feat: add browser GPU render mode (#571)

## Problem

HyperFrames already had `--gpu`, but that flag only controlled FFmpeg hardware encoding. The browser capture path still forced Chrome/WebGL through SwiftShader software GL via `--use-angle=swiftshader`, so WebGL-heavy local renders could leave the biggest bottleneck on the CPU path.

That made the existing flag naming easy to misread: `--gpu` sounded like it accelerated the whole render, but it did not change the browser frame-capture backend.

## What this fixes

- Enables host browser GPU acceleration automatically for local CLI renders.
- Adds `--no-browser-gpu` as the local opt-out for software Chrome/WebGL capture.
- Keeps `--browser-gpu` as an explicit local browser-GPU request.
- Adds `browserGpuMode: "software" | "hardware"` to engine config, with `PRODUCER_BROWSER_GPU_MODE` env support for lower-level producer users.
- Keeps Docker browser capture on the deterministic software path.
- Maps hardware browser GPU mode to platform-native Chrome backends:
  - macOS: Metal-backed ANGLE
  - Windows: D3D11-backed ANGLE
  - Linux: EGL
- Blocks explicit `--browser-gpu --docker` with a clear error because Docker browser GPU passthrough is not cross-platform.
- Clarifies docs so `--gpu` means FFmpeg encoder GPU and browser GPU means Chrome/WebGL capture GPU.
- Keeps encoder backend selection auto-detected from FFmpeg capabilities:
  - NVIDIA: NVENC
  - macOS: VideoToolbox
  - Linux: VAAPI
  - Intel: QSV

## Why two flags

There are two separate GPU surfaces in the render pipeline:

1. Browser GPU controls Chrome frame capture.
   - Affects WebGL, canvas, CSS rendering, compositing, and screenshot capture inside the browser.
   - This is enabled automatically for local CLI renders.
   - Use `--no-browser-gpu` when you want the software browser baseline.

2. `--gpu` controls FFmpeg video encoding.
   - Affects the final encode step after frames have already been captured.
   - The concrete encoder is auto-detected from the host FFmpeg build and hardware.
   - It can be faster for some machines/codecs, but it is not equivalent to browser rendering acceleration.

The controls stay independent because users may want:

- `hyperframes render` for the fast local default with browser GPU capture.
- `hyperframes render --no-browser-gpu` for the software-browser local baseline.
- `hyperframes render --gpu` for browser GPU capture plus hardware FFmpeg encoding.
- `hyperframes render --no-browser-gpu --gpu` for software browser capture plus hardware FFmpeg encoding.
- `hyperframes render --docker` for deterministic browser capture.

## Why `--gpu` does not imply browser GPU

Keeping `--gpu` scoped to FFmpeg encoding avoids a semantic break and keeps the risk profile explicit:

- `--gpu` already means encoder acceleration. Expanding it to also change Chrome capture would silently alter behavior for users who only wanted hardware encoding.
- Browser GPU and encoder GPU have different portability. Encoder GPU can work in Docker when the host exposes the right devices; browser GPU passthrough is not cross-platform, so this PR intentionally blocks explicit `--browser-gpu --docker`.
- The Apple presentation benchmark shows why the controls should stay separate: browser GPU capture was the useful improvement, while macOS VideoToolbox via `--gpu` was slower and produced larger output for this `standard` H.264 run.

If HyperFrames later wants a single umbrella acceleration control, it should be explicit, for example `--acceleration browser|encoder|all` or `--gpu=browser|encoder|all`, rather than changing the meaning of the existing boolean `--gpu`.

## Root cause

`buildChromeArgs()` always injected `--use-gl=angle --use-angle=swiftshader`. `disableGpu` only appended `--disable-gpu`; it did not provide a hardware-GPU mode. That made the public `--gpu` flag look broader than it was, because render capture stayed software-backed even when encoder GPU was requested.

## Verification

### Local checks

- `bun install`
- `bun run build:hyperframes-runtime`
- `bun run --filter @hyperframes/engine test src/config.test.ts src/services/browserManager.test.ts`
- `bun run --filter @hyperframes/cli test src/utils/dockerRunArgs.test.ts src/commands/render.test.ts`
- `bun run --filter @hyperframes/cli typecheck`
- `bun run --filter @hyperframes/engine typecheck`
- `bun run --filter @hyperframes/producer typecheck`
- `cd packages/producer && bunx vitest run src/services/renderOrchestrator.test.ts`
- `bunx oxlint packages/cli/src/commands/render.ts packages/cli/src/commands/render.test.ts packages/cli/src/utils/dockerRunArgs.ts packages/cli/src/utils/dockerRunArgs.test.ts packages/engine/src/config.ts packages/engine/src/config.test.ts packages/engine/src/services/browserManager.ts packages/engine/src/services/browserManager.test.ts packages/producer/src/services/renderOrchestrator.test.ts`
- `bunx oxfmt --check ...` on changed source/docs files
- `git diff --check`
- `bun packages/cli/src/cli.ts render --help | rg -n "browser-gpu|no-browser-gpu|GPU"`
- `bun packages/cli/src/cli.ts render packages/producer/tests/css-spinner-render-compat/src --output /tmp/hf-auto-browser-gpu-smoke.mp4 --workers 1 --quality draft --fps 24 --strict`
  - Render plan prints `GPU: browser GPU (auto)`.
- `bun packages/cli/src/cli.ts render packages/producer/tests/css-spinner-render-compat/src --no-browser-gpu --output /tmp/hf-software-browser-gpu-smoke.mp4 --workers 1 --quality draft --fps 24 --strict`
  - Render plan does not print browser GPU.
- `bun packages/cli/src/cli.ts render packages/producer/tests/css-spinner-render-compat/src --docker --browser-gpu --output /tmp/should-not-render.mp4`
  - Exits 1 with `Browser GPU is local-only`.
- `buildDockerRunArgs()` regression coverage asserts Docker container args include `--no-browser-gpu`, preventing nested container renders from re-enabling browser GPU through the local CLI default.
- `resolveBrowserGpuForCli()` regression coverage asserts `PRODUCER_BROWSER_GPU_MODE=software` opts out when no CLI browser-GPU flag is supplied, while explicit `--browser-gpu` / `--no-browser-gpu` still win.
- `ffmpeg -v error -i /tmp/hf-auto-browser-gpu-smoke.mp4 -f null -`
- `ffmpeg -v error -i /tmp/hf-software-browser-gpu-smoke.mp4 -f null -`
- `ffprobe -v error -show_entries format=duration:stream=codec_name,width,height,r_frame_rate -of json /tmp/hf-browser-gpu-smoke.mp4` -> H.264, 1920x1080, 24fps, 5.0s

### Apple presentation benchmark

Rendered `/Users/miguel07code/Downloads/apple-presentation.zip` as supplied after extracting to `/tmp/hf-apple-profile/apple-presentation`.

Fixed settings:

- 1920x1080
- 30fps
- `standard` quality
- 4240 frames
- 141.32s duration
- 8-worker cap; render auto-calibration used 6 capture workers
- macOS host detected FFmpeg GPU encoder: `videotoolbox`

| Mode | Equivalent flags after this PR | Wall time | vs software-browser baseline | Speed | Capture | Encode | Output |
| --- | --- | ---: | ---: | ---: | ---: | ---: | ---: |
| Software browser + CPU encode | `--no-browser-gpu` | 120.77s | baseline | 1.17x | 97.87s | 10.04s | 8.38MB |
| Browser GPU + CPU encode | default local render | 70.10s | 42.0% faster | 2.02x | 50.72s | 9.91s | 8.39MB |
| Software browser + encoder GPU | `--no-browser-gpu --gpu` | 133.16s | 10.3% slower | 1.06x | 103.58s | 18.31s | 25.43MB |
| Browser GPU + encoder GPU | `--gpu` | 74.12s | 38.6% faster | 1.91x | 46.69s | 17.93s | 25.45MB |

Result: browser GPU capture is the meaningful improvement for this WebGL/browser-capture-heavy presentation. VideoToolbox encoding was slower and produced larger files for this current `standard` H.264 path, so `--gpu` should stay separate and opt-in.

Why `--gpu` plus browser GPU was slower than browser GPU alone: the combined run captured about 4.0s faster than browser GPU alone, but VideoToolbox encoding was about 8.0s slower than CPU x264 encoding, so the encode loss outweighed the capture gain.

### VideoToolbox flag check

I also isolated the encode stage against the already-captured Apple frames to check whether macOS GPU encoding only needed special flags.

`ffmpeg -h encoder=h264_videotoolbox` does not expose a CRF/CQ-style quality option like x264. It exposes bitrate-oriented and VideoToolbox-specific options such as `-b:v`, `-realtime`, `-profile`, `-coder`, `-prio_speed`, `-power_efficient`, and `-allow_sw`. That means our current `-q:v` mapping is not equivalent to x264 CRF and can produce very different bitrate/size behavior.

Measured full-frame encode variants on this host:

| VideoToolbox variant | Encode wall time | Output size | Bitrate |
| --- | ---: | ---: | ---: |
| Current `-q:v 64 -allow_sw 1` | 18.76s | 25.31MB | 1.43 Mbps |
| Current without `-allow_sw 1` | 18.21s | 25.31MB | 1.43 Mbps |
| `-b:v 500k -maxrate 750k -bufsize 1000k -profile high -coder cabac -realtime 1 -prio_speed 1 -power_efficient 0` | 20.58s | 7.42MB | 0.42 Mbps |
| Same with `-b:v 1500k` | 20.84s | 16.70MB | 0.95 Mbps |
| `-b:v 500k -profile baseline -coder cavlc -realtime 1 -prio_speed 1 -power_efficient 0` | 18.11s | 8.94MB | 0.51 Mbps |

Conclusion: VideoToolbox can be made size/bitrate-predictable with explicit `--video-bitrate`, but the tested speed-oriented flags did not make it faster than CPU x264 wall time for this render. That reinforces keeping `--gpu` encoder acceleration explicit and separate from browser GPU capture.

Artifacts from the local benchmark:

- `/tmp/hf-apple-profile/results/cpu.mp4`
- `/tmp/hf-apple-profile/results/browser-gpu.mp4`
- `/tmp/hf-apple-profile/results/encoder-gpu.mp4`
- `/tmp/hf-apple-profile/results/full-gpu.mp4`
- `/tmp/hf-apple-profile/results/summary.json`

All four benchmark MP4s completed `ffprobe` and full `ffmpeg -f null` decode checks.

### Pixel comparison

Compared decoded MP4 output between software-browser and browser-GPU renders:

- Apple presentation:
  - 4240 frames compared
  - 636 exact matching decoded frame hashes
  - 3604 different decoded frame hashes
  - Average PSNR: 57.79 dB
- `css-spinner-render-compat` clean fixture:
  - 120 frames compared
  - 0 exact matching decoded frame hashes
  - Average PSNR: 61.57 dB

Interpretation: browser GPU output is not strict hash/pixel-identical to the software-browser path after lossy H.264 encode, but the measured deltas are visually tiny. Above 50 dB PSNR is typically visually indistinguishable for normal video review. Use `--no-browser-gpu` or Docker when strict cross-run/cross-machine reproducibility matters more than local speed.

### Browser verification

- Started HyperFrames Studio preview for `packages/producer/tests/css-spinner-render-compat/src`.
- Used `agent-browser` to open `http://localhost:5191#project/src` and verify the composition loaded in Studio.
- Screenshots:
  - `/tmp/hf-gpu-browser-proof/preview-loaded.png`
  - `/tmp/hf-gpu-browser-proof/preview-playing.png`
  - `/tmp/hf-gpu-browser-proof/preview-frame-60.png`
- Agent-browser recordings:
  - `/tmp/hf-gpu-browser-proof/preview-playback.webm`
  - `/tmp/hf-gpu-browser-proof/preview-seek.webm`

## Notes

- Browser GPU is enabled automatically for local CLI renders and disabled in Docker.
- `--no-browser-gpu` is the opt-out for software Chrome/WebGL capture.
- `--gpu` remains encoder-only and opt-in.
- The Apple presentation zip has existing lint errors around unmanaged nested videos and imperative media `play()` calls. The benchmark still compares the same supplied source across modes, but it should not be treated as a clean deterministic-composition fixture.

docs/guides/rendering.mdx

@@ -65,7 +65,8 @@ Render your Hyperframes [compositions](/concepts/compositions) to MP4, MOV, or W
 
     **Pros:**
     - Fast startup, no container overhead
-    - Uses your system GPU for hardware-accelerated encoding (with `--gpu`)
+    - Can use your system GPU for Chrome/WebGL capture by default
+    - Can use your system GPU for hardware-accelerated encoding (with `--gpu`)
     - Best for iterative development
 
     **Cons:**
@@ -90,7 +91,8 @@ Render your Hyperframes [compositions](/concepts/compositions) to MP4, MOV, or W
 
     **Cons:**
     - Slower startup due to container initialization
-    - No GPU acceleration inside the container
+    - Browser capture stays on the deterministic software-GL path
+    - GPU encoding requires Docker host GPU passthrough and is not cross-platform on Docker Desktop
 
     <Note>
       Docker mode uses `chrome-headless-shell` with [BeginFrame](/concepts/determinism#how-it-works) control for frame-perfect, deterministic capture.
@@ -121,7 +123,8 @@ Render your Hyperframes [compositions](/concepts/compositions) to MP4, MOV, or W
 | `--video-bitrate` | e.g. `10M`, `5000k` | — | Target bitrate encoding. Cannot combine with `--crf` |
 | `--workers` | 1-8 or `auto` | auto | Parallel render workers (see [Workers](#workers) below) |
 | `--max-concurrent-renders` | 1-10 | 2 | Max simultaneous renders via the producer server (see [Concurrent Renders](#concurrent-renders) below) |
-| `--gpu` | — | off | GPU encoding (NVENC, VideoToolbox, VAAPI) |
+| `--gpu` | — | off | GPU encoding (NVENC, VideoToolbox, VAAPI, QSV) |
+| `--browser-gpu` / `--no-browser-gpu` | — | on locally, off in Docker | Use or opt out of host GPU acceleration for local Chrome/WebGL capture |
 | `--hdr` | — | off | Force HDR output even if no HDR sources are detected (MP4 only). See [HDR Rendering](/guides/hdr) |
 | `--sdr` | — | off | Force SDR output even if HDR sources are detected |
 | `--docker` | — | off | Use Docker for [deterministic rendering](/concepts/determinism) |
@@ -149,6 +152,26 @@ npx hyperframes render --video-bitrate 10M --output controlled.mp4
 
 **Tip**: The default `standard` preset (CRF 18) is visually lossless at 1080p — most people cannot distinguish it from the source. Use `--quality draft` for faster iteration, or `--quality high` / `--crf 10` when file size is no concern.
 
+## GPU Acceleration
+
+Hyperframes has two separate GPU acceleration surfaces:
+
+- `--gpu` uses a hardware video encoder in FFmpeg when one is available. Supported backends include VideoToolbox on macOS, NVENC on NVIDIA systems, VAAPI on Linux, and Intel QSV on supported Windows/Linux hosts.
+- Browser GPU uses the host GPU for local Chrome/WebGL capture. It is enabled automatically for local renders and disabled in Docker. Use `--no-browser-gpu` to opt out.
+
+```bash Terminal
+# Add hardware FFmpeg encoding to the default local browser-GPU render
+npx hyperframes render --gpu --output encoded-fast.mp4
+
+# Opt out of hardware Chrome/WebGL capture
+npx hyperframes render --no-browser-gpu --output software-browser.mp4
+
+# Use browser GPU plus hardware FFmpeg encoding
+npx hyperframes render --gpu --output gpu.mp4
+```
+
+Browser GPU capture is local-mode only. It maps to platform-native Chrome GPU backends: Metal on macOS, D3D11 on Windows, and EGL on Linux. Use `--no-browser-gpu` or Docker mode when exact cross-machine reproducibility matters more than local render speed.
+
 ## Workers
 
 Each render worker launches a **separate Chrome browser process** to capture frames in parallel. More workers can speed up rendering, but each one consumes ~256 MB of RAM and significant CPU.

docs/guides/troubleshooting.mdx

@@ -128,9 +128,10 @@ If your issue is about a specific coding mistake (animations not working, video
 
     1. Use `--quality draft` during development for faster encoding
     2. Run `npx hyperframes benchmark` to find the optimal worker count for your system
-    3. Use `--gpu` for hardware-accelerated encoding (local mode only)
+    3. Local Chrome/WebGL GPU capture is enabled automatically; compare with `--no-browser-gpu` if troubleshooting
-    4. Reduce `--fps` to 24 if 30fps is not needed
+    4. Use `--gpu` for hardware-accelerated encoding (local mode only)
-    5. Check that your composition does not have unnecessary elements or overly complex animations
+    5. Reduce `--fps` to 24 if 30fps is not needed
+    6. Check that your composition does not have unnecessary elements or overly complex animations
 
     See [Rendering: Options](/guides/rendering#options) for all available flags.
   </Accordion>

docs/packages/cli.mdx

@@ -530,6 +530,12 @@ This is suppressed in CI environments, non-TTY shells, and when `HYPERFRAMES_NO_
 
     # With options
     npx hyperframes render --output output.mp4 --fps 60 --quality high
+
+    # Opt out of local browser GPU capture
+    npx hyperframes render --no-browser-gpu --output cpu-browser.mp4
+
+    # Add hardware FFmpeg encoding
+    npx hyperframes render --gpu --output gpu.mp4
     ```
 
     | Flag | Values | Default | Description |
@@ -543,7 +549,8 @@ This is suppressed in CI environments, non-TTY shells, and when `HYPERFRAMES_NO_
     | `--hdr` | — | off | Force HDR output even if no HDR sources are detected. MP4 only. See [HDR Rendering](/guides/hdr) |
     | `--sdr` | — | off | Force SDR output even if HDR sources are detected |
     | `--workers` | 1-8 | 4 | Parallel render workers |
-    | `--gpu` | — | off | GPU encoding (NVENC, VideoToolbox, VAAPI) |
+    | `--gpu` | — | off | GPU encoding (NVENC, VideoToolbox, VAAPI, QSV) |
+    | `--browser-gpu` / `--no-browser-gpu` | — | on locally, off in Docker | Use or opt out of host GPU acceleration for local Chrome/WebGL capture |
     | `--docker` | — | off | Use Docker for [deterministic rendering](/concepts/determinism) |
     | `--quiet` | — | off | Suppress verbose output |
 

docs/packages/engine.mdx

@@ -175,7 +175,7 @@ await applyFaststart(inputPath, outputPath);
 
 // Detect GPU encoding support
 const gpu = await detectGpuEncoder();
-// gpu: "nvenc" | "videotoolbox" | "vaapi" | null
+// gpu: "nvenc" | "videotoolbox" | "vaapi" | "qsv" | null
 ```
 
 #### WebM with VP9 Alpha

docs/packages/producer.mdx

@@ -213,18 +213,31 @@ npx hyperframes render --docker --output output.mp4
 
 The producer supports hardware-accelerated encoding for faster renders:
 
-| Platform | Encoder | Flag |
+| Platform | Encoder | Selection |
-|----------|---------|------|
+|----------|---------|-----------|
 | NVIDIA | NVENC | Auto-detected |
 | macOS | VideoToolbox | Auto-detected |
 | Linux | VAAPI | Auto-detected |
+| Intel | QSV | Auto-detected |
 
-GPU encoding is automatically used when available. To check your system's capabilities:
+When GPU encoding is enabled, Hyperframes detects the available FFmpeg hardware encoder automatically. To check your system's capabilities:
 
 ```bash
 npx hyperframes doctor
 ```
 
+The CLI enables local Chrome/WebGL GPU capture automatically and supports `--no-browser-gpu` as an opt-out. When using the producer API directly, pass an engine config override:
+
+```typescript
+import { resolveConfig } from '@hyperframes/producer';
+
+const job = createRenderJob({
+  fps: 30,
+  quality: 'standard',
+  producerConfig: resolveConfig({ browserGpuMode: 'hardware' }),
+});
+```
+
 ## Additional Exports
 
 The producer also re-exports key engine functionality for convenience:

docs/quickstart.mdx

@@ -237,6 +237,6 @@ The agent handles scaffolding, animation, and rendering. See the [prompting guid
     Start from built-in examples like Warm Grain and Swiss Grid
   </Card>
   <Card title="Rendering" icon="film" href="/guides/rendering">
-    Explore render options: quality presets, Docker mode, and GPU encoding
+    Explore render options: quality presets, Docker mode, and GPU acceleration
   </Card>
 </CardGroup>

packages/cli/src/commands/render.test.ts

@@ -0,0 +1,102 @@
+import { afterEach, beforeEach, describe, expect, it, vi } from "vitest";
+
+const producerState = vi.hoisted(() => ({
+  createdJobs: [] as Array<Record<string, unknown>>,
+  resolveConfigCalls: [] as Array<Record<string, unknown>>,
+}));
+
+vi.mock("../utils/producer.js", () => ({
+  loadProducer: vi.fn(async () => ({
+    resolveConfig: vi.fn((overrides: Record<string, unknown>) => {
+      producerState.resolveConfigCalls.push(overrides);
+      return { ...overrides, resolved: true };
+    }),
+    createRenderJob: vi.fn((config: Record<string, unknown>) => {
+      producerState.createdJobs.push(config);
+      return { config, progress: 100 };
+    }),
+    executeRenderJob: vi.fn(async () => undefined),
+  })),
+}));
+
+vi.mock("../telemetry/events.js", () => ({
+  trackRenderComplete: vi.fn(),
+  trackRenderError: vi.fn(),
+}));
+
+describe("renderLocal browser GPU config", () => {
+  const savedEnv = new Map<string, string | undefined>();
+
+  function setEnv(key: string, value: string) {
+    savedEnv.set(key, process.env[key]);
+    process.env[key] = value;
+  }
+
+  beforeEach(() => {
+    producerState.createdJobs = [];
+    producerState.resolveConfigCalls = [];
+    savedEnv.clear();
+  });
+
+  afterEach(() => {
+    for (const [key, value] of savedEnv) {
+      if (value === undefined) {
+        delete process.env[key];
+      } else {
+        process.env[key] = value;
+      }
+    }
+    vi.clearAllMocks();
+  });
+
+  it("passes an explicit software override for --no-browser-gpu even when env requests hardware", async () => {
+    setEnv("PRODUCER_BROWSER_GPU_MODE", "hardware");
+
+    const { renderLocal } = await import("./render.js");
+    await renderLocal("/tmp/project", "/tmp/out.mp4", {
+      fps: 30,
+      quality: "standard",
+      format: "mp4",
+      gpu: false,
+      browserGpu: false,
+      hdrMode: "auto",
+      quiet: true,
+    });
+
+    expect(producerState.resolveConfigCalls).toContainEqual({ browserGpuMode: "software" });
+    expect(producerState.createdJobs[0]?.producerConfig).toMatchObject({
+      browserGpuMode: "software",
+      resolved: true,
+    });
+  });
+
+  it("passes an explicit hardware override for default local browser GPU", async () => {
+    const { renderLocal } = await import("./render.js");
+    await renderLocal("/tmp/project", "/tmp/out.mp4", {
+      fps: 30,
+      quality: "standard",
+      format: "mp4",
+      gpu: false,
+      browserGpu: true,
+      hdrMode: "auto",
+      quiet: true,
+    });
+
+    expect(producerState.resolveConfigCalls).toContainEqual({ browserGpuMode: "hardware" });
+    expect(producerState.createdJobs[0]?.producerConfig).toMatchObject({
+      browserGpuMode: "hardware",
+      resolved: true,
+    });
+  });
+
+  it("resolves browser GPU from CLI flags, Docker mode, and env fallback", async () => {
+    const { resolveBrowserGpuForCli } = await import("./render.js");
+
+    expect(resolveBrowserGpuForCli(false, undefined, undefined)).toBe(true);
+    expect(resolveBrowserGpuForCli(false, undefined, "hardware")).toBe(true);
+    expect(resolveBrowserGpuForCli(false, undefined, "software")).toBe(false);
+    expect(resolveBrowserGpuForCli(false, true, "software")).toBe(true);
+    expect(resolveBrowserGpuForCli(false, false, "hardware")).toBe(false);
+    expect(resolveBrowserGpuForCli(true, undefined, "hardware")).toBe(false);
+  });
+});

packages/cli/src/commands/render.ts

@@ -9,6 +9,7 @@ export const examples: Example[] = [
   ["High quality at 60fps", "hyperframes render --fps 60 --quality high --output hd.mp4"],
   ["Deterministic render via Docker", "hyperframes render --docker --output deterministic.mp4"],
   ["Parallel rendering with 6 workers", "hyperframes render --workers 6 --output fast.mp4"],
+  ["Opt out of browser GPU render", "hyperframes render --no-browser-gpu --output cpu.mp4"],
   ["HDR output (auto-detected)", "hyperframes render --output hdr-output.mp4"],
 ];
 import { cpus, freemem, tmpdir } from "node:os";
@@ -99,6 +100,11 @@ export default defineCommand({
       description: "Target video bitrate such as 10M. Mutually exclusive with --crf.",
     },
     gpu: { type: "boolean", description: "Use GPU encoding", default: false },
+    "browser-gpu": {
+      type: "boolean",
+      description:
+        "Use host GPU acceleration for Chrome/WebGL capture. Enabled by default for local renders; use --no-browser-gpu to opt out.",
+    },
     quiet: {
       type: "boolean",
       description: "Suppress verbose output",
@@ -186,6 +192,8 @@ export default defineCommand({
 
     const useDocker = args.docker ?? false;
     const useGpu = args.gpu ?? false;
+    const browserGpuArg = args["browser-gpu"];
+    const useBrowserGpu = resolveBrowserGpuForCli(useDocker, browserGpuArg);
     const quiet = args.quiet ?? false;
     const strictAll = args["strict-all"] ?? false;
     const strictErrors = (args.strict ?? false) || strictAll;
@@ -197,6 +205,15 @@ export default defineCommand({
       process.exit(1);
     }
 
+    if (useDocker && browserGpuArg === true) {
+      errorBox(
+        "Browser GPU is local-only",
+        "--browser-gpu uses the host Chrome GPU backend. Docker mode keeps browser rendering deterministic and does not expose a cross-platform Chrome GPU backend.",
+        "Run without --docker, or use --gpu for Docker GPU encoding where your Docker host supports GPU passthrough.",
+      );
+      process.exit(1);
+    }
+
     let crf: number | undefined;
     if (crfRaw != null) {
       const parsed = Number(crfRaw);
@@ -227,6 +244,13 @@ export default defineCommand({
           c.dim(" \u2192 " + outputPath),
       );
       console.log(c.dim("   " + fps + "fps \u00B7 " + quality + " \u00B7 " + workerLabel));
+      if (useGpu || useBrowserGpu) {
+        const gpuModes = [
+          useGpu ? "encoder GPU" : null,
+          useBrowserGpu ? "browser GPU (auto)" : null,
+        ].filter(Boolean);
+        console.log(c.dim("   GPU: " + gpuModes.join(" + ")));
+      }
       console.log("");
     }
 
@@ -312,6 +336,7 @@ export default defineCommand({
         format,
         workers,
         gpu: useGpu,
+        browserGpu: useBrowserGpu,
         hdrMode: args.sdr ? "force-sdr" : args.hdr ? "force-hdr" : "auto",
         crf,
         videoBitrate,
@@ -324,6 +349,7 @@ export default defineCommand({
         format,
         workers,
         gpu: useGpu,
+        browserGpu: useBrowserGpu,
         hdrMode: args.sdr ? "force-sdr" : args.hdr ? "force-hdr" : "auto",
         crf,
         videoBitrate,
@@ -340,6 +366,7 @@ interface RenderOptions {
   format: "mp4" | "webm" | "mov";
   workers?: number;
   gpu: boolean;
+  browserGpu: boolean;
   hdrMode: "auto" | "force-hdr" | "force-sdr";
   crf?: number;
   videoBitrate?: string;
@@ -347,6 +374,17 @@ interface RenderOptions {
   browserPath?: string;
 }
 
+export function resolveBrowserGpuForCli(
+  useDocker: boolean,
+  browserGpuArg: boolean | undefined,
+  envMode = process.env.PRODUCER_BROWSER_GPU_MODE,
+): boolean {
+  if (useDocker) return false;
+  if (browserGpuArg !== undefined) return browserGpuArg;
+  if (envMode === "software") return false;
+  return true;
+}
+
 const DOCKER_IMAGE_PREFIX = "hyperframes-renderer";
 
 function dockerImageTag(version: string): string {
@@ -464,6 +502,7 @@ async function renderDocker(
       format: options.format,
       workers: options.workers,
       gpu: options.gpu,
+      browserGpu: options.browserGpu,
       hdrMode: options.hdrMode,
       crf: options.crf,
       videoBitrate: options.videoBitrate,
@@ -508,7 +547,7 @@ async function renderDocker(
   printRenderComplete(outputPath, elapsed, options.quiet);
 }
 
-async function renderLocal(
+export async function renderLocal(
   projectDir: string,
   outputPath: string,
   options: RenderOptions,
@@ -530,6 +569,9 @@ async function renderLocal(
     format: options.format,
     workers: options.workers,
     useGpu: options.gpu,
+    producerConfig: producer.resolveConfig({
+      browserGpuMode: options.browserGpu ? "hardware" : "software",
+    }),
     hdrMode: options.hdrMode,
     crf: options.crf,
     videoBitrate: options.videoBitrate,

packages/cli/src/docs/rendering.md

@@ -19,11 +19,14 @@ Requires: Docker installed and running.
 - `-w, --workers` — Parallel workers 1-8 (default: auto)
 - `--crf` — Override encoder CRF (mutually exclusive with `--video-bitrate`)
 - `--video-bitrate` — Target video bitrate such as `10M` (mutually exclusive with `--crf`)
-- `--gpu` — Use GPU encoding (NVENC, VideoToolbox, VAAPI)
+- `--gpu` — Use GPU encoding (NVENC, VideoToolbox, VAAPI, QSV)
+- `--browser-gpu` / `--no-browser-gpu` — Use or opt out of host GPU acceleration for local Chrome/WebGL capture (enabled by default for local renders, disabled in Docker)
 - `-o, --output` — Custom output path
 
 ## Tips
 
 - Use `draft` quality for fast previews during development
+- Local renders use browser GPU capture automatically; use `--no-browser-gpu` to compare against the software-browser path
+- Use `--gpu` when a local render also benefits from hardware FFmpeg encoding
 - Use `npx hyperframes benchmark` to find optimal settings
 - 4 workers is usually the sweet spot for most compositions

packages/cli/src/utils/dockerRunArgs.test.ts

@@ -6,6 +6,7 @@ const BASE: DockerRenderOptions = {
   quality: "standard",
   format: "mp4",
   gpu: false,
+  browserGpu: false,
   hdrMode: "auto",
   crf: undefined,
   videoBitrate: undefined,
@@ -42,6 +43,7 @@ describe("buildDockerRunArgs", () => {
         "standard",
         "--format",
         "mp4",
+        "--no-browser-gpu",
       ]
     `);
   });
@@ -91,6 +93,7 @@ describe("buildDockerRunArgs", () => {
         "18",
         "--quiet",
         "--gpu",
+        "--no-browser-gpu",
         "--hdr",
       ]
     `);
@@ -139,6 +142,11 @@ describe("buildDockerRunArgs", () => {
     expect(on).toContain("--gpu");
   });
 
+  it("forces software browser capture inside Docker", () => {
+    const args = buildDockerRunArgs({ ...FIXED_INPUT, options: BASE });
+    expect(args).toContain("--no-browser-gpu");
+  });
+
   it("forwards every renderer-shaped option (regression tripwire for silent drops)", () => {
     const args = buildDockerRunArgs({
       ...FIXED_INPUT,
@@ -148,6 +156,7 @@ describe("buildDockerRunArgs", () => {
         format: "webm",
         workers: 8,
         gpu: true,
+        browserGpu: false,
         hdrMode: "force-hdr",
         crf: 16,
         videoBitrate: undefined,
@@ -165,6 +174,7 @@ describe("buildDockerRunArgs", () => {
     expect(args).toContain("16");
     expect(args).toContain("--quiet");
     expect(args).toContain("--gpu");
+    expect(args).toContain("--no-browser-gpu");
     expect(args).toContain("--hdr");
   });
 

packages/cli/src/utils/dockerRunArgs.ts

@@ -24,6 +24,7 @@ export interface DockerRenderOptions {
   format: "mp4" | "webm" | "mov";
   workers?: number;
   gpu: boolean;
+  browserGpu: boolean;
   hdrMode: "auto" | "force-hdr" | "force-sdr";
   crf?: number;
   videoBitrate?: string;
@@ -59,6 +60,7 @@ export function buildDockerRunArgs(input: DockerRunArgsInput): string[] {
     ...(options.videoBitrate ? ["--video-bitrate", options.videoBitrate] : []),
     ...(options.quiet ? ["--quiet"] : []),
     ...(options.gpu ? ["--gpu"] : []),
+    ...(options.browserGpu ? [] : ["--no-browser-gpu"]),
     ...(options.hdrMode === "force-hdr" ? ["--hdr"] : []),
     ...(options.hdrMode === "force-sdr" ? ["--sdr"] : []),
   ];

packages/engine/src/config.test.ts

@@ -29,6 +29,7 @@ describe("resolveConfig", () => {
     expect(config.quality).toBe("standard");
     expect(config.format).toBe("jpeg");
     expect(config.jpegQuality).toBe(80);
+    expect(config.browserGpuMode).toBe("software");
     expect(config.audioGain).toBe(1);
     expect(config.debug).toBe(false);
   });
@@ -66,6 +67,20 @@ describe("resolveConfig", () => {
     expect(config.disableGpu).toBe(false);
   });
 
+  it("reads browser GPU mode from env", () => {
+    setEnv("PRODUCER_BROWSER_GPU_MODE", "hardware");
+
+    const config = resolveConfig();
+    expect(config.browserGpuMode).toBe("hardware");
+  });
+
+  it("falls back to software browser GPU mode for invalid env values", () => {
+    setEnv("PRODUCER_BROWSER_GPU_MODE", "native");
+
+    const config = resolveConfig();
+    expect(config.browserGpuMode).toBe("software");
+  });
+
   it("explicit overrides take precedence over env vars", () => {
     setEnv("PRODUCER_CORES_PER_WORKER", "5");
 

packages/engine/src/config.ts

@@ -31,6 +31,11 @@ export interface EngineConfig {
   // ── Browser ──────────────────────────────────────────────────────────
   chromePath?: string;
   disableGpu: boolean;
+  /**
+   * Chrome/WebGL rendering backend. "software" keeps the existing SwiftShader
+   * path for reproducible output; "hardware" lets Chrome use the host GPU.
+   */
+  browserGpuMode: "software" | "hardware";
   enableBrowserPool: boolean;
   browserTimeout: number;
   protocolTimeout: number;
@@ -113,6 +118,7 @@ export const DEFAULT_CONFIG: EngineConfig = {
   largeRenderThreshold: 1000,
 
   disableGpu: false,
+  browserGpuMode: "software",
   enableBrowserPool: false,
   browserTimeout: 120_000,
   protocolTimeout: 300_000,
@@ -158,6 +164,11 @@ export function resolveConfig(overrides?: Partial<EngineConfig>): EngineConfig {
     if (raw === undefined) return fallback;
     return raw === "true";
   };
+  const envBrowserGpuMode = (): EngineConfig["browserGpuMode"] => {
+    const raw = env("PRODUCER_BROWSER_GPU_MODE");
+    if (raw === "hardware" || raw === "software") return raw;
+    return DEFAULT_CONFIG.browserGpuMode;
+  };
 
   // Env-var layer (backward compat)
   const fromEnv: Partial<EngineConfig> = {
@@ -171,6 +182,7 @@ export function resolveConfig(overrides?: Partial<EngineConfig>): EngineConfig {
 
     chromePath: env("PRODUCER_HEADLESS_SHELL_PATH"),
     disableGpu: envBool("PRODUCER_DISABLE_GPU", DEFAULT_CONFIG.disableGpu),
+    browserGpuMode: envBrowserGpuMode(),
     enableBrowserPool: envBool("PRODUCER_ENABLE_BROWSER_POOL", DEFAULT_CONFIG.enableBrowserPool),
     browserTimeout: envNum("PRODUCER_PUPPETEER_LAUNCH_TIMEOUT_MS", DEFAULT_CONFIG.browserTimeout),
     protocolTimeout: envNum(

packages/engine/src/services/browserManager.test.ts

@@ -0,0 +1,48 @@
+import { describe, expect, it } from "vitest";
+
+import { buildChromeArgs } from "./browserManager.js";
+
+describe("buildChromeArgs browser GPU mode", () => {
+  const base = { width: 1920, height: 1080 };
+
+  it("uses SwiftShader software GL by default for reproducible local renders", () => {
+    const args = buildChromeArgs(base);
+    expect(args).toContain("--use-gl=angle");
+    expect(args).toContain("--use-angle=swiftshader");
+    expect(args).not.toContain("--enable-gpu-rasterization");
+  });
+
+  it("uses Metal-backed ANGLE for hardware browser GPU mode on macOS", () => {
+    const args = buildChromeArgs({ ...base, platform: "darwin" }, { browserGpuMode: "hardware" });
+    expect(args).toContain("--use-gl=angle");
+    expect(args).toContain("--use-angle=metal");
+    expect(args).toContain("--enable-gpu-rasterization");
+    expect(args).not.toContain("--use-angle=swiftshader");
+  });
+
+  it("uses D3D11-backed ANGLE for hardware browser GPU mode on Windows", () => {
+    const args = buildChromeArgs({ ...base, platform: "win32" }, { browserGpuMode: "hardware" });
+    expect(args).toContain("--use-gl=angle");
+    expect(args).toContain("--use-angle=d3d11");
+    expect(args).toContain("--enable-gpu-rasterization");
+    expect(args).not.toContain("--use-angle=swiftshader");
+  });
+
+  it("uses EGL for hardware browser GPU mode on Linux", () => {
+    const args = buildChromeArgs({ ...base, platform: "linux" }, { browserGpuMode: "hardware" });
+    expect(args).toContain("--use-gl=egl");
+    expect(args).toContain("--enable-gpu-rasterization");
+    expect(args).not.toContain("--use-gl=angle");
+    expect(args).not.toContain("--use-angle=swiftshader");
+  });
+
+  it("keeps --disable-gpu authoritative when requested", () => {
+    const args = buildChromeArgs(
+      { ...base, platform: "darwin" },
+      { browserGpuMode: "hardware", disableGpu: true },
+    );
+    expect(args).toContain("--disable-gpu");
+    expect(args).toContain("--use-angle=swiftshader");
+    expect(args).not.toContain("--use-angle=metal");
+  });
+});

packages/engine/src/services/browserManager.ts

@@ -238,12 +238,18 @@ export interface BuildChromeArgsOptions {
   width: number;
   height: number;
   captureMode?: CaptureMode;
+  platform?: NodeJS.Platform;
 }
 
 export function buildChromeArgs(
   options: BuildChromeArgsOptions,
-  config?: Partial<Pick<EngineConfig, "disableGpu" | "chromePath">>,
+  config?: Partial<Pick<EngineConfig, "browserGpuMode" | "disableGpu" | "chromePath">>,
 ): string[] {
+  const platform = options.platform ?? process.platform;
+  const gpuDisabled = config?.disableGpu ?? DEFAULT_CONFIG.disableGpu;
+  const browserGpuMode = gpuDisabled
+    ? "software"
+    : (config?.browserGpuMode ?? DEFAULT_CONFIG.browserGpuMode);
   // Chrome flags tuned for headless rendering performance. The set below is a
   // fairly standard "headless-for-capture" configuration — similar profiles
   // appear in Puppeteer's defaults, Playwright, Remotion, and Chrome's own
@@ -254,8 +260,7 @@ export function buildChromeArgs(
     "--disable-dev-shm-usage",
     "--enable-webgl",
     "--ignore-gpu-blocklist",
-    "--use-gl=angle",
+    ...getBrowserGpuArgs(browserGpuMode, platform),
-    "--use-angle=swiftshader",
     "--font-render-hinting=none",
     "--force-color-profile=srgb",
     `--window-size=${options.width},${options.height}`,
@@ -301,9 +306,28 @@ export function buildChromeArgs(
     );
   }
 
-  const gpuDisabled = config?.disableGpu ?? DEFAULT_CONFIG.disableGpu;
   if (gpuDisabled) {
     chromeArgs.push("--disable-gpu");
   }
   return chromeArgs;
 }
+
+function getBrowserGpuArgs(
+  mode: EngineConfig["browserGpuMode"],
+  platform: NodeJS.Platform,
+): string[] {
+  if (mode === "software") {
+    return ["--use-gl=angle", "--use-angle=swiftshader"];
+  }
+
+  switch (platform) {
+    case "darwin":
+      return ["--use-gl=angle", "--use-angle=metal", "--enable-gpu-rasterization"];
+    case "win32":
+      return ["--use-gl=angle", "--use-angle=d3d11", "--enable-gpu-rasterization"];
+    case "linux":
+      return ["--use-gl=egl", "--enable-gpu-rasterization"];
+    default:
+      return ["--enable-gpu-rasterization"];
+  }
+}

packages/producer/src/services/renderOrchestrator.test.ts

@@ -206,6 +206,7 @@ function createConfig(): EngineConfig {
     minParallelFrames: 120,
     largeRenderThreshold: 1000,
     disableGpu: false,
+    browserGpuMode: "software",
     enableBrowserPool: false,
     browserTimeout: 120000,
     protocolTimeout: 300000,