Listing Lab
A product copywriting tool that runs entirely in the browser. Built on a small open-source model and performing inference locally through Transformers.js, it does not depend on a backend API and generates product titles, descriptions, bullet points, and search tags for independent sellers.
Timeline (last 90 days)
Listing Lab is a product-copy generation tool that runs entirely in the browser, based on the open-weight SmolLM2-360M-Instruct model, loaded via Transformers.js, and does not depend on any backend inference API.
The Listing Lab author submitted the project in the second week of the Hacktoberfest open-source AI challenge.
Listing Lab's inference in the browser prefers WebGPU acceleration and falls back to running on the CPU via WASM when unsupported.
The model files downloaded on first use by Listing Lab are cached by the browser, so subsequent use requires no repeated downloads.
Listing Lab does not provide a prompt API, nor does it have a prompt database; sellers review and copy the results themselves.
Listing Lab abandoned the output approach that relied on a strict schema and switched to editable plain-text drafts, to avoid silent UI failures caused by the model missing fixed fields.
When Listing Lab downloads the model for the first time, the interface displays the number of bytes downloaded and the percentage.
Listing Lab chose open-weight models, so the product has no pay-per-use inference cost and its marginal cost approaches zero.
The Listing Lab project has currently completed its production build, and the public demo address and code repository will be added after deployment verification is completed.
The Listing Lab author lists "first user test" as the next step before release, wanting to verify whether sellers are willing to accept the first download for the sake of privacy, which platform they need most, and whether they are willing to pay a one-time fee rather than subscribe.
All Facts (10)
The Listing Lab author lists "first user test" as the next step before release, wanting to verify whether sellers are willing to accept the first download for the sake of privacy, which platform they need most, and whether they are willing to pay a one-time fee rather than subscribe.
50%UnverifiedListing Lab is a product-copy generation tool that runs entirely in the browser, based on the open-weight SmolLM2-360M-Instruct model, loaded via Transformers.js, and does not depend on any backend inference API.
50%UnverifiedListing Lab's inference in the browser prefers WebGPU acceleration and falls back to running on the CPU via WASM when unsupported.
50%UnverifiedThe model files downloaded on first use by Listing Lab are cached by the browser, so subsequent use requires no repeated downloads.
50%UnverifiedListing Lab does not provide a prompt API, nor does it have a prompt database; sellers review and copy the results themselves.
50%UnverifiedListing Lab abandoned the output approach that relied on a strict schema and switched to editable plain-text drafts, to avoid silent UI failures caused by the model missing fixed fields.
50%UnverifiedWhen Listing Lab downloads the model for the first time, the interface displays the number of bytes downloaded and the percentage.
50%UnverifiedListing Lab chose open-weight models, so the product has no pay-per-use inference cost and its marginal cost approaches zero.
50%UnverifiedThe Listing Lab project has currently completed its production build, and the public demo address and code repository will be added after deployment verification is completed.
50%UnverifiedThe Listing Lab author submitted the project in the second week of the Hacktoberfest open-source AI challenge.
50%