Serhii Yaremych

Added two little surface-light tricks to the glass: a specular hotspot, the sharp shine that slides around, and a soft fresnel sheen that lights up the edges at grazing angles. Together they make it feel less like a flat pane and more like something rounded you could squish

My take on the iOS Siri liquid glass, running on Android. All real-time on the GPU, no video, no assets. Lens refraction, colored ribbons, fresnel rim, specular sheen, and a contact shadow through the glass. The shader is very messy, but I hope I manage to clean it up sometime. RenderEffectX

Leaves in depth :D

Leaves in depth :D

A for-fun Android project: one flat photo in, a tilt-reactive 2.5D depth card out. Depth estimation, subject cutout, background inpaint, per-pixel parallax, all on-device, mostly CPU. Kotlin+Jetpack Compose, tested on a 2019 Xiaomi Mi 9. Thread, with the how-it-works. 🧵

Bonus: the depth map read as a relief. The slope at each pixel tells which way that spot faces, that's a normal map. Tilt moves a virtual light: spots facing the light catch a sheen, spots leaning away go dark(Blinn-Phong). Full-frame it read as hammered metal, so it stayed a sandbox.

A for-fun Android project: one flat photo in, a tilt-reactive 2.5D depth card out. Depth estimation, subject cutout, background inpaint, per-pixel parallax, all on-device, mostly CPU. Kotlin+Jetpack Compose, tested on a 2019 Xiaomi Mi 9. Thread, with the how-it-works. 🧵

The payoff. Per frame, the inpainted plate is backward-warped by depth, pivoted at the subject's depth: the subject plane holds still, far pixels drift one way, near ones the other. The cutout stays rigid on top, warping a flat person in 3D reads uncanny; moving only the background reads natural.

A for-fun Android project: one flat photo in, a tilt-reactive 2.5D depth card out. Depth estimation, subject cutout, background inpaint, per-pixel parallax, all on-device, mostly CPU. Kotlin+Jetpack Compose, tested on a 2019 Xiaomi Mi 9. Thread, with the how-it-works. 🧵

Stage 2 — cut out the subject, rebuild the background. BiRefNet gives the mask; a guided filter refines it only in a thin band around the edge, full-frame refining adds halo. The hole is filled with a push-pull pyramid: average known colors down to 1×1, refill on the way up. No model, runs once.

A for-fun Android project: one flat photo in, a tilt-reactive 2.5D depth card out. Depth estimation, subject cutout, background inpaint, per-pixel parallax, all on-device, mostly CPU. Kotlin+Jetpack Compose, tested on a 2019 Xiaomi Mi 9. Thread, with the how-it-works. 🧵

The whole thing, coarse. Two small models run once per photo, depth (how far is every pixel) and the subject mask, then classic image processing: refine the mask edge, cut the subject out, fill the hole behind it. Per frame it's just pixels: the warp, plus a glare and rim light riding the tilt.

RenderEffectX is my follow-on from Imla. Imla captures the Compose tree into a HardwareBuffer but presents it through a SurfaceView. This one keeps that pipeline, drops the surface, and draws the result inline. Same idea as RenderEffect, but down to API 29. #AndroidDev

Three animatable demo scenes, each a fragment shader rebuilt per frame: a music player pushed through a CRT (adapted from a Shadertoy shader), a glitch tearing a real photo with channel split and dropout, and a chromatic ripple that radiates from your touch.

RenderEffectX is my follow-on from Imla. Imla captures the Compose tree into a HardwareBuffer but presents it through a SurfaceView. This one keeps that pipeline, drops the surface, and draws the result inline. Same idea as RenderEffect, but down to API 29. #AndroidDev

The pipeline: record the composable into a HardwareBuffer (HardwareRenderer + ImageReader, no glReadPixels), import it zero-copy as an EGLImage on a GL thread, run a GLSL ES fragment shader, wrap the result as a hardware Bitmap (Bitmap.wrapHardwareBuffer, API 29+), draw it inline.

RenderEffectX is my follow-on from Imla. Imla captures the Compose tree into a HardwareBuffer but presents it through a SurfaceView. This one keeps that pipeline, drops the surface, and draws the result inline. Same idea as RenderEffect, but down to API 29. #AndroidDev

Okay, I finally had time to update #Imla, my experimental GPU backdrop blur for Jetpack Compose. The main change is how it renders: instead of a surface per blurred region, the whole UI is captured once per frame into a HardwareBuffer that every blur layer samples on the GPU. 🧵

The fun part is the output: instead of rendering into the default framebuffer and eglSwapBuffers, I hand the finished HardwareBuffer straight to SurfaceFlinger via a SurfaceControl transaction (API 29+). I rarely see SurfaceControl used this way in apps.

Okay, I finally had time to update #Imla, my experimental GPU backdrop blur for Jetpack Compose. The main change is how it renders: instead of a surface per blurred region, the whole UI is captured once per frame into a HardwareBuffer that every blur layer samples on the GPU. 🧵

What's actually new: - separable Gaussian blur (replacing my old custom Kawase) - progressive blur: sharp top, blurred bottom, via a brush mask - a modifier DSL: ImlaHost + Modifier.effectGroup() / effectLayer { }

Need a bit more work… Android 11, btw. Still fully on the hardware path. #AndroidDev

One more blur pass… right? #AndroidDev #JetpackCompose