Another Tool for the Creative Process

01.Set the rules once

Two things shape every translation before a single line is drafted, and your team sets both from the console or the API, per creator, per stream or per video:

• The Do Not Translate list.Catchphrases, names, brands and slang stay exactly as said, in every language: the host's signature exclamation, the friends' nicknames, the game characters.
• The context brief. One sentence on what the channel, stream or video is: who is on screen and what kind of video. A cooking show and a courtroom drama translate the same sentence differently.

The narrower setting wins: a rule on one video overrides the creator's default for that video only.

02.Pick the line that lands

Some lines can land more than one way. A pun, a jab, a sentence that reads as sincere or sarcastic depending on the actor's face. For those, the translation comes back as three options, each tagged with its intent and emotion: the pun kept and sheepish, the plain meaning and sincere, a new joke and playful. Your creative director picks one. The tool never decides what a character meant.

03.Edit against the original

Every translated line sits beside its source line, so a reviewer who speaks both languages can rewrite any of them before anything renders. In the console, every video pauses at that review screen and nothing renders until your team approves it; teams that trust the flow can skip the stop. Through the API the same gate is one setting, and a dub can be released language by language as each reviewer clears it.

04.The files that come back

Each language arrives ready for your edit bay or your upload queue:

• The finished video, up to 4K.
• The full mix as true 5.1 surround or stereo, your pick, in 24-bit PCM WAV, plus the voice alone for your own mix.
• Subtitles in .srt and .vtt, built from the reviewed transcript, so they match what the dub says.
• The translated title and description, with links, @handles and promo codes left as written.

The dub is yours to post anywhere.

05.Who decides what

Your team keeps every creative decision: the names and phrases that never change, the brief, the pick between three readings of a joke, any line rewritten by hand, the moment a language goes out. The tool takes the work no one wants to do thirty times over: drafting each line in every language, voicing it in the speaker's own voice, lip-syncing the mouth to the new words, carrying the laughter, music and room sound through, and building the subtitles and metadata.

06.Machine learning for the tedious part.

Machine learning is good at repetition: the same voice, the same timing, the same face, held steady line after line and language after language. That is where Familiar uses it. Judgment about tone, humor and meaning stays with the people who made the work, and the review screen is where they apply it.

All of the data we trained with ourselves are actually generated by the Familiar team.

07.Not generative AI, but translation to enable creatives.

There is no prompt. Nothing new is made up. Generative AI in the traditional sense creates new content from a prompt, which is a fine way to make new things. A dub needs the opposite: your video goes in, and the same actor, the same performance and the same timing come out in another language. Familiar is a multimodal translation tool.

QUESTIONS, ANSWERED

REFERENCES

1. [1]Familiar FAQ
2. [2]Familiar API: creators, Do Not Translate, context
3. [3]Familiar API: context brief
4. [4]Familiar API: Do Not Translate
5. [5]Familiar API: transcript and review
6. [6]Familiar API: outputs
7. [7]Familiar API: translated titles and descriptions