space.remanso.recording

apoena.dev

Documentation

An audio recording — a past stream, a podcast episode, a voice note. A recording at rkey R is the recording of the space.remanso.note at rkey R in the same repo: one note, at most one recording, paired by convention rather than by a field, so a reader that knows the note's did and rkey reaches the audio with a single getRecord. A recording whose rkey matches no note stands alone and belongs to nobody. Legacy path: a recording is also referenced from a note by at-uri embedded in the note's content.

main record

An audio recording — a past stream, a podcast episode, a voice note. A recording at rkey R is the recording of the space.remanso.note at rkey R in the same repo: one note, at most one recording, paired by convention rather than by a field, so a reader that knows the note's did and rkey reaches the audio with a single getRecord. A recording whose rkey matches no note stands alone and belongs to nobody. Legacy path: a recording is also referenced from a note by at-uri embedded in the note's content.

Record Key tid Timestamp-based ID

Properties

audio blob Required

The recording itself. video/mp4 and video/webm are accepted alongside audio/* because a PDS sniffs the container rather than trusting the upload's Content-Type: an audio-only MP4 or M4A sniffs as video/mp4, and a sniffer that reads no further than the EBML magic calls an audio-only WebM video/webm.

maxSize: 50.0 MB
createdAt string datetime Required

An RFC 3339 formatted timestamp.

credits array of ref #credit Optional

Attribution for openly licensed music mixed into the recording. Only licences that require it are listed — a CC0 track is deliberately absent, because CC0 asks for nothing and an entry here would imply otherwise.

maxLength: 64 items
durationSec integer Optional

Playback length in seconds.

language string Optional

The language spoken in the recording. ISO 639-3, the same code space as space.remanso.note's `language`, so a note and its recording agree on how a language is spelled. A recording whose speech was never identified simply omits this.

maxLength: 10 bytes
Known values: afr, ara, aze, bel, ben, bul, cat, ces, ckb, cmn, dan, deu, ell, eng, est, eus, fin, fra, hau, heb, hin, hrv, hun, hye, ind, isl, ita, jpn, kat, kaz, kor, lit, mar, mkd, nld, nob, pes, pol, por, ron, run, rus, slk, spa, srp, swe, tgl, tur, ukr, vie
recordedAt string datetime Optional

When the audio was originally captured — a past stream's air date, not the upload time.

title string Optional

No description available.

maxLength: 1000 bytes
transcript ref #transcript Optional

No description available.

View raw schema 
{
  "key": "tid",
  "type": "record",
  "record": {
    "type": "object",
    "required": [
      "audio",
      "createdAt"
    ],
    "properties": {
      "audio": {
        "type": "blob",
        "accept": [
          "audio/*",
          "video/mp4",
          "video/webm"
        ],
        "maxSize": 50000000,
        "description": "The recording itself. video/mp4 and video/webm are accepted alongside audio/* because a PDS sniffs the container rather than trusting the upload's Content-Type: an audio-only MP4 or M4A sniffs as video/mp4, and a sniffer that reads no further than the EBML magic calls an audio-only WebM video/webm."
      },
      "title": {
        "type": "string",
        "maxLength": 1000
      },
      "credits": {
        "type": "array",
        "items": {
          "ref": "#credit",
          "type": "ref"
        },
        "maxLength": 64,
        "description": "Attribution for openly licensed music mixed into the recording. Only licences that require it are listed — a CC0 track is deliberately absent, because CC0 asks for nothing and an entry here would imply otherwise."
      },
      "language": {
        "type": "string",
        "maxLength": 10,
        "description": "The language spoken in the recording. ISO 639-3, the same code space as space.remanso.note's `language`, so a note and its recording agree on how a language is spelled. A recording whose speech was never identified simply omits this.",
        "knownValues": [
          "afr",
          "ara",
          "aze",
          "bel",
          "ben",
          "bul",
          "cat",
          "ces",
          "ckb",
          "cmn",
          "dan",
          "deu",
          "ell",
          "eng",
          "est",
          "eus",
          "fin",
          "fra",
          "hau",
          "heb",
          "hin",
          "hrv",
          "hun",
          "hye",
          "ind",
          "isl",
          "ita",
          "jpn",
          "kat",
          "kaz",
          "kor",
          "lit",
          "mar",
          "mkd",
          "nld",
          "nob",
          "pes",
          "pol",
          "por",
          "ron",
          "run",
          "rus",
          "slk",
          "spa",
          "srp",
          "swe",
          "tgl",
          "tur",
          "ukr",
          "vie"
        ]
      },
      "createdAt": {
        "type": "string",
        "format": "datetime"
      },
      "recordedAt": {
        "type": "string",
        "format": "datetime",
        "description": "When the audio was originally captured — a past stream's air date, not the upload time."
      },
      "transcript": {
        "ref": "#transcript",
        "type": "ref"
      },
      "durationSec": {
        "type": "integer",
        "description": "Playback length in seconds."
      }
    }
  },
  "description": "An audio recording — a past stream, a podcast episode, a voice note. A recording at rkey R is the recording of the space.remanso.note at rkey R in the same repo: one note, at most one recording, paired by convention rather than by a field, so a reader that knows the note's did and rkey reaches the audio with a single getRecord. A recording whose rkey matches no note stands alone and belongs to nobody. Legacy path: a recording is also referenced from a note by at-uri embedded in the note's content."
}
credit object

One openly licensed work used in the recording, and where its terms live.

Properties

creator string Required

No description available.

maxLength: 1000 bytes
license string Required

The licence's short id, e.g. "by".

maxLength: 64 bytes
licenseUrl string uri Required

A valid URI.

sourceUrl string uri Required

The work's landing page at its provider.

title string Required

No description available.

maxLength: 1000 bytes
View raw schema 
{
  "type": "object",
  "required": [
    "title",
    "creator",
    "license",
    "licenseUrl",
    "sourceUrl"
  ],
  "properties": {
    "title": {
      "type": "string",
      "maxLength": 1000
    },
    "creator": {
      "type": "string",
      "maxLength": 1000
    },
    "license": {
      "type": "string",
      "maxLength": 64,
      "description": "The licence's short id, e.g. \"by\"."
    },
    "sourceUrl": {
      "type": "string",
      "format": "uri",
      "description": "The work's landing page at its provider."
    },
    "licenseUrl": {
      "type": "string",
      "format": "uri"
    }
  },
  "description": "One openly licensed work used in the recording, and where its terms live."
}
transcript object

What was said, as text. A recording is published before anything has transcribed it, so audio with no transcript is the normal starting state rather than an incomplete record — which is why this is optional and written by a later putRecord. `text` is the whole transcript in reading order and is always present; `segments` is that same text cut into timestamped spans. The two never disagree: a client renders either one without reconciling them, and a writer that edits one rewrites both.

Properties

createdAt string datetime Required

When this transcript was written, which is later than the recording's own createdAt, and is bumped on every correction.

segments array of ref#transcriptSegment Optional

The transcript cut into timestamped spans, one paragraph each — the granularity a reader scans and seeks by, not the two-second slivers a speech model emits. Absent when the audio yielded no usable timestamps, which happens on very short clips.

maxLength: 4096 items
text string Required

The transcript in reading order, paragraphs separated by a blank line. Equal to `segments` joined the same way, when segments are present.

maxLength: 100000 bytes
View raw schema 
{
  "type": "object",
  "required": [
    "text",
    "createdAt"
  ],
  "properties": {
    "text": {
      "type": "string",
      "maxLength": 100000,
      "description": "The transcript in reading order, paragraphs separated by a blank line. Equal to `segments` joined the same way, when segments are present."
    },
    "segments": {
      "type": "array",
      "items": {
        "ref": "#transcriptSegment",
        "type": "ref"
      },
      "maxLength": 4096,
      "description": "The transcript cut into timestamped spans, one paragraph each — the granularity a reader scans and seeks by, not the two-second slivers a speech model emits. Absent when the audio yielded no usable timestamps, which happens on very short clips."
    },
    "createdAt": {
      "type": "string",
      "format": "datetime",
      "description": "When this transcript was written, which is later than the recording's own createdAt, and is bumped on every correction."
    }
  },
  "description": "What was said, as text. A recording is published before anything has transcribed it, so audio with no transcript is the normal starting state rather than an incomplete record — which is why this is optional and written by a later putRecord. `text` is the whole transcript in reading order and is always present; `segments` is that same text cut into timestamped spans. The two never disagree: a client renders either one without reconciling them, and a writer that edits one rewrites both."
}
transcriptSegment object

One timestamped span of the transcript. Offsets are milliseconds from the start of the audio: the lexicon data model has no float, and rounding to whole seconds would land a seek target after the words the timestamp labels.

Properties

endMs integer Required

Offset from the start of the audio, in milliseconds. Never before `startMs`.

minimum: 0
startMs integer Required

Offset from the start of the audio, in milliseconds.

minimum: 0
text string Required

No description available.

maxLength: 10000 bytes
View raw schema 
{
  "type": "object",
  "required": [
    "startMs",
    "endMs",
    "text"
  ],
  "properties": {
    "text": {
      "type": "string",
      "maxLength": 10000
    },
    "endMs": {
      "type": "integer",
      "minimum": 0,
      "description": "Offset from the start of the audio, in milliseconds. Never before `startMs`."
    },
    "startMs": {
      "type": "integer",
      "minimum": 0,
      "description": "Offset from the start of the audio, in milliseconds."
    }
  },
  "description": "One timestamped span of the transcript. Offsets are milliseconds from the start of the audio: the lexicon data model has no float, and rounding to whole seconds would land a seek target after the words the timestamp labels."
}

Lexicon Garden

@