Media operators

The Media family contains 5 operators. Image / audio / video generation, transcription, OCR.

Operators in this family

• media.generate
• media.ocr
• media.process
• media.transcribe
• media.transform

media.generate

Generates an image from a text prompt via OpenRouter (or any OpenAI-compatible image endpoint). Persists the image to tenant storage.

Version 1.0.0 · Semantics side_effect · Secrets OPENROUTER_API_KEY

Config

Field	Type	Required	Default	Description
model	string	no	—	Image model id (e.g. ‘openai/dall-e-3’).
prompt	string	yes	—	Text prompt describing the image to generate.
quality	”standard” | “hd”	no	"standard"
return_b64	boolean	no	false	Also include the base64-encoded image bytes inline in the output.
size	”256x256” | “512x512” | “1024x1024” | “1024x1792” | “1792x1024”	no	"1024x1024"

Input ports

Port	Type	Required	Description
input	any	no	Input

Output ports

Port	Type	Required	Description
output	object	no	Image

media.ocr

Extracts text from an image using a vision-capable LLM. Counts against the tenant AI budget.

Version 1.0.0 · Semantics side_effect · Visibility advanced · Secrets OPENROUTER_API_KEY

Config

Field	Type	Required	Default	Description
lang	string	no	—	Optional ISO-639 language hint for the OCR model.
model	string	no	—	Vision-capable chat model id (e.g. ‘openai/gpt-4o-mini’).
prompt	string	no	—	Override the instruction sent to the vision model.
source	”url” | “storage”	yes	—
storage_path	string	no	—	Tenant storage path (required when source=storage).
url	string	no	—	Source URL (required when source=url). Subject to SSRF guard.

Input ports

Port	Type	Required	Description
input	any	no	Input

Output ports

Port	Type	Required	Description
output	object	no	Text

media.process

Processes images, audio, or video (resize, transcribe, extract frames).

Version 1.0.0 · Semantics passthrough · Visibility advanced

Config

Field	Type	Required	Default	Description
operation	”resize” | “transcribe” | “extract_frames” | “thumbnail”	yes	"resize"
options	object	no	—

Input ports

Port	Type	Required	Description
input	any	yes	Media

Output ports

Port	Type	Required	Description
output	any	no	Output

media.transcribe

Transcribes audio to text via a Whisper-compatible endpoint. Counts against the tenant AI budget.

Version 1.0.0 · Semantics side_effect · Secrets OPENROUTER_API_KEY

Config

Field	Type	Required	Default	Description
filename	string	no	—	Multipart filename hint (e.g. ‘audio.mp3’). Some providers require an extension to detect format.
language	string	no	—	Optional ISO-639 language hint for the transcription model.
model	string	no	—	Whisper-compatible model id (e.g. ‘openai/whisper-1’).
response_format	”text” | “json” | “verbose_json”	no	"verbose_json"
source	”url” | “storage”	yes	—
storage_path	string	no	—	Tenant storage path (required when source=storage).
url	string	no	—	Source URL (required when source=url). Subject to SSRF guard.

Input ports

Port	Type	Required	Description
input	any	no	Input

Output ports

Port	Type	Required	Description
output	object	no	Transcript

media.transform

Applies local image transformations (resize, re-format, strip metadata). No external API call.

Version 1.0.0 · Semantics side_effect · Idempotent

Config

Field	Type	Required	Default	Description
ops	object[]	yes	—
source	”url” | “storage”	yes	—	Where to read the source image from.
storage_path	string	no	—	Tenant storage path (required when source=storage).
url	string	no	—	Source URL (required when source=url). Subject to SSRF guard.

Input ports

Port	Type	Required	Description
input	any	no	Input

Output ports

Port	Type	Required	Description
output	object	no	Image

Source schemas

Every operator above is rendered from its JSON schema in /opt/glyph/schemas/operators/:

• schemas/operators/media.generate.json
• schemas/operators/media.ocr.json
• schemas/operators/media.process.json
• schemas/operators/media.transcribe.json
• schemas/operators/media.transform.json