AI with a budget and a record.
Image and video generation is being built inside the production: many models through one gate, a ceiling on every call, a budget for every person, and provenance on every asset.
One gate, every generation
- Studio budget01
- Person’s budget02
- Ceiling03
- Route04
- Queue05
- Meter06
- Provenance07
One gate, in order
No surface in the product calls a model directly. Every generation will pass through the same seven checks.
- 1
The studio’s budget
Is there credit, and has the studio’s own limit been reached?
- 2
The person’s budget
Has this person reached the limit set for them, or for their seat class?
- 3
The ceiling
What is the most this call could cost? Above the ceiling, the person sees the number and confirms.
- 4
The route
Which model does this job, chosen by capability — replaceable, never locked in.
- 5
The queue
The work runs on the provider’s machines; Genreline submits and receives.
- 6
The meter
The cost is recorded against the person and the production.
- 7
Provenance
Which model made the asset, from what, and under what terms — written onto the asset.
What already works
The first three checks govern text today, inside the assistant.
- A per-call ceiling on AI spend, with confirmation above it — for text AIAvailable
- Per-member and per-seat-class budgets, each visible to its holderAvailable
- PrimeOS — the production assistant inside the workAvailable
- Rights per asset — releases write rights; clients read disclosuresAvailable
- Clearance panel — disclosure obligations for synthetic performers, read to the client who bears themAvailable
Being built
- The Stage — image and video generation, takes, versions, cost per resultComing
- Storyboard frame generation and animaticsComing
- Continuity — character, location, wardrobe and style held across every shotComing
- Model Arena — the same prompt through several models, side by sideComing
- Studio Kits — your studio’s looks, characters and presets, reusableComing
- Remaster — upscale and restore existing footageComing
- Translation with lip-syncComing
- Aspect-ratio conversion at any lengthComing
- Variant output — cutdowns, captions, localisationsComing
What Genreline will not claim
Genreline does not train models. It routes work to the models a studio chooses.
So it will not tell you that any model is “commercially safe” — that depends on the model and its terms, not on us — and it offers no IP indemnity for what a model produces.
What it will give you instead is the record: for every generated asset, which model made it, from what, and under what terms. And for every likeness, the signed release that says whether it may be used at all.
Everything being built is on the roadmap.
Open a studio, invite your team, add your first client. The record starts with the first approval.