ABSTRACT
01.Two terms, plainly
Machine learning is the broad field. IBM describes it as algorithms that learn the patterns of training data and then make accurate inferences about new data. Show a model enough examples of an input and its right output, and it learns the mapping between them.
Generative AI is one use of machine learning. IBM defines it as artificial intelligence that can create original content, such as text, images, video, audio or software code, in response to a user's prompt or request. It is a fine tool for making new things: a concept frame, a first draft, a song sketch.
Translation asks for something else. The work already exists, and the job is to carry it into another language intact.
| Generative AI | Translation with Familiar | |
|---|---|---|
| Input | A prompt | Your finished video |
| What is new | The content itself | The words in another language, and the voice and mouth that carry them |
| Whose likeness | Whoever the prompt describes | The actor on screen, with consent |
| Review | Varies by tool | Your creative directors, line by line, before the render |
| Output | New text, images, video, audio or code | The same performance, up to 4K, in true 5.1 or stereo |
02.Six questions to ask any vendor
Is there a prompt?
No. Your video goes in, and the same video comes out in another language.
Is anything invented?
No. The words are a translation of what was said, and nothing new is made up. The scene stays as it was: the laughter, the music, the ambience.
Whose voice and face are in the output, and with whose consent?
The actor's own. The voice comes from the clip itself, and the face on screen is the actor's, performing the new language. Familiar only translates people with their consent.
What data was the model trained on?
All of the data we trained with ourselves are actually generated by the Familiar team.
Who reviews before release?
Your creative directors. A line that can land more than one way, a joke or a jab, comes back as three translations with different intent and emotion; they pick one, edit any line against the original, and only then trigger the dub. A Do Not Translate list keeps names and catchphrases exactly as said.
What files come back?
Each language arrives as a finished video up to 4K, with the full mix in true 5.1 surround or stereo, your pick, in 24-bit PCM WAV. Alongside it: subtitles in .srt and .vtt from the reviewed transcript, the translated title and description, and the voice track alone for your own mix.
03.Machine learning for the tedious part.
On Spider-Man: Into the Spider-Verse, Sony Pictures Imageworks artists hand-drew ink lines for each angle and expression of a character, and a machine learning model learned from those drawings to give animators a first-pass prediction. Artists nudged each prediction into place, and their corrections trained the next round, according to effects supervisor Pav Grochola's account on fxguide.
Dubbing's tedious part is the translation, the voice and the lip-sync, for every line in every language. Familiar does that part. Casting, performance and every creative call stay with the people who made the work.
We're another tool for the creative process.
QUESTIONS
Is Familiar generative AI?
No. Generative AI in the traditional sense makes new content from a prompt. Familiar has no prompt and invents nothing: your video goes in, the same performance comes out in another language.
Can our creative team change a line before it renders?
Yes. Every translated line waits for your creative directors, who can edit it against the original or pick among three translations where a line reads more than one way. The dub renders only after that.
Which languages does Familiar cover?
Dubs come out in 30 languages, any to any. Footage can come in from 99 source languages.
REFERENCES
- [1]IBM, What is machine learning?, read October 24, 2026
- [2]IBM, What is generative AI?, read October 24, 2026
- [3]fxguide, Ink Lines and Machine Learning (Pav Grochola, Sony Pictures Imageworks), May 20, 2019
- [4]Why Familiar Is Not Traditional Generative AI
- [5]What Spider-Verse Got Right About Machine Learning
- [6]API docs: transcript and review
Multimodal Translation
They're a team of 8 with PhDs, backed by Samsung, RedBird, Gold House. Our mission
FAMILIAR · THE LAUNCH FILM · 2:31