Audio decisions

01.What the model decides in audio

The model hears everything in the source: the laughter, the crowd, a shout, a speaker no one can make out. It finds the main voice and keeps the rest as it was.

The decision is which voices are translated and what stays in the source language. Each one lands in the decision record with type: "audio".

02.The kinds

TRANSLATE

Whether a voice is translated at all: the main voice, a line off mic, a background speaker no one can make out. The main voice is translated; the background speaker is left as it was.

KEEP_ORIGINAL

A chant, a yell or a phrase that stays in the source language. The stadium chant at 41.2 to 44.0 s stays in Korean, so does the yell across the kitchen and the catchphrase the fans know.

PRESERVE

A sound kept exactly as it was. The laughter and the crowd come through untouched in every language.

03.In the record

The decisions resource (in the works) filters to audio with type=audio. The chant, as the record holds it:

{
  "dub_id": "dub_31xk9m",
  "lang": "en",
  "decisions": [
    {
      "id": "dec_7h2k0b",
      "type": "audio",
      "kind": "keep_original",
      "segment": null,
      "start": 41.20,
      "end": 44.05,
      "speaker": null,
      "source": "crowd chant",
      "options": [
        { "id": "opt_1", "text": "keep in Korean", "reasoning": "A stadium chant; the crowd is the scene, not a speaker." },
        { "id": "opt_2", "text": "translate", "reasoning": "The words carry the next line's joke." }
      ],
      "chosen": "opt_1",
      "chosen_by": "model",
      "status": "open",
      "memory": null
    }
  ],
  "has_more": false
}

04.Override

The override (in the works) takes exactly one of keep_original: true or choose. Add remember: "creator" to carry the choice to the next episode.

curl -X PATCH https://api.thefamiliarlab.com/v1/dubs/dub_31xk9m/decisions/dec_7h2k0b \
  -H "Authorization: Bearer $FAMILIAR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{ "keep_original": true, "remember": "creator" }'

curl -X PATCH https://api.thefamiliarlab.com/v1/dubs/dub_31xk9m/decisions/dec_7h2k0b \
  -H "Authorization: Bearer $FAMILIAR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{ "choose": "opt_2" }'

{
  "id": "dec_7h2k0b",
  "type": "audio",
  "kind": "keep_original",
  "chosen": "opt_2",
  "chosen_by": "you",
  "status": "settled",
  "memory": null
}

PATCH works at in_review and at complete, then POST /render again. During rendering it returns 409 decision_locked; an unknown option returns 422 option_not_found.

05.Settings

Two request fields on POST /v1/dubs (in the works) set the defaults. Nested fields are written dotted in multipart.

ask requires review=manual; on an auto dub it returns 422 decision_setting_requires_review. The dub waits at in_review until every open decision is settled or you render.

06.On the review screen

The review screen (in the works) shows each audio decision as the sound's span on the episode timeline, with the options and the reasoning. One tap keeps it in the original; another remembers the choice for this creator.

07.Access