Production

Premiere

Jalaran Premiere assembles the publication package for a finished video — titles, description, thumbnail and captions — from assets the rest of the pipeline already produced.

Open Premiere →Create free account

Who Premiere is for

For creators who have finished the video and now face the least enjoyable hour of the process. Particularly if you publish a series, where the thumbnail has to look like the same character every time and usually does not.

What Premiere does

Premiere generates title and description options, composites a thumbnail from the character’s own locked reference image, and derives captions from the phoneme timeline the audio carries. The three parts are independent, not a pipeline — a package with text and a thumbnail but no captions is a normal finished state rather than a broken one. The thumbnail uses image compositing rather than a generative model on purpose: a freshly generated face per thumbnail is the character-drift problem Lore exists to prevent, reintroduced in a different asset.

AI title and description variants in the character’s voice
Thumbnail composited from the locked reference image, never regenerated
SRT captions derived from a real phoneme timeline
Independent sub-steps — a partial package is a valid end state

How Premiere works

Create a package
One record for one video. The three sub-steps below run independently and in any order.
Generate the text
Title variants, a description, and short overlay text for the thumbnail — clamped to a length that stays legible at feed size.
Composite the thumbnail
Built from the character’s locked reference image plus overlay text. No generative model, so the character cannot drift.
Derive captions
Captions come from the audio’s own phoneme timeline, so they are the actual words rather than an approximation.

What Premiere does not do

Captions require audio that carries a phoneme timeline, which in practice means audio from Voice. There is deliberately no waveform fallback: an approximate transcript is not a degraded caption, it is a wrong one, and a viewer has no way to tell. The thumbnail is a composite of your own image and text — it will not invent a scene. Premiere produces assets and stops there; uploading is a separate, deliberate step handled by Broadcast.

Common questions

Why will it not generate captions from any audio file?

Because captions are graded on being the actual words. Puppet can estimate mouth shapes from a waveform because an approximate mouth still beats a frozen face, but an approximate caption is simply incorrect and looks authoritative. So audio without a timeline is an error here rather than a guess.

Does the thumbnail use AI image generation?

No. It composites your character’s own locked reference image with overlay text using ordinary image processing. Generating a fresh face each time would reintroduce exactly the character drift the pipeline is built to prevent.

Does Premiere publish the video?

No. It produces the package. Uploading is Broadcast’s job, and keeping them separate means assembling a package never has a side effect on a real account.