@giveitsmaller/contracts 0.9.0 → 0.16.0

package/availability/availability.json

@@ -1,6 +1,13 @@
 {
-  "capabilities_version": 60,
+  "capabilities_version": 89,
   "endpoints": {
+    "DELETE /api/auth/api-keys/{apiKeyId}": {
+      "auth": "required",
+      "availability": "stable",
+      "identity_scoped": true,
+      "operation_id": "revokeApiKey",
+      "required_tier": null
+    },
     "GET /api/operations/schema": {
       "auth": "optional",
       "availability": "stable",
@@ -36,6 +43,13 @@
       "operation_id": "getCreditsUsage",
       "required_tier": null
     },
+    "GET /api/workflows": {
+      "auth": "required",
+      "availability": "stable",
+      "identity_scoped": true,
+      "operation_id": "listWorkflows",
+      "required_tier": null
+    },
     "GET /api/workflows/{id}/downloads": {
       "auth": "optional",
       "availability": "stable",
@@ -71,6 +85,13 @@
       "operation_id": "readiness",
       "required_tier": null
     },
+    "PATCH /api/auth/profile": {
+      "auth": "required",
+      "availability": "stable",
+      "identity_scoped": true,
+      "operation_id": "updateProfile",
+      "required_tier": null
+    },
     "POST /api/audio-watermark/decode": {
       "auth": "required",
       "availability": "stable",
@@ -78,6 +99,34 @@
       "operation_id": "decodeAudioWatermark",
       "required_tier": null
     },
+    "POST /api/auth/api-keys": {
+      "auth": "required",
+      "availability": "stable",
+      "identity_scoped": true,
+      "operation_id": "createApiKey",
+      "required_tier": null
+    },
+    "POST /api/auth/change-password": {
+      "auth": "required",
+      "availability": "stable",
+      "identity_scoped": true,
+      "operation_id": "changePassword",
+      "required_tier": null
+    },
+    "POST /api/auth/confirm-email-change": {
+      "auth": "anonymous",
+      "availability": "stable",
+      "identity_scoped": false,
+      "operation_id": "confirmEmailChange",
+      "required_tier": null
+    },
+    "POST /api/auth/forgot-password": {
+      "auth": "anonymous",
+      "availability": "stable",
+      "identity_scoped": false,
+      "operation_id": "forgotPassword",
+      "required_tier": null
+    },
     "POST /api/auth/login": {
       "auth": "anonymous",
       "availability": "stable",
@@ -92,6 +141,27 @@
       "operation_id": "logoutUser",
       "required_tier": null
     },
+    "POST /api/auth/register": {
+      "auth": "anonymous",
+      "availability": "stable",
+      "identity_scoped": false,
+      "operation_id": "registerUser",
+      "required_tier": null
+    },
+    "POST /api/auth/reset-password": {
+      "auth": "anonymous",
+      "availability": "stable",
+      "identity_scoped": false,
+      "operation_id": "resetPassword",
+      "required_tier": null
+    },
+    "POST /api/auth/verify-email": {
+      "auth": "anonymous",
+      "availability": "stable",
+      "identity_scoped": false,
+      "operation_id": "verifyEmail",
+      "required_tier": null
+    },
     "POST /api/contact": {
       "auth": "anonymous",
       "availability": "stable",
@@ -190,7 +260,7 @@
       "options": {
         "folder_structure": {
           "default": "flat",
-          "description": "flat = all files in root, by_job = subfolder per source job",
+          "description": "flat = all files together at the top level; by_job = a separate subfolder for each source",
           "type": "enum",
           "values": [
             "flat",
@@ -211,7 +281,7 @@
     "audio_overlay": {
       "availability": "planned",
       "default": false,
-      "description": "Mix a secondary audio asset (overlay) over a primary audio or\nvideo base. Use cases: DJ tags, podcast intros/outros, station\nIDs, jingles, branded mid-rolls.\n\n**Industry naming.** This is \"audio overlay\" or \"audio branding\"\n\u2014 NOT \"audio watermark\". Audio watermarking is steganographic\n(imperceptible identifier embedded in audio for ownership /\nforensic tracking) and is owned by the separate `audio_watermark`\noperation (planned per ticket I20).\n\nMulti-input (Path B with `role: base` + `role: overlay` per\n`JobInputV2.role`). `min_inputs: 2 = max_inputs: 2` for V2.0;\nmulti-overlay support is advertised as a planned feature\n(`features.multi_overlay_stack`) but Lambda-side implementation\ndeferred.\n\nThe overlay can be uploaded directly, imported externally, or\nreferenced from a vault connection per `WorkflowSource` (from\nI10). The base is whatever single-asset source the workflow\nupstream has \u2014 for `mime_groups.audio`, the base is the primary\naudio track; for `mime_groups.video`, the overlay is mixed into\nthe video's audio track and the visual track passes through\nuntouched.\n\nPer ADR-0001 \u00a71.3 (Tension 1 \u2014 `planned` operations return\n`feature_not_available` 422 until Lambda support ships). Per\nADR-0004 \u00a7\"V2 JobDefinition\" + plan v5 \u00a7F1 round 6.\n",
+      "description": "Mix a secondary audio asset (overlay) over a primary audio or\nvideo base. Use cases: DJ tags, podcast intros/outros, station\nIDs, jingles, branded mid-rolls.\n\n**Industry naming.** This is \"audio overlay\" or \"audio branding\"\n\u2014 NOT \"audio watermark\". Audio watermarking is steganographic\n(imperceptible identifier embedded in audio for ownership /\nforensic tracking) and is owned by the separate `audio_watermark`\noperation (planned per ticket I20).\n\nMulti-input (Path B with `role: base` + `role: overlay` per\n`JobInputV2.role`). `min_inputs: 2 = max_inputs: 2` for V2.0;\nmulti-overlay support is advertised as a planned feature\n(`features.multi_overlay_stack`) but Lambda-side implementation\ndeferred.\n\nEach input is a `MultiInputSource` \u2014 imported externally,\nreferenced from a vault connection, or an upstream `job_output`\n(uploads are NOT referenced directly inside inputs[]). An uploaded\noverlay or base enters via a `passthrough` source job referenced by\n`{ type: job_output, from: <id> }` (per ticket 4som89Uh). The base\nis whatever single-asset source the workflow\nupstream has \u2014 for `mime_groups.audio`, the base is the primary\naudio track; for `mime_groups.video`, the overlay is mixed into\nthe video's audio track and the visual track passes through\nuntouched.\n\nPer ADR-0001 \u00a71.3 (Tension 1 \u2014 `planned` operations return\n`feature_not_available` 422 until Lambda support ships). Per\nADR-0004 \u00a7\"V2 JobDefinition\" + plan v5 \u00a7F1 round 6.\n",
       "features": {
         "multi_overlay_stack": {
           "availability": "planned",
@@ -250,7 +320,7 @@
               "depends_on": {
                 "mode": "duck"
               },
-              "description": "Side-chain compression ratio. Higher values produce more\naggressive ducking. Only applies when `mode: duck`.\n",
+              "description": "How strongly the base ducks \u2014 higher values dip it more\naggressively. Only applies when `mode: duck`.\n",
               "max": 20.0,
               "min": 1.0,
               "type": "float"
@@ -270,7 +340,7 @@
               "depends_on": {
                 "mode": "duck"
               },
-              "description": "Side-chain threshold (dB). Base audio is ducked when the\noverlay exceeds this level. Only applies when `mode: duck`.\n",
+              "description": "Level (dB) the overlay must reach before the base audio\nducks. Only applies when `mode: duck`.\n",
               "max": 0.0,
               "min": -60.0,
               "type": "float"
@@ -319,7 +389,7 @@
             },
             "mode": {
               "default": "mix",
-              "description": "How the overlay combines with the base audio:\n- `mix`: simple sum (overlay + base; both audible).\n- `duck`: side-chain compress the base under the overlay\n  (base attenuates while overlay plays \u2014 broadcast pattern).\n- `replace`: mute the base for the overlay's duration\n  (overlay only; useful for hard-cut station IDs).\n",
+              "description": "How the overlay combines with the base audio:\n- `mix`: both play together (overlay added on top; both audible).\n- `duck`: the base audio dips while the overlay plays, then\n  returns \u2014 the broadcast voice-over pattern.\n- `replace`: mute the base for the overlay's duration\n  (overlay only; useful for hard-cut station IDs).\n",
               "per_value_availability": {
                 "duck": {
                   "availability": "planned"
@@ -402,7 +472,7 @@
               "depends_on": {
                 "mode": "duck"
               },
-              "description": "Side-chain compression ratio. Only applies when `mode: duck`.",
+              "description": "How strongly the base ducks \u2014 higher values dip it more aggressively. Only applies when `mode: duck`.",
               "max": 20.0,
               "min": 1.0,
               "type": "float"
@@ -422,7 +492,7 @@
               "depends_on": {
                 "mode": "duck"
               },
-              "description": "Side-chain threshold (dB). Only applies when `mode: duck`.",
+              "description": "Level (dB) the overlay must reach before the base audio ducks. Only applies when `mode: duck`.",
               "max": 0.0,
               "min": -60.0,
               "type": "float"
@@ -471,7 +541,7 @@
             },
             "mode": {
               "default": "mix",
-              "description": "Same modes as the audio mime_group. `replace` mutes the\nbase video's audio for the overlay's duration.\n",
+              "description": "Same modes as for audio overlays. `replace` mutes the\nbase video's audio for the overlay's duration.\n",
               "per_value_availability": {
                 "duck": {
                   "availability": "planned"
@@ -543,7 +613,17 @@
           }
         }
       },
-      "min_inputs": 2
+      "min_inputs": 2,
+      "per_role_cardinality": {
+        "base": {
+          "max": 1,
+          "min": 1
+        },
+        "overlay": {
+          "max": 1,
+          "min": 1
+        }
+      }
     },
     "audio_to_video": {
       "availability": "beta",
@@ -561,6 +641,7 @@
             "audio/flac",
             "audio/ogg",
             "audio/opus",
+            "audio/mp4",
             "audio/x-m4a"
           ],
           "options": {
@@ -571,7 +652,7 @@
             },
             "framerate": {
               "default": 2,
-              "description": "Output video framerate. Default `2` fps \u2014 combined with\nFFmpeg `-tune stillimage`, keeps the file size tiny for\nthe common case of a single still overlay. Raise for\nanimated overlays (not supported in V1 \u2014 overlay must be\na still image) or platform requirements (e.g. YouTube\nrecommends 24+ fps; some platforms accept lower).\n",
+              "description": "Output video framerate. Default `2` fps keeps the file size\ntiny for the common case of a single still overlay. Raise for\nanimated overlays (not supported in V1 \u2014 overlay must be\na still image) or platform requirements (e.g. YouTube\nrecommends 24+ fps; some platforms accept lower).\n",
               "max": 60,
               "min": 1,
               "type": "integer"
@@ -588,7 +669,7 @@
             },
             "output_format": {
               "default": "mp4",
-              "description": "Output container format. V1 supports `mp4` only. Future\nadditions (webm, mov) await Lambda follow-up.\n",
+              "description": "Output container format. V1 supports `mp4` only; webm and mov\nare planned additions.\n",
               "type": "enum",
               "values": [
                 "mp4"
@@ -647,7 +728,7 @@
             },
             "method": {
               "default": "auto",
-              "description": "Watermarking algorithm:\n- `auto`: server selects the best available method for the\n  input (currently maps to `psychoacoustic`; will widen as\n  `neural` ships).\n- `psychoacoustic`: classical masking-curve embedding.\n  Resemble's open-source PerTh library provides the path-\n  to-prod reference implementation. Survives MP3/AAC at\n  moderate bitrates and broadcast chains.\n- `neural`: ML-based embedding (deep generative). Higher\n  survival under aggressive transcoding + analog-hole\n  capture; longer encoding time + non-deterministic.\n",
+              "description": "Watermarking algorithm:\n- `auto`: server selects the best available method for the\n  input (currently maps to `psychoacoustic`; will widen as\n  `neural` ships).\n- `psychoacoustic`: hides the mark below what the ear can\n  hear, so it stays inaudible. Survives MP3/AAC at moderate\n  bitrates and broadcast chains.\n- `neural`: machine-learning embedding. Higher survival\n  under aggressive re-encoding and microphone re-recording;\n  longer encoding time and non-deterministic.\n",
               "per_value_availability": {
                 "neural": {
                   "availability": "planned"
@@ -661,13 +742,13 @@
               ]
             },
             "payload": {
-              "description": "Optional caller-supplied payload to embed (max 256\ncharacters). Use cases: ownership marker, distribution\nchannel ID, license key. When omitted, the server\ngenerates a UUID-v7 `watermark_id` and embeds that\ninstead; the generated ID is returned on\n`OperationResult` so the caller can subsequently call\nthe decode endpoint to verify.\n",
+              "description": "Optional caller-supplied payload to embed (max 256\ncharacters). Use cases: ownership marker, distribution\nchannel ID, license key. When omitted, the server\ngenerates an identifier automatically and embeds that\ninstead; the generated ID is returned on\n`OperationResult` so the caller can subsequently call\nthe decode endpoint to verify.\n",
               "required": false,
               "type": "string"
             },
             "robustness": {
               "default": "compression_resistant",
-              "description": "Survival profile:\n- `light`: minimal payload-strength; smallest perceptual\n  footprint. Survives a single re-encode pass at moderate\n  bitrate.\n- `compression_resistant` (default): survives multiple\n  MP3/AAC re-encodes at typical streaming bitrates.\n- `broadcast_grade`: survives broadcast compression\n  chains, EQ, dynamic-range compression, and analog-hole\n  capture (microphone re-recording). Higher embedding\n  time + slightly larger perceptual footprint.\n",
+              "description": "Survival profile:\n- `light`: minimal payload-strength; smallest perceptual\n  footprint. Survives a single re-encode pass at moderate\n  bitrate.\n- `compression_resistant` (default): survives multiple\n  MP3/AAC re-encodes at typical streaming bitrates.\n- `broadcast_grade`: survives broadcast compression\n  chains, EQ, dynamic-range compression, and microphone\n  re-recording. Higher embedding time + slightly larger\n  perceptual footprint.\n",
               "type": "enum",
               "values": [
                 "light",
@@ -716,7 +797,7 @@
             },
             "method": {
               "default": "auto",
-              "description": "Watermarking algorithm. Same semantics as the `audio`\nmime_group; embedded into the video's audio track only.\n",
+              "description": "Watermarking algorithm. Same as for an audio input;\nembedded into the video's audio track only.\n",
               "per_value_availability": {
                 "neural": {
                   "availability": "planned"
@@ -730,13 +811,13 @@
               ]
             },
             "payload": {
-              "description": "Optional caller-supplied payload to embed (max 256\ncharacters). When omitted, the server generates a UUID-v7\n`watermark_id` and embeds that instead.\n",
+              "description": "Optional caller-supplied payload to embed (max 256\ncharacters). When omitted, the server generates an\nidentifier automatically and embeds that instead.\n",
               "required": false,
               "type": "string"
             },
             "robustness": {
               "default": "compression_resistant",
-              "description": "Survival profile. Same semantics as audio mime_group.",
+              "description": "Survival profile. Same as for an audio input.",
               "type": "enum",
               "values": [
                 "light",
@@ -820,12 +901,14 @@
               ]
             },
             "trim_end": {
-              "description": "Trim from end in seconds",
+              "availability": "planned",
+              "description": "Trim from end in seconds. Gated `planned` (qZI5EK9j) \u2014 see `trim_start`.",
               "min": 0,
               "type": "float"
             },
             "trim_start": {
-              "description": "Trim from beginning in seconds",
+              "availability": "planned",
+              "description": "Trim from beginning in seconds. Gated `planned` (qZI5EK9j): no compression worker implements per-clip trim yet, so the server returns `feature_not_available` rather than accepting and silently ignoring it. Flips to `stable` when a worker lands; to trim today, chain a `split` op.\n",
               "min": 0,
               "type": "float"
             }
@@ -961,32 +1044,6 @@
             "image/avif"
           ],
           "options": {
-            "auto_orient": {
-              "default": true,
-              "description": "Auto-rotate based on EXIF orientation",
-              "type": "boolean"
-            },
-            "fit": {
-              "default": "max",
-              "depends_on": {
-                "height": "set",
-                "logic": "or",
-                "width": "set"
-              },
-              "description": "Resize mode. max = fit within bounds, crop = fill and trim, scale = force exact dimensions. Only applies when width or height is set.",
-              "type": "enum",
-              "values": [
-                "max",
-                "crop",
-                "scale"
-              ]
-            },
-            "height": {
-              "description": "Maximum output height in pixels",
-              "max": 16384,
-              "min": 1,
-              "type": "integer"
-            },
             "icc_profile": {
               "default": "preserve",
               "description": "ICC color profile handling. preserve = keep original, strip = remove, srgb = convert to sRGB",
@@ -1046,12 +1103,6 @@
               "max": 100,
               "min": 1,
               "type": "integer"
-            },
-            "width": {
-              "description": "Maximum output width in pixels",
-              "max": 16384,
-              "min": 1,
-              "type": "integer"
             }
           }
         },
@@ -1059,7 +1110,7 @@
           "mimes": [
             "video/mp4",
             "video/webm",
-            "video/ogg"
+            "video/quicktime"
           ],
           "options": {
             "audio_bitrate": {
@@ -1110,7 +1161,7 @@
             },
             "encoding_mode": {
               "default": "crf",
-              "description": "crf = constant quality (variable file size), target_size = constrained file size (two-pass)",
+              "description": "crf = constant quality (variable file size), target_size = constrained file size",
               "type": "enum",
               "values": [
                 "crf",
@@ -1119,7 +1170,7 @@
             },
             "faststart": {
               "default": true,
-              "description": "Move MP4 moov atom to start for progressive web playback (MP4 only)",
+              "description": "Move the MP4/MOV moov atom to start for progressive web playback. MP4/MOV only; an explicit faststart on a WebM output is rejected as invalid_options (no moov atom).",
               "type": "boolean"
             },
             "fit": {
@@ -1145,9 +1196,9 @@
               "type": "float"
             },
             "height": {
-              "description": "Output height in pixels",
+              "description": "Output height in pixels. Must be an EVEN integer >= 2 (yuv420p chroma-subsampling); odd values are rejected as invalid_options by the encoder.",
               "max": 4320,
-              "min": 1,
+              "min": 2,
               "type": "integer"
             },
             "preset": {
@@ -1170,24 +1221,26 @@
               "depends_on": {
                 "encoding_mode": "target_size"
               },
-              "description": "Target output file size in bytes (min 1MB). Two-pass encoding to meet this target.",
+              "description": "Target output file size in bytes (min 1MB).",
               "min": 1048576,
               "type": "integer"
             },
             "trim_end": {
-              "description": "Trim from end in seconds",
+              "availability": "planned",
+              "description": "Trim from end in seconds. Gated `planned` (qZI5EK9j) \u2014 see `trim_start`.",
               "min": 0,
               "type": "float"
             },
             "trim_start": {
-              "description": "Trim from beginning in seconds",
+              "availability": "planned",
+              "description": "Trim from beginning in seconds. Gated `planned` (qZI5EK9j): no compression worker implements per-clip trim yet, so the server returns `feature_not_available` rather than accepting and silently ignoring it. Flips to `stable` when a worker lands; to trim today, chain a `split` op.\n",
               "min": 0,
               "type": "float"
             },
             "width": {
-              "description": "Output width in pixels",
+              "description": "Output width in pixels. Must be an EVEN integer >= 2 (yuv420p chroma-subsampling); odd values are rejected as invalid_options by the encoder.",
               "max": 7680,
-              "min": 1,
+              "min": 2,
               "type": "integer"
             }
           },
@@ -1414,7 +1467,7 @@
     "custom_luma": {
       "availability": "planned",
       "default": false,
-      "description": "Apply a caller-uploaded luma matte (transition-mask) over a base\nvideo to produce a custom luma-matte transition effect. Inspired\nby Cloudinary's deprecated luma-matte custom transitions and\nintended as a paid-tier (`pro`+) creative feature.\n\n**NOT** related to FFmpeg's `xfade=custom` filter (which is an\nexpression-mode parameter on the same `xfade` filter). The\n`merge.transition.custom` value is tagged `planned` as an advisory\nthat points callers at this dedicated operation for caller-uploaded\nluma-matte transitions.\n\nMulti-input: exactly one input with `role: base` (the source\nvideo being transitioned) + exactly one input with\n`role: transition_mask` (the luma matte \u2014 a video whose pixel\nluminance drives the transition reveal \u2014 bright = revealed,\ndark = held). The mask can be uploaded directly, imported\nexternally, or referenced from a vault connection per\n`WorkflowSource` (from I10).\n\nPer ADR-0001 \u00a71.3 (Tension 1 \u2014 `planned` operations return\n`feature_not_available` HTTP 422 until Lambda support ships).\nPer ADR-0004 \u00a7\"V2 JobDefinition\" + plan v5 \u00a7F4 round 6\n(codex-corrected \u2014 `xfade=custom` is an expression, not a\ntransition; this operation is the deliberate alternative).\n",
+      "description": "Apply a caller-uploaded luma matte (transition-mask) over a base\nvideo to produce a custom luma-matte transition effect. Inspired\nby Cloudinary's deprecated luma-matte custom transitions and\nintended as a paid-tier (`pro`+) creative feature.\n\n**NOT** related to FFmpeg's `xfade=custom` filter (which is an\nexpression-mode parameter on the same `xfade` filter). The\n`merge.transition.custom` value is tagged `planned` as an advisory\nthat points callers at this dedicated operation for caller-uploaded\nluma-matte transitions.\n\nMulti-input: exactly one input with `role: base` (the source\nvideo being transitioned) + exactly one input with\n`role: transition_mask` (the luma matte \u2014 a video whose pixel\nluminance drives the transition reveal \u2014 bright = revealed,\ndark = held). Each input is a `MultiInputSource` \u2014 imported\nexternally, referenced from a vault connection, or an upstream\n`job_output` (uploads are NOT referenced directly inside inputs[]).\nAn uploaded base or mask enters via a `passthrough` source job\nreferenced by `{ type: job_output, from: <id> }` (per ticket\n4som89Uh).\n\nPer ADR-0001 \u00a71.3 (Tension 1 \u2014 `planned` operations return\n`feature_not_available` HTTP 422 until Lambda support ships).\nPer ADR-0004 \u00a7\"V2 JobDefinition\" + plan v5 \u00a7F4 round 6\n(codex-corrected \u2014 `xfade=custom` is an expression, not a\ntransition; this operation is the deliberate alternative).\n",
       "input_model": "multi",
       "max_inputs": 2,
       "mime_groups": {
@@ -1486,11 +1539,21 @@
         }
       },
       "min_inputs": 2,
+      "per_role_cardinality": {
+        "base": {
+          "max": 1,
+          "min": 1
+        },
+        "transition_mask": {
+          "max": 1,
+          "min": 1
+        }
+      },
       "required_tier": "pro"
     },
     "image_watermark": {
       "default": false,
-      "description": "Apply an image overlay (file or external source) onto a base media\nasset. Multi-input: exactly one input with role: base (the source\nasset) + exactly one with role: overlay (the watermark). The overlay\ncan be an upload, an external_import handle, or a vault connection\nper WorkflowSource (ticket I10).\n\nStable today for static-image bases (image/jpeg, image/png,\nimage/webp); animated GIF bases are advertised as `planned` via\nthe parallel `image_gif` mime_group \u2014 schema is contract-defined,\ndispatch returns `feature_not_available` (422) until Lambda\nsupport ships. Per Tension 1 (ADR-0001) + plan v5 \u00a7F4 round 6.\n\n**Video bases are NOT supported by `image_watermark`.** Use the\ndedicated `video_watermark` operation\n(`schemas/operations/video_watermark.yaml`) which routes through\nFFmpeg overlay + re-encode. Per ADR-0013.\n\n**Audio is explicitly NOT supported.** For audio overlay (DJ tags,\npodcast intros, jingles), use the `audio_overlay` operation\n(declared in schemas/operations/audio_overlay.yaml at\n`availability: planned`).\n\nSource capped at 25 megapixels for image bases (V1 watermark\nprecedent retained); per-frame caps for animated GIF bases will be\ndefined when the corresponding Lambda support ships.\nPer ADR-0004 \u00a7\"V2 JobDefinition\" + plan v5 \u00a7F1.\n",
+      "description": "Apply an image overlay onto a base media\nasset. Multi-input: exactly one input with role: base (the source\nasset) + exactly one with role: overlay (the watermark). Each input\nis a `MultiInputSource` \u2014 an external_import handle, a vault\nconnection, or an upstream `job_output` (uploads are NOT referenced\ndirectly inside inputs[]). To use an uploaded base or overlay, feed\nit through a `passthrough` source job and reference that job via\n`{ type: job_output, from: <id> }` (per ticket 4som89Uh).\n\nStable today for static-image bases (image/jpeg, image/png,\nimage/webp); animated GIF bases are advertised as `planned` via\nthe parallel `image_gif` mime_group \u2014 schema is contract-defined,\ndispatch returns `feature_not_available` (422) until Lambda\nsupport ships. Per Tension 1 (ADR-0001) + plan v5 \u00a7F4 round 6.\n\n**Video bases are NOT supported by `image_watermark`.** Use the\ndedicated `video_watermark` operation\n(`schemas/operations/video_watermark.yaml`) which routes through\nFFmpeg overlay + re-encode. Per ADR-0013.\n\n**Audio is explicitly NOT supported.** For audio overlay (DJ tags,\npodcast intros, jingles), use the `audio_overlay` operation\n(declared in schemas/operations/audio_overlay.yaml at\n`availability: planned`).\n\nSource capped at 25 megapixels for image bases (V1 watermark\nprecedent retained); per-frame caps for animated GIF bases will be\ndefined when the corresponding Lambda support ships.\nPer ADR-0004 \u00a7\"V2 JobDefinition\" + plan v5 \u00a7F1.\n",
       "features": {
         "multi_overlay_stack": {
           "availability": "planned",
@@ -1509,7 +1572,7 @@
           "options": {
             "anchor": {
               "default": "bottom_right",
-              "description": "9-grid anchor position on the base image. Mirrors\nMux/Cloudinary positioning. Combined with margin_x and\nmargin_y for fine-grained offset from the anchor point.\n",
+              "description": "9-grid anchor position on the base image. Combined with\nmargin_x and margin_y for fine-grained offset from the\nanchor point.\n",
               "type": "enum",
               "values": [
                 "top_left",
@@ -1598,7 +1661,17 @@
           }
         }
       },
-      "min_inputs": 2
+      "min_inputs": 2,
+      "per_role_cardinality": {
+        "base": {
+          "max": 1,
+          "min": 1
+        },
+        "overlay": {
+          "max": 1,
+          "min": 1
+        }
+      }
     },
     "merge": {
       "default": false,
@@ -1647,7 +1720,7 @@
             },
             "transition": {
               "default": "none",
-              "description": "Default transition between clips. Audio-only catalog\n(acrossfade-backed); see FORMAT.md \u00a7\"Audio vs visual\ntransition catalogs (merge)\" for additions policy. Future\nvalues (e.g. `fade_curve` variants) will be added with\n`per_value_availability` per ADR-0001 \u00a71.4.\n",
+              "description": "Transition between audio clips. `none` joins them with a\nhard cut; `crossfade` blends one into the next.\n",
               "type": "enum",
               "values": [
                 "none",
@@ -1678,6 +1751,16 @@
                 "none",
                 "crossfade"
               ]
+            },
+            "trim_end": {
+              "description": "Trim from end in seconds (per-input). Applied to this input\nBEFORE any transition/crossfade. Matches the `compress` trim\nsemantic. No `max` \u2014 the upper bound is the input's own\nduration, enforced by the worker, not statically bounded\nhere.\n",
+              "min": 0,
+              "type": "float"
+            },
+            "trim_start": {
+              "description": "Trim from beginning in seconds (per-input). Applied to this\ninput BEFORE any transition/crossfade. Matches the\n`compress` trim semantic. No `max` \u2014 the upper bound is the\ninput's own duration, enforced by the worker, not statically\nbounded here.\n",
+              "min": 0,
+              "type": "float"
             }
           }
         },
@@ -1745,7 +1828,7 @@
               "depends_on": {
                 "output_type": "video"
               },
-              "description": "Transition effect between images in the slideshow output.\nCatalog from FFmpeg 7.1.1 xfade filter (per ADR-0007). See\nper_value_availability for per-value rollout state. The\n`none` value is a sentinel (\"no transition / hard cut\"),\nnot an FFmpeg preset.\n",
+              "description": "Transition effect between images in the slideshow output.\n`none` means a hard cut with no transition. Some effects\nare still rolling out \u2014 check their availability.\n",
               "per_value_availability": {
                 "circleclose": {
                   "availability": "beta"
@@ -1896,7 +1979,8 @@
         "video": {
           "mimes": [
             "video/mp4",
-            "video/webm"
+            "video/webm",
+            "video/quicktime"
           ],
           "options": {
             "codec": {
@@ -1953,7 +2037,7 @@
                   "always"
                 ]
               },
-              "description": "CRF (quality-targeted) vs target_size (two-pass to fixed file size).\n`target_size` is currently `codec=h264` only \u2014 libx265 / libvpx-vp9 / libsvtav1 each\nrequire a different two-pass invocation and are tracked as separate per-codec\nLambda follow-ups. The `per_value_depends_on` block above gates\n`encoding_mode=target_size` itself on `codec \u2208 [h264]`, so a request with\n`encoding_mode=target_size + codec \u2208 {h265, vp9, av1}` is rejected as\n`invalid_options` regardless of whether `target_size_bytes` is supplied \u2014\nclosing the gap left by the previous option-level `depends_on` (which only\ntriggered when `target_size_bytes` was present in the payload). Per ticket\n[`bsV3FWM5`](https://trello.com/c/bsV3FWM5).\n",
+              "description": "CRF aims for a quality level and lets the file size vary;\ntarget_size aims for a fixed output file size. `target_size`\ncurrently works with `codec=h264` only \u2014 a request pairing it\nwith another codec is rejected as `invalid_options`.\n",
               "per_value_depends_on": {
                 "target_size": {
                   "codec": [
@@ -2031,13 +2115,13 @@
                   "always"
                 ]
               },
-              "description": "Two-pass target file size in bytes (minimum 1 MiB). Required when\n`encoding_mode=target_size`. h264-only \u2014 the codec gate now lives on\n`encoding_mode.per_value_depends_on.target_size.codec` (per ticket\n[`bsV3FWM5`](https://trello.com/c/bsV3FWM5)); this option's own\n`depends_on` no longer duplicates it.\n\nPractical floor: a ~100 kbps post-AAC video bitrate (i.e. `target_size_bytes`\ncorresponding to less than ~100 kbps of video after subtracting the 160 kbps\nAAC audio track over the merged duration) is silently floored by libx264 and\nthe resulting file will not honour the requested `target_size_bytes`.\nPublishers targeting very small outputs should size accordingly.\n",
+              "description": "Target file size in bytes (minimum 1 MiB). Required when\n`encoding_mode=target_size`, which is h264-only.\n\nPractical floor: if the target works out to less than\n~100 kbps of video (after subtracting the ~160 kbps AAC audio\ntrack over the merged duration), the encoder floors it and the\nresulting file will not honour the requested `target_size_bytes`.\nPublishers targeting very small outputs should size accordingly.\n",
               "min": 1048576,
               "type": "integer"
             },
             "transition": {
               "default": "none",
-              "description": "Default transition between clips. Catalog from FFmpeg 7.1.1\nxfade filter (per ADR-0007). See per_value_availability for\nper-value rollout state.\n",
+              "description": "Default transition between clips. Some effects are still\nrolling out \u2014 check their availability.\n",
               "per_value_availability": {
                 "circleclose": {
                   "availability": "beta"
@@ -2296,6 +2380,16 @@
                 "wipetr",
                 "wipebl"
               ]
+            },
+            "trim_end": {
+              "description": "Trim from end in seconds (per-input). Applied to this input\nBEFORE any transition/crossfade. Matches the `compress` trim\nsemantic. No `max` \u2014 the upper bound is the input's own\nduration, enforced by the worker, not statically bounded\nhere.\n",
+              "min": 0,
+              "type": "float"
+            },
+            "trim_start": {
+              "description": "Trim from beginning in seconds (per-input). Applied to this\ninput BEFORE any transition/crossfade. Matches the\n`compress` trim semantic. No `max` \u2014 the upper bound is the\ninput's own duration, enforced by the worker, not statically\nbounded here.\n",
+              "min": 0,
+              "type": "float"
             }
           },
           "processing_class": {
@@ -2324,14 +2418,108 @@
       },
       "min_inputs": 2
     },
+    "passthrough": {
+      "availability": "beta",
+      "default": false,
+      "description": "Inert lossless source operation. A single-input source job whose\nSOLE operation is `passthrough` emits its source bytes UNCHANGED \u2014\nno compression, no transformation, no Lambda dispatch. The API\nself-completes the job at workflow-create/publish time: the job's\nterminal output IS the upload's `{bucket, key}` unchanged.\n\n**Purpose.** Lets an uploaded file enter a multi-input operation\n(`merge` / `archive` / `image_watermark` / etc.) LOSSLESSLY. Since\nV2 narrows `JobInputV2.source` to exclude upload-direct (`MultiInputSource`\n\u2014 job_output / external_import / connection only), an upload that\nmust feed a multi-input op enters via a `passthrough` source job\nreferenced downstream by `{ type: job_output, from: <id> }`. This\npreserves billing / DAG / lineage that direct upload-in-inputs[]\nwould bypass.\n\n**Distinct from `operations: []`.** An empty `operations[]` on a\nsingle-input upload job KEEPS its implicit-compress meaning (see\n`POST /api/workflows` description). `passthrough` is the EXPLICIT\nlossless opt-out chosen via the existing rule \"non-empty\n`operations[]` without `compress` = compression opt-out\" \u2014 it is a\nsole, named operation, not an empty chain.\n\n**Never published to SNS.** Because the API self-completes the job\nat publish, `passthrough` is INTENTIONALLY absent from the AsyncAPI\nrouting enums (`OperationType`, `OperationRequestAttributes.operation_type`)\nand has no `ops-passthrough` queue. It never appears on the wire as\na request message.\n\nMedia-agnostic (no `mime_groups`): it accepts any uploaded MIME and\ncopies it verbatim, so there are no per-MIME options or constraints.\n`availability: beta` \u2014 the API self-completion path is LIVE (the\ninputs[]-narrowing + passthrough self-complete mechanism is deployed),\nso workflow-create accepts `passthrough` source jobs and MUST NOT\nreturn `feature_not_available`. Activated per ticket\n[`4som89Uh`](https://trello.com/c/4som89Uh) + ADR-0004 (planned\u2192beta\nflip; self-complete code deployed API-side).\n",
+      "input_model": "single"
+    },
+    "render_variants": {
+      "availability": "planned",
+      "description": "Generate multiple resized / re-encoded image variants from a single source image in one operation (image fan-out).",
+      "input_model": "single",
+      "mime_groups": {
+        "image": {
+          "availability": "planned",
+          "mimes": [
+            "image/jpeg",
+            "image/png",
+            "image/webp"
+          ],
+          "options": {
+            "targets": {
+              "description": "One output variant per entry (max 20). Each variant is resized /\nre-encoded independently from the single decoded source; the\noperation succeeds or fails **atomically** \u2014 there are no partial\noutputs. Each result output echoes its target's `id` back as\n`target_id` so callers can map outputs to the variants they asked\nfor.\n\nDimensions: `fit: max` needs at least one of `width` / `height`\n(aspect preserved, output may be smaller than the box). `fit: crop`\nand `fit: scale` require BOTH `width` and `height` (they produce an\nexact-size output). A target whose dimensions exceed the source is\nrejected (`invalid_options`) unless `allow_upscale: true`.\n",
+              "items": {
+                "properties": {
+                  "allow_upscale": {
+                    "default": false,
+                    "description": "When false (default), a target whose dimensions exceed the source is rejected (`invalid_options`). Set true to permit upscaling. fit max never upscales.",
+                    "type": "boolean"
+                  },
+                  "background": {
+                    "default": "#FFFFFF",
+                    "description": "Background colour (hex #RRGGBB), applied only when flattening a transparent source into an opaque format (e.g. a PNG with alpha rendered to JPEG). Ignored otherwise.",
+                    "pattern": "^#[0-9A-Fa-f]{6}$",
+                    "type": "string"
+                  },
+                  "fit": {
+                    "default": "max",
+                    "description": "Resize mode. max = fit within the box, aspect preserved\n(output may be smaller); needs at least one dimension.\ncrop = fill and trim to exact width x height. scale = force\nexact width x height. crop and scale require both dimensions.\n",
+                    "enum": [
+                      "max",
+                      "crop",
+                      "scale"
+                    ],
+                    "type": "string"
+                  },
+                  "height": {
+                    "description": "Target height in pixels. Required together with width for fit crop/scale; for fit max, at least one of width/height is required.",
+                    "maximum": 16384,
+                    "minimum": 1,
+                    "type": "integer"
+                  },
+                  "id": {
+                    "description": "Caller-assigned identifier, unique within the array\n(duplicate ids are rejected as `invalid_options`). Echoed\nverbatim as `target_id` on the matching result output \u2014 the\naddressing contract.\n",
+                    "maxLength": 64,
+                    "pattern": "^[A-Za-z0-9._-]+$",
+                    "type": "string"
+                  },
+                  "output_format": {
+                    "description": "Output format for this variant. When omitted, the source format is preserved.",
+                    "enum": [
+                      "jpeg",
+                      "png",
+                      "webp",
+                      "avif"
+                    ],
+                    "type": "string"
+                  },
+                  "quality": {
+                    "default": 85,
+                    "description": "Encode quality (lossy formats only \u2014 jpeg / webp / avif). Accepted but ignored for png (lossless).",
+                    "maximum": 100,
+                    "minimum": 1,
+                    "type": "integer"
+                  },
+                  "width": {
+                    "description": "Target width in pixels. Required together with height for fit crop/scale; for fit max, at least one of width/height is required.",
+                    "maximum": 16384,
+                    "minimum": 1,
+                    "type": "integer"
+                  }
+                },
+                "required": [
+                  "id"
+                ],
+                "type": "object"
+              },
+              "maxItems": 20,
+              "minItems": 1,
+              "required": true,
+              "type": "array"
+            }
+          }
+        }
+      }
+    },
     "split": {
       "availability": "beta",
       "default": false,
-      "description": "Fan one input file into N outputs across GIF, PDF, audio, and\nvideo MIME families. Single-input \u2014 one input file per job, fanned\ninto N outputs per the per-mime-group catalog. Mirrors the\n`merge` / `convert` catalog-split-by-mime-group pattern.\n\n**Three modes** (audio + video; the mode discriminator gates which\nof the three options applies per the `depends_on` rules in this\nschema):\n- `interval`: split every N numeric-seconds.\n- `count`: split into N equal-duration pieces (integer 2..=200).\n- `cut_points`: explicit cut points in numeric-seconds (strictly\n  increasing, no duplicates, each > 0 and < probed duration).\n\n**GIF + PDF use range-based selection** instead of three modes:\n- GIF (`image_gif`): `frame_range` REQUIRED + `output_format`\n  enum {png,webp,jpg}.\n- PDF (`document_pdf`): `page_range` OR `page_groups` (mutually\n  exclusive per `depends_on`).\n\n**Wire format**: numeric seconds (floats allowed for sub-second\nprecision). NOT ISO 8601 duration strings. Matches FFmpeg /\nCloudinary / Shotstack / AWS MediaConvert conventions.\n\n**200-output hard cap** per ADR-0009 \u00a7D5\n(`OperationResult.outputs[].maxItems: 200`). Preflight rejects\nrequests exceeding this BEFORE work as `invalid_options` (422).\nCap math by mode:\n- `interval`: `ceil(probed_duration / interval) <= 200`\n- `count`: enforced by the option's `min: 2 max: 200` range\n- `cut_points`: `cut_points.len() + 1 <= 200`\n- PDF: resolved page count after expansion `<= 200`\n  (cross-ref [`WgCqnMRa`](https://trello.com/c/WgCqnMRa))\n- GIF: resolved frame count after expansion `<= 200`\n\n**Output naming**: 3-digit zero-padded \u2014 `output-001` ..\n`output-200`. Matches the convert PDF\u2192N precedent.\n\n**Output envelope binding** (per ADR-0014 + ADR-0009 \u00a7D2):\n- `image_gif`: `PositionIndexed` (frame stream-position 0-based\n  ordinal)\n- `document_pdf`: `PageIndexed` (1-based gapless page index)\n- `audio`: `PositionIndexed` (cut stream-position ordinal)\n- `video`: `PositionIndexed` (cut stream-position ordinal)\n\n**`precision` flag** (audio + video only): `fast` (default) =\npacket-boundary approximate cuts; `exact` = re-encode-aligned\nprecise cuts. Mime_group scoping is the gate \u2014 precision does\nNOT apply to GIF or PDF (frame extraction is exact and PDF page\nselection is exact by definition).\n\n**Long-form video** routes to a separate `split-video-fargate`\nLambda (same `split` OperationType, different worker; routing\nvia `processing_class`). No schema-side discriminator beyond MIME\ndetection + `processing_class`.\n\n`availability: beta` for the `audio` mime_group (worker live on\nstaging per lambdas #161 \u2014 shape-stable + opt-in, MUST NOT 422).\nThe `image_gif`, `document_pdf`, and `video` mime_groups stay\n`availability: planned` until their cross-repo Lambda workers ship\n([`vKI0CFDu`](https://trello.com/c/vKI0CFDu) + lambdas L1); dispatch\nreturns `feature_not_available` (422) for those groups until then.\n\nPer ADR-0014.\n",
+      "description": "Fan one input file into N outputs across GIF, PDF, audio, and\nvideo MIME families. Single-input \u2014 one input file per job, fanned\ninto N outputs per the per-mime-group catalog. Mirrors the\n`merge` / `convert` catalog-split-by-mime-group pattern.\n\n**Three shared modes** (audio + video; the mode discriminator\ngates which option applies per the `depends_on` rules in this\nschema):\n- `interval`: split every N numeric-seconds.\n- `count`: split into N equal-duration pieces (integer 2..=200).\n- `cut_points`: explicit cut points in numeric-seconds (strictly\n  increasing, no duplicates, each > 0 and < probed duration).\n\n**Plus an audio-only `silence` mode** (`beta`, gated by the\n`silence_mode_audio` feature): cut at detected silence gaps via\nthe worker's ffmpeg `silencedetect` filter. Video stays\nthree-mode.\n\n**GIF + PDF use range-based selection** instead of three modes:\n- GIF (`image_gif`): `frame_range` REQUIRED + `output_format`\n  enum {png,webp,jpg}.\n- PDF (`document_pdf`): `page_range` OR `page_groups` (mutually\n  exclusive per `depends_on`).\n\n**Wire format**: numeric seconds (floats allowed for sub-second\nprecision). NOT ISO 8601 duration strings. Matches FFmpeg /\nCloudinary / Shotstack / AWS MediaConvert conventions.\n\n**200-output hard cap** per ADR-0009 \u00a7D5\n(`OperationResult.outputs[].maxItems: 200`). Preflight rejects\nrequests exceeding this BEFORE work as `invalid_options` (422).\nCap math by mode:\n- `interval`: `ceil(probed_duration / interval) <= 200`\n- `count`: enforced by the option's `min: 2 max: 200` range\n- `cut_points`: `cut_points.len() + 1 <= 200`\n- `silence` (audio): resolved segment count (detected gaps + 1)\n  `<= 200`; preflight rejects as `invalid_options` (422)\n- PDF: resolved page count after expansion `<= 200`\n  (cross-ref [`WgCqnMRa`](https://trello.com/c/WgCqnMRa))\n- GIF: resolved frame count after expansion `<= 200`\n\n**Output naming**: 3-digit zero-padded \u2014 `output-001` ..\n`output-200`. Matches the convert PDF\u2192N precedent.\n\n**Output envelope binding** (per ADR-0014 + ADR-0009 \u00a7D2):\n- `image_gif`: `PositionIndexed` (frame stream-position 0-based\n  ordinal)\n- `document_pdf`: `PageIndexed` (1-based gapless page index)\n- `audio`: `PositionIndexed` (cut stream-position ordinal)\n- `video`: `PositionIndexed` (cut stream-position ordinal)\n\n**`precision` flag** (audio + video only): `fast` (default) =\npacket-boundary approximate cuts; `exact` = re-encode-aligned\nprecise cuts. Mime_group scoping is the gate \u2014 precision does\nNOT apply to GIF or PDF (frame extraction is exact and PDF page\nselection is exact by definition).\n\n**Long-form video** routes to a separate `split-video-fargate`\nLambda (same `split` OperationType, different worker; routing\nvia `processing_class`). No schema-side discriminator beyond MIME\ndetection + `processing_class`.\n\n`availability: beta` for the `audio` and `video` mime_groups\n(workers live on staging \u2014 shape-stable + opt-in, MUST NOT 422).\nVideo activates **short-form only**: `video.processing_class.short_form`\nis `beta`; `long_form` (the `split-video-fargate` worker) stays\n`availability: planned` until that worker ships. The `image_gif` and\n`document_pdf` mime_groups also stay `availability: planned` until their\ncross-repo Lambda workers ship ([`vKI0CFDu`](https://trello.com/c/vKI0CFDu)\n+ lambdas L1); dispatch returns `feature_not_available` (422) for the\nstill-planned groups/classes until then.\n\nPer ADR-0014.\n",
       "features": {
         "silence_mode_audio": {
-          "availability": "planned",
-          "description": "Silence-detect cut mode for audio (`mode: silence` \u2014\nCinavia/Resemble territory, server-side silence detection).\nOUT of V1 (Lambda support not yet shipped); declared as a\nfuture-work feature here so SDK consumers see the gate.\nPicked up as a separate post-launch ticket.\n"
+          "availability": "beta",
+          "description": "Silence-detect cut mode for audio (`mode: silence` \u2014\nserver-side silence detection in the audio split worker).\n`beta`: shape-stable + opt-in. The worker is built; the mode\nships dark until the API regenerates enum validation to accept\n`mode: silence`. The `silence` enum value carries a matching\n`per_value_availability: beta` on `audio.mode`.\n"
         }
       },
       "input_model": "single",
@@ -2383,22 +2571,57 @@
             },
             "mode": {
               "default": "interval",
-              "description": "Split strategy:\n- `interval`: cut every `interval` seconds.\n- `count`: cut into `count` equal-duration pieces.\n- `cut_points`: cut at explicit timestamps.\n",
+              "description": "Split strategy:\n- `interval`: cut every `interval` seconds.\n- `count`: cut into `count` equal-duration pieces.\n- `cut_points`: cut at explicit timestamps.\n- `silence`: cut at detected silence gaps (audio-only,\n  `beta` via the `silence_mode_audio` feature gate).\n",
+              "per_value_availability": {
+                "silence": {
+                  "availability": "beta"
+                }
+              },
               "type": "enum",
               "values": [
                 "interval",
                 "count",
-                "cut_points"
+                "cut_points",
+                "silence"
               ]
             },
             "precision": {
               "default": "fast",
-              "description": "Cut precision. Under `fast`, audio cuts are\npacket-boundary approximate (drift bounded by codec\npacket size \u2014 typically <= 23ms for AAC, <= 26ms for\nMP3). Under `exact`, the Lambda re-encodes to align cuts\nprecisely at the requested timestamps (slower, larger\noutput, but sample-accurate). Avoid \"keyframe\" language\n\u2014 that's video-only; audio drift is packet-boundary.\n",
+              "description": "Cut precision. Under `fast`, audio cuts are\npacket-boundary approximate (drift bounded by codec\npacket size \u2014 typically <= 23ms for AAC, <= 26ms for\nMP3). Under `exact`, the Lambda re-encodes to align cuts\nprecisely at the requested timestamps (slower, larger\noutput, but sample-accurate).\n",
               "type": "enum",
               "values": [
                 "fast",
                 "exact"
               ]
+            },
+            "silence_handling": {
+              "default": "keep_at_end",
+              "depends_on": {
+                "mode": "silence"
+              },
+              "description": "How detected silence is handled at segment boundaries.\n`keep_at_end`: trailing silence stays appended to the\npreceding segment (no separate silence-only outputs).\nOnly valid when `mode: silence`.\n",
+              "type": "enum",
+              "values": [
+                "keep_at_end"
+              ]
+            },
+            "silence_min_duration": {
+              "default": 0.75,
+              "depends_on": {
+                "mode": "silence"
+              },
+              "description": "Minimum duration of a silence gap, in numeric seconds\n(sub-second allowed), before it is treated as a cut point.\n`min: 0.001` is the strict-positive sub-second idiom\n(cf. `interval`). Only valid when `mode: silence`.\n",
+              "min": 0.001,
+              "type": "float"
+            },
+            "silence_threshold_db": {
+              "default": -50.0,
+              "depends_on": {
+                "mode": "silence"
+              },
+              "description": "Noise floor below which audio counts as silence, in dBFS\n(negative; 0.0 = full scale). Strictly negative in\npractice; `max: 0.0` is the declarative ceiling (a 0 dB\n\"silence\" floor is degenerate, like the strict bounds\ndocumented in prose for `cut_points`). Only valid when\n`mode: silence`.\n",
+              "max": 0.0,
+              "type": "float"
             }
           }
         },
@@ -2463,10 +2686,11 @@
           }
         },
         "video": {
-          "availability": "planned",
+          "availability": "beta",
           "mimes": [
             "video/mp4",
-            "video/webm"
+            "video/webm",
+            "video/quicktime"
           ],
           "options": {
             "count": {
@@ -2540,7 +2764,7 @@
               "required_tier": "pro"
             },
             "short_form": {
-              "availability": "planned",
+              "availability": "beta",
               "constraints": {
                 "max_input_duration": "PT5M",
                 "max_input_size_bytes": 524288000,
@@ -2621,13 +2845,13 @@
             },
             "rotation": {
               "default": -45.0,
-              "description": "Rotation angle in degrees. Default -45\u00b0 suits tiled mode\n(angled diagonal pattern). Clamped to range. Tiled canvas is\ncapped at 25 megapixels pre-rotation with 5000 px maximum\nside length.\n",
+              "description": "Rotation angle in degrees. Default -45\u00b0 suits tiled mode\n(angled diagonal pattern). Clamped to range.\n",
               "max": 360.0,
               "min": -360.0,
               "type": "float"
             },
             "text": {
-              "description": "Watermark text. Must be non-empty. Rendered with bundled\nLiberation Sans (SIL OFL). A single text overlay is capped\nat 4 megapixels before rendering fails.\n",
+              "description": "Watermark text. Must be non-empty.\n",
               "required": true,
               "type": "string"
             },
@@ -2776,7 +3000,7 @@
           "mimes": [
             "video/mp4",
             "video/webm",
-            "video/ogg"
+            "video/quicktime"
           ],
           "options": {
             "fit": {
@@ -2920,7 +3144,7 @@
               "type": "float"
             },
             "text": {
-              "description": "Watermark text. Must be non-empty. Rendered with bundled\nLiberation Sans (SIL OFL).\n",
+              "description": "Watermark text. Must be non-empty.\n",
               "required": true,
               "type": "string"
             },
@@ -2974,7 +3198,7 @@
     "video_watermark": {
       "availability": "beta",
       "default": false,
-      "description": "Apply an image overlay (file or external source) onto a base video\nasset using FFmpeg's `overlay` filter. Multi-input role-based:\nexactly one input with `role: base` (the source video) + exactly\none with `role: overlay` (the watermark image). The overlay can be\nan upload, an external_import handle, or a vault connection per\n`WorkflowSource`.\n\n**Dedicated operation, not an extension of `image_watermark`.**\n`image_watermark` is pure-Rust/no-FFmpeg and handles\nimage-on-image overlay only. `video_watermark` requires FFmpeg's\noverlay filter + re-encode + a different deploy unit (no\nstream-copy escape \u2014 overlay = pixel modification). Per ADR-0013.\n\n**Audio passthrough.** The base video's audio stream is preserved\nunchanged via FFmpeg stream-copy \u2014 the watermark applies to the\nvisual track only.\n\nActivated to `availability: beta` (Wave A, ticket\n[`c3uthIP4`](https://trello.com/c/c3uthIP4)): the cross-repo Lambda\nbackend is live (API accepts + publishes via the operations SNS\ntopic), so the operation and its `short_form` processing class are\nshape-stable + opt-in and MUST NOT return `feature_not_available`.\n`long_form` stays `planned` (no live Fargate worker yet \u2014 its\nlong-form path is not provisioned) and the `multi_overlay_stack`\nfeature stays `planned` (Lambda support unconfirmed); requesting\nthose planned sub-features DOES still return `feature_not_available`\nuntil they are separately activated. Per ADR-0013 + lambdas\npre-launch epic Wave A (ticket\n[`SlluxMBN`'s sibling `4NrRPCgh`](https://trello.com/c/4NrRPCgh)).\n",
+      "description": "Apply an image overlay onto a base video\nasset using FFmpeg's `overlay` filter. Multi-input role-based:\nexactly one input with `role: base` (the source video) + exactly\none with `role: overlay` (the watermark image). Each input is a\n`MultiInputSource` \u2014 an external_import handle, a vault connection,\nor an upstream `job_output` (uploads are NOT referenced directly\ninside inputs[]). To use an uploaded base or overlay, feed it through\na `passthrough` source job and reference it via `{ type: job_output,\nfrom: <id> }` (per ticket 4som89Uh).\n\n**Dedicated operation, not an extension of `image_watermark`.**\n`image_watermark` is pure-Rust/no-FFmpeg and handles\nimage-on-image overlay only. `video_watermark` requires FFmpeg's\noverlay filter + re-encode + a different deploy unit (no\nstream-copy escape \u2014 overlay = pixel modification). Per ADR-0013.\n\n**Audio passthrough.** The base video's audio stream is preserved\nunchanged via FFmpeg stream-copy \u2014 the watermark applies to the\nvisual track only.\n\nActivated to `availability: beta` (Wave A, ticket\n[`c3uthIP4`](https://trello.com/c/c3uthIP4)): the cross-repo Lambda\nbackend is live (API accepts + publishes via the operations SNS\ntopic), so the operation and its `short_form` processing class are\nshape-stable + opt-in and MUST NOT return `feature_not_available`.\n`long_form` stays `planned` (no live Fargate worker yet \u2014 its\nlong-form path is not provisioned) and the `multi_overlay_stack`\nfeature stays `planned` (Lambda support unconfirmed); requesting\nthose planned sub-features DOES still return `feature_not_available`\nuntil they are separately activated. Per ADR-0013 + lambdas\npre-launch epic Wave A (ticket\n[`SlluxMBN`'s sibling `4NrRPCgh`](https://trello.com/c/4NrRPCgh)).\n",
       "features": {
         "multi_overlay_stack": {
           "availability": "planned",
@@ -3071,7 +3295,7 @@
       }
     }
   },
-  "schema_version": "2.31.0",
+  "schema_version": "2.64.0",
   "source_commit": null,
   "user_tier": null,
   "workflow_features": {
@@ -3093,10 +3317,10 @@
         "type": {
           "per_value_availability": {
             "all_outputs": {
-              "availability": "planned"
+              "availability": "stable"
             },
             "explicit": {
-              "availability": "planned"
+              "availability": "stable"
             },
             "terminal": {
               "availability": "stable"