Drape virtual try-on
Nailthy Tang's Drape experiment wires live speech, wardrobe metadata, and a realtime try-on feed.
by @nailthy62
Overview
Nailthy Tang's Drape experiment streams your voice, tracks what you are wearing, and lets Jev pick the next garment from a finite closet list so the preview updates live on camera. She cites about $0.0011 per decision and about 620 ms per swap on X. Product home is weardrape.app; the madewithjev build page documents the loop.
Problem: Fashion try-on demos choke when every outfit change needs a slow generative rerender instead of a fast discrete closet pick.
Built for: Creators experimenting with realtime wardrobe UX for Drape and similar virtual try-on products.
First indexed on Jev Directory: 2026-09-19
Creator and team
- Name
- Nailthy Tang
- Organization
- Drape
- Handle
- @nailthy62
How Jev is used
- Role in the product flow
- Realtime wardrobe Choice from transcript plus current outfit state
- Primitives
- ChoiceScore
- State in
- Live speech transcript, detected or labeled current outfit pieces, and enumerated closet SKUs the renderer can swap.
- Decision out
- Next garment Choice (and supporting relevance scores) consumed by the try-on pipeline without free-form styling prose.
- User talks through preferences while the camera feed runs
- ASR transcript and outfit tags land in structured state
- Jev Choice selects the next closet item that matches the request
- Renderer swaps the outfit on the live preview
This is a closet-picker loop, not open-ended image generation every frame. Speech becomes text; the current look is already structured; Jev only answers which SKU comes next from a list the renderer understands. That keeps latency in the sub-second band Nailthy quoted on X. The directory embeds her launch video via remote twimg with referrerPolicy no-referrer like other showcase cards. Treat the experiment as a product pattern for Drape rather than a standalone open repo in this listing.
Sourced performance claims
- Author post cites about $0.0011 per decision and about 620 ms per decision on the public X clip.Source: x.com/nailthy62/status/2101388186916454439
Features and stack
Features
- Voice-driven outfit changes on a live feed
- Finite closet Choice instead of unbounded generative edits
- Author cost and latency figures on the X post
- Linked from madewithjev.com/builds/drape-virtual-try-on
Stack
- TypeSafe Jev
- Speech to text
- Realtime try-on renderer
- Drape product stack
Pricing: Experiment metrics from the author post; production Drape pricing lives on weardrape.app.
Demo
Showcase clip with the same lightbox player as the homepage. Click to play.
Drape realtime virtual try-on
You narrate the outfit like a podcast host; Jev hears the transcript, checks what you are already wearing, and yeets the next look onto camera before your coffee cools.
Links
FAQ
- Is there a public GitHub repo for this clip?
- This listing tracks the Drape experiment and X proof. Follow weardrape.app for the product; no separate repo slug is claimed here.
- Does Jev generate new clothing pixels?
- The described loop picks among existing closet items the renderer can swap, keeping decisions typed and fast.
- Where do the cost numbers come from?
- They are author-reported on the X post embedded in the showcase. Re-measure on your own stack before quoting in decks.
Related learn guides
Original Jev guidance that pairs with this product pattern.
- Jev use cases
The patterns builders actually search for: moderation, routing, triage, RAG verify, and agent gates.
- System One model
The model family behind Jev: parallel typed questions, one forward pass, probabilities you can threshold.
Related products
Hand-picked neighbors with rich profiles or overlapping tags.