This is the competitor companion to the Furniture Virtual Try-On: H5-first + 2D composition + image-to-video plan. That piece covers how to build; this one covers who the players are, where they’re strong, and where they die. Data comes from Grok deep research + an 8-source competitor scan (57 entries, hallucinated entries like the phantom “MeiKeEr” stripped out), then adversarially verified by a 105-agent workflow (25 claims → 4 survived + 21 killed). Prices are point-in-time as of 2026-08 and directional — verify on the official site; items flagged as uncertain during verification are marked.

The real yardstick

Furniture “try-on” isn’t new, but most players do one of two things: build 3D rendering tools for designers, or build “AI restyle the whole room” toys for consumers. The need stuck in the middle — a furniture-store guide closing a sale in 15 minutes, demoing on the spot, and sending the finished video out over WeChat — is a clear gap.

So the yardstick isn’t “whose render looks best.” It’s “who lets a guide close in 15 minutes in front of a customer and ship a shareable video.” By that measure, most celebrity products fail.

The four camps

Master comparison table

The main players by camp, with pricing and the four capabilities (real photo / place specific item / judge size / output video) visible at a glance — then the per-camp breakdown:

✓ = verified, ✗ = rejected in verification, (uncertain) = unresolved; Collov / Room3D prices — check the official site.

Reading the table: only Kujiale / 3vjia (both need a pro subscription + 3D builds), Wayfair / Home Depot (US retail internal tools), and MeltFlex AI (niche) hold all four. And “one photo + one product image → faithful render + size hint + finished video” has no dedicated player in China; a B2B tool for offline furniture-store guides (on-site demo + WeChat delivery) is a clear gap. Competitors fall in three tiers — pro SaaS (Kujiale/3vjia), AI try-on (RoomGPT/Collov/REimagineHome), lightweight tools (Houzz/Planner5D); the mainstream is 2D rendering, real-time AR is just a premium selling point.

Per-camp fatal flaws below.

1. The 3D design-tool camp: Kujiale’s professional moat

Players: Kujiale (the leader, hundreds of thousands of renders a day), 3vjia, Alibaba’s Meiping Meiwu / Tangping.

Approach: build the apartment in 3D first, then drop in 3D furniture models, then render. The most professional-looking of the four camps — but the learning curve is too steep for a guide. Designers train on it; a guide sitting across from a customer can’t get through it in 15 minutes, and video export needs a pro subscription.

Pricing: Kujiale ¥29.8–166/month (the ¥166/month pro tier is the anchor we use for our multi-seat store pricing).

Fatal flaw: it serves designers, not guides. It needs a 3D apartment and 3D furniture models, and merchants don’t have them — modeling one piece costs ¥100–500; a one-person company can’t build a model library. This is the shared death-blow for both the 3D and AR camps.

2. The AR try-on camp: IKEA’s “see it, can’t ship it”

Players: IKEA Place / IKEA Kreativ, Amazon View in Room, Wayfair, Home Depot.

Approach: place furniture live in the phone camera — the strongest on-site feel of the four camps.

Verified hard facts (3/3 votes): IKEA Place does use ARKit (iOS) / ARCore (Android); IKEA Kreativ only places IKEA’s own furniture (locked to its own catalog).

Fatal flaw: ① it can’t produce a shippable finished piece — AR preview is screen-record-only, you can’t get a clean showcase video to send a customer; ② precision gets complaints; ③ Android in China has no Google ARCore, so the experience is broken.

Myths to kill (≥2 votes rejected in adversarial verification — stop citing):

• “ARKit/ARCore makes phone furniture placement work out of the box” is fantasy — real builds must handle scale anchoring / shadows / lighting consistency themselves; room-scan precision is ~5–8cm and environment-dependent (high ceilings, dark surfaces, featureless white walls all drift); precise measurement needs LiDAR, only on iPhone Pro.
• “AR lifts conversion 25–40% / cuts returns 64%” has no primary source — a few aggregator sites copy each other; it’s a marketing number, don’t cite it.

3. The AI photo-makeover camp: RoomGPT et al. — closest, missing the key step

Players: RoomGPT, REimagineHome, Collov, Interior AI, SofaBrain, Decor8 AI, Planner5D, Room3D.

Approach: upload a room photo, AI returns a render. Closest to what we’re building — but almost all do “restyle the whole room,” and can’t faithfully place a specific product, don’t handle size, don’t output video.

Pricing: see the master table above (with verification status). Collov / Room3D prices are verification-rejected or uncertain — check the site before citing; RoomGPT’s $9 is likely a 25–30 credit pack, not a monthly subscription.

Fatal flaw: output is a static image — the customer can’t drag furniture around to re-place it, only regenerate; and AI redraws the furniture distorted (texture, logo, proportions), which is fatal when a guide is selling a specific product.

In one line: RoomGPT et al. solved “produce a render” but not “place that piece + judge size + ship a video.”

4. The general AI-retouch camp: cheap, but distorted

Players: Doubao / Jimeng P-tu, Meitu, Gemini (nano-banana), Qwen-Image-Edit, Flux Kontext.

Approach: general retouching — it can “P a sofa into the living room,” free/cheap.

Verified hard facts: Flux Kontext’s real price is ~$0.04/image (PiAPI Pro tier, not the rumored $0.02); commercial use of FLUX.1 Kontext [dev] needs a separate Black Forest Labs license — PiAPI’s “free commercial use” only holds when you call through its paid platform.

Fatal flaw: furniture is often drawn distorted — fatal when selling a specific product; no size, no video. It’s “retouching,” not a “try-on tool.”

How we cut in: differentiation, already shipped

We cut this gap with lynxfurnish (live, a finished-product service for furniture-store guides). Tech route: H5 + 2D composition + image-to-video, no 3D/AR. Differentials vs each camp:

• vs RoomGPT et al.: ① we support drag-to-rearrange (they’re static images, can’t); ② we output a shareable MP4 (they’re screenshot/screen-record only); ③ we preserve the product image — AI only does light/shadow fusion, the furniture body isn’t redrawn a single pixel (they redraw it distorted).
• vs IKEA Kreativ: not locked to a catalog (a guide places any in-store product), and ships a finished video.
• vs Kujiale: a guide can finish in 15 minutes, no 3D apartment, no pro subscription.
• Size: we don’t claim “precise measurement,” only three buckets — “loose / just-right / won’t fit.” The user gives one reference dimension (wall width / ceiling height) and we compute the footprint ratio, with “for reference, re-measure on-site” labeled throughout. We deliberately dodge single-photo metric depth — physically unreliable, and the consumer products that claim to do it get complained about.

What to copy, what to avoid

Copy: have the user give one reference dimension instead of measuring the room (lowest barrier); brand-watermark the finished piece (the guide feels it’s their store’s official output, and it spreads for you); unlimited regeneration (the customer keeps editing until happy, then pays).

Avoid: locking to your own catalog (IKEA Kreativ only fits IKEA); a pure-AR route (no shippable piece + no ARCore in China); high per-use pricing (RoomGPT $1–2/use doesn’t retain); pure 2D paste without light/shadow fusion (looks fake, won’t close); claiming “AI precise room-measurement” (physically unreliable, gets complaints).

Competitor data here passed adversarial verification (105 agents / 25 claims → 4 survived + 21 killed); full verification log is in the companion piece’s appendix. Prices are point-in-time as of 2026-08 and directional — verify on the official site; Collov / Room3D prices are verification-uncertain, check the site before citing. Companion plan piece: Furniture Virtual Try-On: H5-first + 2D composition + image-to-video plan. Product is live: furnish.lxlynx.com.