what we do, day in and day out.
We lay the architecture out clearly so that toy factories take fewer wrong turns. We have seen too many projects reworked at the mass-production stage because the platform was chosen wrong, compliance was never scheduled up front, or data ownership was never thought through - with both time and money going down with them. Settling the business structure first and discussing how deep the customization should go second is the order we always insist on.
Reliability is not perfection in a single link, it is stability across the whole chain. We treat consistency as a bottom line, because behind every module that leaves the factory is a child looking forward to the first conversation with their toy. That is worth repeating.
Shenzhen Nablai Intelligent Technology Co., Ltd. — AI module specialists for smart toys
📧 contact@nablai.com.cn 🌐 www.nablai.com.cn