Deployment — Server-Side Render Pipeline

wity-scene provides a single public Lambda entry point (witySceneRender) that accepts a scene XML document and returns a rendered output URL. Internally it inspects the scene and routes to the appropriate specialized Lambda — callers never need to know about the internal pipeline.

For the full API reference of the gateway, see witySceneRender.

---

Public entry point — witySceneRender

caller → witySceneRender (gateway)
              │ outputFormat?
              ├─ "mp4"  → witySceneToVideo? → witySceneCompose? → final URL
              ├─ "pdf"  → witySceneToPdf  → S3 URL
              └─ "pptx" → witySceneToPptx → S3 URL

Input

{
  "sceneXml":      "<wity-scene>...</wity-scene>",
  "outputFormat":  "mp4",
  "options": {
    "fps":          30,
    "sceneWidth":   1920,
    "sceneHeight":  1080,
    "fontManifest": { "Inter": "https://cdn.example.com/Inter.ttf" }
  }
}

outputFormat defaults to "mp4". Supported values: "mp4", "pdf", "pptx".

PDF-specific options:

Option	Type	Default	Description
options.ts	number	0	Scene time (seconds) to snapshot
options.timestamps	number[]	—	One page per timestamp (overrides ts)
options.variant	"standard" | "print"	"standard"	Screen (96 DPI) or high-res (300 DPI)
options.dpi	number	96 / 300	Override render DPI

PPTX-specific options:

Option	Type	Default	Description
options.ts	number	0	Scene time (seconds) to snapshot
options.timestamps	number[]	—	One slide per timestamp (overrides ts)

Response

{ "url": "https://...", "fileSize": 9876543, "pipeline": ["witySceneToVideo", "witySceneCompose"] }

For "pdf": response also includes "pages": number. For "pptx": response also includes "slides": number.

pipeline lists which downstream Lambdas were invoked — useful for observability and cost attribution.

Lambda config

Setting	Value
Function name	witySceneRender
Memory	512 MB
Timeout	660 s (11 min — accommodates two sequential 300 s downstream calls for mp4)
Ephemeral storage	512 MB
Runtime	Node.js 20
No FFmpeg layer needed	This Lambda only parses and routes — no media processing

Environment variables

Variable	Default	Description
SCENE_TO_VIDEO_FUNCTION	witySceneToVideo	Graphics compiler Lambda
SCENE_COMPOSE_FUNCTION	witySceneCompose	Full compositor Lambda
SCENE_TO_PDF_FUNCTION	witySceneToPdf	PDF renderer Lambda
SCENE_TO_PPTX_FUNCTION	witySceneToPptx	PPTX renderer Lambda

IAM note: The execution role must have lambda:InvokeFunction permission on all downstream Lambdas.

---

Routing logic

Routing is determined by outputFormat:

outputFormat: "mp4" (default)

The gateway additionally inspects element tags to pick the minimal pipeline:

Scene content	Route
Only ws-rect / ws-text / ws-image	witySceneToVideo only
Graphics + ws-video / ws-audio	witySceneToVideo → witySceneCompose
Only ws-video / ws-audio	witySceneCompose only (graphicsMp4Url: null)

outputFormat: "pdf" or "pptx"

Single downstream Lambda call. No content inspection — all element types are accepted; ws-video and ws-audio are silently skipped in the snapshot renderer (graphics only).

All routing is entirely internal. The caller provides sceneXml + outputFormat and receives a URL.

---

Adding a new output format

The gateway uses a PIPELINE_REGISTRY keyed by outputFormat. To add a new format:

1. Deploy a specialized Lambda.
2. Add one entry to PIPELINE_REGISTRY in handler.js.
3. Add the env var to the gateway config.

No changes to any existing Lambda.

---

Downstream Lambdas (internal)

These are internal infrastructure. Callers must not invoke them directly — use witySceneRender.

witySceneToVideo — graphics compiler

Renders ws-rect, ws-text, ws-image → PNG frame sequence → silent MP4.

Package: @wity/scene-to-video

Setting	Value
Memory	3008 MB · Timeout 300 s · Ephemeral 4096 MB
FFmpeg layer	arn:aws:lambda:ap-south-1:175033217214:layer:ffmpeg:1

Input: { sceneXml, fontManifest?, fps? } → { url, fileSize }

witySceneCompose — full compositor

Blends ws-video + ws-audio + optional graphics overlay via FFmpeg filter_complex.

Package: @wity/scene-compose

Setting	Value
Memory	2048 MB · Timeout 300 s · Ephemeral 4096 MB
FFmpeg layer	arn:aws:lambda:ap-south-1:175033217214:layer:ffmpeg:1

Input: { sceneXml, graphicsMp4Url?, fps?, sceneWidth?, sceneHeight? } → { url, fileSize }

witySceneToPdf — PDF renderer

Snapshots scene at one or more timestamps via node-canvas → pdf-lib → PDF.

Package: @wity/scene-to-pdf

Setting	Value
Memory	2048 MB · Timeout 120 s · Ephemeral 1024 MB
No FFmpeg layer needed

Input: { sceneXml, fontManifest?, ts?, timestamps?, variant?, dpi? } → { url, fileSize, pages }

witySceneToPptx — PPTX renderer

Snapshots scene at one or more timestamps via node-canvas → pptxgenjs → PowerPoint.

Package: @wity/scene-to-pptx

Setting	Value
Memory	1024 MB · Timeout 120 s · Ephemeral 1024 MB
No FFmpeg layer needed

Input: { sceneXml, fontManifest?, ts?, timestamps? } → { url, fileSize, slides }

---

Analysis services (separate from rendering)

Analysis services are standalone Lambdas — they are not routed through witySceneRender.

wityAudioProfile — audio loudness analysis

Downloads all ws-video and ws-audio media, decodes audio via FFmpeg, measures windowed RMS loudness per element, and returns per-element + power-mixed loudness profiles in dBFS.

Not an output format — callers invoke this Lambda directly.

See the wityAudioProfile reference for full input/output documentation.

Setting	Value
Memory	1024 MB · Timeout 120 s · Ephemeral 2048 MB
FFmpeg layer	arn:aws:lambda:ap-south-1:175033217214:layer:ffmpeg:1

Input: { sceneXml, options?: { windowMs?, sampleRate? } } → AudioProfileResult

---

S3 bucket setup

All downstream Lambdas write to wity-user-generated-content under separate prefixes:

Lambda	Prefix
witySceneToVideo	scene-compiled/
witySceneCompose	scene-composed/
witySceneToPdf	scene-pdf/
witySceneToPptx	scene-pptx/

The gateway Lambda does not write to S3 — it only orchestrates.