A 90-second client spot used to cost my production team five figures and six weeks: two shoot days, a five-person crew, three weeks in the edit. The camera time was the cheap part. The ten handoffs between script and final cut were the expensive part. That gap, between one good clip and one finished video, is where AI video work lives in 2026. This year the pipeline for it arrived.
What got cheap, and what didn't
In the July 2026 head-to-head on this blog, the frontier trio, Kling 3.0, Veo 3.1, and Sora 2, was already generating sound in the same pass as the pixels, with native 4K on one of them. Which one wins a scene still reshuffles monthly; that fight stays in the comparison piece. None of them carries anything between shots. A model ships a clip. It does not ship your episode.
How the pipeline is built now
On x-rush.ai the Video module splits into three tools. Video Studio makes single clips: text-to-video, from one still image, or AI edits of footage you already have. Meme Video covers short-form template work. AI Drama is the series desk where that script-to-screen checklist lives now. One Series Workspace holds one series; a single-episode short is the same object with one episode, so adding episode 2 migrates nothing. Six stages per episode, each ending on your confirmation:
- Story and script: episode goal, outline, scene-structured script; a full draft is one 0.5-credit text unit
- Shot planning: a structured shot list from framing to duration, no storyboard images required
- Characters and wardrobe: versions bound per episode, reference images at 1 credit each
- Scenes and props: the same binding, plus day/night and weather variants
- Shot production: reuse a result, upload footage, generate from text, or start from one confirmed image
- Sound and assembly: dialogue, narration, music, effects, and subtitles over the adopted clips
No stage offers platform templates; you start from an idea, a pasted script, or an empty desk.
What a generation unit costs
Prices are fixed per unit, shown before you confirm: text 0.5 credits, an image 1, audio 2, music 5, video 10. The Free plan refills 5 credits a day, less than one video unit, so Free is for planning and rehearsal. Daily shooting starts on Lite at $15 a month, and concurrent units cap at 1 on Free, 2 on Lite, 3 on Pro.
What it still gets wrong
Generated dialogue does not move mouths; lip-sync does not exist here, so a line lands as edited sound over a face that never said it. Assembly joins clips with hard cuts, no transitions, no mid-clip splitting. Whether a face stayed consistent across forty shots is your review call: candidates wait, you adopt or reject, and nothing is auto-adopted or graded. Frame-exact product shots are still out of reach.
Where this goes next
Clips will get longer, sound will arrive with more of them, and the three-name frontier will reshuffle before this page's January review. The six stages are the stable part; models swap out underneath. If you have a pilot script in a drawer, stage one is a paste box, and pasting is free.