Skip to content

Releases: ElevenSpectacles/virtual-try-on

v5.0.1

Choose a tag to compare

@github-actions github-actions released this 30 Oct 10:57
56e537b

5.0.1 (2026-10-25)

Bug Fixes

  • lenses cast no colourless contact shadow (grey patch on the cheek) (960708c)
  • lenses cast no colourless contact shadow (grey patch on the cheek) (cf823d6)

Documentation

  • note the nuxi prepare step for the generated agent skill (2bc65e7)

v5.0.0

Choose a tag to compare

@github-actions github-actions released this 20 Oct 17:02
6b12849

5.0.0 (2026-10-20)

⚠ BREAKING CHANGES

  • the frameYaw and templeWidth props and the templeWidthBoost calibration field are removed. Delete the props; old calibration manifests keep working (the field is ignored).

Features

  • capture() snapshot of the composited try-on (32c7328)
  • capture() snapshot of the composited try-on (5f794bc), closes #93
  • face measurements, face shape and catalog-wide fit ranking (e4f5a13)
  • face measurements, face shape and catalog-wide fit ranking (584905b), closes #98
  • freeze() / unfreeze() to compare frames on a held pose (fdb4016)
  • freeze() / unfreeze() to compare frames on a held pose (9b7c2ea), closes #97
  • GLB cache and neighbour prefetch for instant frame switching (0712f63)
  • GLB cache and neighbour prefetch for instant frame switching (fd9b6d3), closes #94
  • lens darkness in physical units, gradient lenses, colour QA tooling (28093fa)
  • lens darkness in physical units, gradient lenses, colour QA tooling (20499d9), closes #85
  • lens reflections from the camera feed and the room's light direction (4598ace)
  • lens reflections from the camera feed and the room's light direction (6ca35aa), closes #84
  • match the webcam with grain, a slight blur and light wrap (57c8411)
  • match the webcam with grain, a slight blur and light wrap (d2ee85a), closes #86
  • multiply contact shadow and add soft contact occlusion (9ac1493)
  • multiply contact shadow and add soft contact occlusion (974c669), closes #88
  • one-command model prep, model lint and calibration validation (a90c52a)
  • one-command model prep, model lint and calibration validation (2a3d5a4), closes #103
  • open temple arms at the hinge to meet the tracked head (63231a1)
  • open temple arms at the hinge to meet the tracked head (383bac1), closes #83
  • opt-in tinted lens shadow (90b7d7c)
  • opt-in tinted lens shadow (6a02658), closes #88
  • optional card calibration for a ±1 mm PD (35e6f74)
  • optional card calibration for a ±1 mm PD (3fb7b4f), closes #96
  • optional ear bend of the temple tips behind the ear (45026eb)
  • optional ear bend of the temple tips behind the ear (a5cb4f5), closes #83
  • pad-contact nose seat (opt-in) and a pantoscopic tilt audit (1f739bc)
  • pad-contact nose seat (opt-in) and a pantoscopic tilt audit (db9253e), closes #87
  • remove frameYaw, templeWidth and templeWidthBoost (cd1369d), closes #101
  • size the frame true to size from the iris-measured PD (8d31c3f)
  • size the frame true to size from the iris-measured PD (962b7d4), closes #92

Bug Fixes

  • contact-occlusion spheres follow the hinged temple arms and the ear bend (5a45def)
  • contact-occlusion spheres follow the hinged temple arms and the ear bend (7a1efed)
  • eight cross-feature bugs found by auditing the Roadmap 2 merge (d63e44e)
  • eight cross-feature bugs found by auditing the Roadmap 2 merge (6842ac1)
  • gate the contact-shadow receiver on the prop again (6710bee)
  • hold the frame through short tracking dropouts; turnBack and tooDark hints (3d58b3f)
  • hold the frame through short tracking dropouts; turnBack and tooDark hints (0cfa7e7), closes #99

Performance

  • time-to-first-frame budget: measure, parallelise, preconnect, self-host (eb417fe)
  • time-to-first-frame budget: measure, parallelise, preconnect, self-host (ab1fd99), closes #95

Documentation

  • agents: internal scene/experience composables and the 400-line SFC rule (4cb2e83)
  • drop the company name from the license note (334f1ee)
  • frame-removal research spike (41006f9)
  • frame-removal research spike (e363726), closes #100
  • public demo on GitHub Pages and README overhaul (2561e6e)
  • public demo on GitHub Pages and README overhaul (512c6fb), closes #104
  • record the landed roadmap-2 PRs (55fe695)
  • record the landed roadmap-2 PRs ([eedcd6c](eedcd6cbccff99f737c...
Read more

v4.8.1

Choose a tag to compare

@github-actions github-actions released this 14 Oct 16:40
ae05ec1

4.8.1 (2026-10-14)

Documentation

  • import ref in the usage example (980c62d)
  • import ref in the usage example (725c0b0)
  • use camelCase prop names in self-hosting section (0c3a2f1)
  • use camelCase prop names in self-hosting section (08b6741)

v4.8.0

Choose a tag to compare

@github-actions github-actions released this 13 Oct 13:02
0ed369c

4.8.0 (2026-10-13)

Features

  • contact shadow from a shadow map of the frame (#6) (d59fe50)
  • contact shadow from a shadow map of the frame (#6) (a69146e)
  • fit verdict, frame width vs the tracked face's width (#12) (4046a25)
  • fit verdict, frame width vs the tracked face's width (#12) (ff72909)
  • opt-in hair occlusion of the temple arms (#9) (a9425b9)
  • opt-in hair occlusion of the temple arms (#9) (b4d42c5)
  • physically-based lens, Fresnel opacity and AR-coating sheen (#10) (0e60039)
  • physically-based lens, Fresnel opacity and AR-coating sheen (#10) (5bbfc24)

Bug Fixes

  • import useTryOnFrameLoop in the experience component (caceb3b)
  • keep rotation smoothing on and a still head still under latency prediction (647332b), closes #15
  • seat the frame on rigid bridge landmarks, not the iris (#11) (4889694)
  • seat the frame on rigid bridge landmarks, not the iris (#11) (c23d9d4)

Performance

  • sample room lighting in the worker, one frame in four (#5) (7de3122)

Documentation

  • README for hair occlusion (#9) (0cd2601)
  • README for the fit verdict, physical lens, contact shadow and lighting state (c7486f1)
  • README for the fit verdict, physical lens, contact shadow and lighting state (62ccf6a)
  • README note on fetching real footage (e64a0ac)
  • research spike on a web path to TrueDepth / metric depth (#19) (d563bd7)
  • research spike on a web path to TrueDepth / metric depth (#19) (a8e74c9)
  • run trim-temple-tips once per model (612d8a8)

Refactoring

  • drop the never-built lensRefraction tier and unused noseBridgeOffset (f056ba6)
  • keep both workers' message types in types/ (efb6a89)
  • move fit-verdict and room-lighting glue into composables (54a2c43)
  • move frame placement into useTryOnPlacement (fbd4263)
  • one placement pipeline for the component and the playground (43ecc80)
  • run the playground prototype on useTryOnPlacement (d80a23f)

v4.7.1

Choose a tag to compare

@github-actions github-actions released this 11 Oct 18:52
54084a2

4.7.1 (2026-10-10)

Documentation

  • add Graphify rules to AGENTS.md (278f8eb)
  • add Graphify rules to AGENTS.md (d7011d0)

v4.7.0

Choose a tag to compare

@github-actions github-actions released this 09 Oct 18:20
bc6d221

4.7.0 (2026-10-09)

Features

  • move the contact shadow down onto the cheeks (505ca93)
  • move the contact shadow down onto the cheeks (#6) (aa71a36)

Bug Fixes

  • centre the contact shadow on the bridge contact, not the head (20307f0)

v4.6.0

Choose a tag to compare

@github-actions github-actions released this 09 Oct 10:52
2a466aa

4.6.0 (2026-10-09)

Features

  • device capability tiering with frame-time step-down (ef666af)
  • fixed-pose frame for a deterministic contact-shadow check (#6) (d25b594)
  • improve face tracking (c8962e7)
  • light the frame from the room and add a contact shadow (ea4fac2)
  • opt-in session metrics event (#17) (1af61b3)

Bug Fixes

  • apply the tier prop reactively (a38eca9)
  • apply the tier prop reactively, not only at setup (8db4e28)

Refactoring

  • schedule detection through a camera-frame helper (9eacbd9)

v4.5.0

Choose a tag to compare

@github-actions github-actions released this 08 Oct 19:27
b16da1e

4.5.0 (2026-10-08)

Features

  • enable the face-mesh occluder by default, configurable globally (2f6a1c8)
  • enable the face-mesh occluder by default, configurable globally (e80c8b6)
  • hold frame scale on MediaPipe blink blendshapes; bump MediaPipe to 1.1.0 (769878d)
  • hold frame scale on MediaPipe blink blendshapes; bump MediaPipe to 1.1.0 (6293095), closes #7

v4.4.0

Choose a tag to compare

@github-actions github-actions released this 08 Oct 15:10
fbb4648

4.4.0 (2026-10-08)

Features

  • render the frame only on a tracked face and drop the VueUse peer (127c6ef)
  • render the frame only on a tracked face and drop the VueUse peer (d29f869)

Bug Fixes

  • deps: update nuxt kit, TresJS and dev tooling to latest minors (e2ec4c0)
  • deps: update nuxt kit, TresJS and dev tooling to latest minors (c2be24f)

v4.2.0

Choose a tag to compare

@github-actions github-actions released this 08 Oct 12:46
e914719

4.2.0 (2026-10-08)

Features

  • ship TresJS and three with the module (b30f2f2)
  • ship TresJS and three with the module (6d07391)

Bug Fixes

  • stand the frame off the face so it tracks turned heads (f1f3671)
  • stand the frame off the face so it tracks turned heads (46b1a82)
  • stop the occluder cutting off the near temple on turned heads (0c55ac4)
  • stop the occluder cutting off the near temple on turned heads (6ec70ce)