Conformance

Validated against 164 real-world encode cases, continuously.

Every claim on this page is generated from Transcodely's validation catalog — a suite that submits real jobs to real workers and asserts on the produced media, manifests, DRM and HDR signalling, and pricing. Of the 164 cases, 106 drive an encode to completion; the rest verify validation boundaries, pricing, and job lifecycle.

164

validation cases

106

real encodes to completion

34 / 40

capabilities validated

23

coverage categories

What every case actually checks

A green case is not "the API returned 200". It means a job ran end to end and the output held up under inspection.

Real jobs, real workers

Each case provisions real object storage (Cloudflare R2), submits through the public API exactly as documented, and runs on real Hetzner worker VMs. No mocks, no stubs.

Media, probed

Produced files are downloaded and inspected with ffprobe — codec, profile, pixel format, resolution, framerate, and colour signalling are checked against what was requested.

Manifests, parsed

HLS and DASH manifests are fetched and parsed for variant count, audio and subtitle renditions, and encryption signalling — proving a stream was produced, not merely priced.

Pricing, recomputed

The locked pricing snapshot returned on each output is recomputed from its components (codec, resolution, framerate, quality, feature, minimum) and checked for internal consistency.

Artifacts, counted

Thumbnails, sprite WebVTT tracks, and subtitle sidecars are listed from storage and counted against what the job asked for.

Honest about gaps

When the worker cannot yet produce a feature, the case keeps its true assertion and is tracked as a gap — reported, never quietly passed. Those gaps are listed in full below.

Coverage by category

Cases are grouped by what they exercise. "Validated" counts the cases the suite expects to pass today; anything in progress is called out separately.

Input validation (rejections)

42/42

Codec × container

11/11

Codec settings

10/10

HDR

8/9

1 tracked, in progress

Input formats

8/8

Pricing

8/8

Thumbnails

8/8

Quality (VMAF)

7/7

Streaming (HLS / DASH)

7/7

Adaptive bitrate

6/6

Presets

6/6

Subtitles

6/6

Audio

5/5

Path templates

5/5

Watermark / overlay

5/5

Captions

3/4

1 tracked, in progress

DRM

4/4

Job lifecycle

4/4

Clip

3/3

Encode-failure handling

2/2

Resolution

2/2

Encoding pipeline

1/1

Multi-output jobs

1/1

Capabilities exercised

Each value below is exactly what the API accepts on the wire. A value is validated only when a real job produces it and the harness asserts on the output. Values still in progress are not yet validated end to end — some are accepted-but-unproduced, others are gated at create.

Video codecs

H.264 H.265 (HEVC) VP9 AV1

Containers & packaging

MP4 WebM MKV MOV HLS DASH Adaptive (HLS + DASH)

DRM systems

Widevine FairPlay PlayReady

Encryption schemes

CENC CBCS

HDR formats

HDR10 HDR10+ HLG Dolby Vision (profile 5) · in progress Dolby Vision (profile 8) · in progress

HDR modes

Passthrough Tone-map Force

Subtitle operations

Passthrough Convert Burn-in Extract generate

Subtitle formats

SRT WebVTT TTML · in progress ASS/SSA · in progress

Thumbnail modes

Single frame Interval Sprite sheet Timestamps animated

Content-aware encoding

Per-title · in progress Auto-ABR · in progress

In progress — tracked, not yet validated

These features are accepted and priced by the API, but the encoder does not yet produce them end to end. Each still has a validation case that asserts the true intended behaviour — it stays here, counted honestly, until it passes.

HDR10+ force on H.265 — output must carry PQ/BT.2020 10-bit AND dynamic HDR10+ metadata

HDR10+ dynamic metadata (SMPTE ST 2094-40) is unimplemented: the worker encodes hdr10_plus through the same static-HDR10 branch as hdr10, so the output carries no per-scene metadata

Retro-caption a hosted video — sidecar under the video prefix, attached with in_manifest=false

retro-caption input: worker cannot read an HLS manifest input, and hosted videos keep no single-file source

Why you can trust these numbers

This page is not hand-written. It is generated from the API's validation catalog and synced into the site as a snapshot; the same catalog is the production-readiness gate for the encoding pipeline. When a capability's coverage changes, the number here changes with it — and a drift check fails the build if the page and the catalog ever disagree.