Skip to content
JBIStudios
Book a call

Resources · White paper & articles

Sharp thinking on AI localization.

The white paper, plus working notes from the studio floor — what AI actually does well, where it breaks, and how the hybrid workflow holds broadcast spec.

White paper

White paper · 2026 edition

Nobody can prove their AI voices sound human

The industry’s own quality metric has been taken apart in peer-reviewed work. Everyone still quotes it. If you buy voice, you are being sold a number that does not measure the thing you are paying for — read this before your next vendor call.

  • Why the industry's quality score cannot see the thing you are buying
  • Casting instead of cloning — and what that changes about rights
  • What emotion costs, and how one-pass pricing hides it
  • What to ask a voice vendor before you sign

We’ll email you about the white paper and occasional studio notes. No spam, unsubscribe anytime.

A VU meter needle reading high on a completely blank dial

Articles

A reel of audio stamped before it leaves the desk as the vendor walks away

Compliance

If It Was Made by AI, You Have to Be Able to Show It: Disclosure and Provenance

The EU AI Act's transparency duties apply from 2 August 2026 and land on the party that publishes. After-the-fact detection does not work, so the only workable answer is recording provenance while the work happens.

Read the article

A long paper trail of stamped seals unrolled beside a studio microphone

Compliance

Cloning a Voice: Consent, Licensing, Compensation, Disclosure

Cloning a voice is fast. The consent, the reuse terms, and the records that prove them are the slow part, and the exposure lands on the brand that publishes. Here is what to check before you sign.

Read the article

Film reels sorted by hand into three colour-coded bins

Localization

Subtitle, Human Dub, AI Dub, Live: The 2026 Decision Grid

The choice is no longer subtitle or dub. Four delivery modes, sorted by latency and by stakes, with one sorting question: what does an error cost once it is published?

Read the article

A reviewer pauses over one line on a page of otherwise immaculate text

Localization

Machine Translation and Post-Editing: The Review Became the Job

The machine made the first draft free. What you buy now is the review that catches what the draft gets confidently wrong. The terms, the 2026 state of the art, and the four error types with a method for each.

Read the article

Two staffed stations lit among six robot stations on one production line

AI Studio

Who Steers the Machine: Where the Line Sits, and What Breaks When Nobody Holds It

AI drafts the translation, the timing, and the voice. A named person decides when that is good enough and answers for it. Here is the workflow, the pricing shift, three documented failures, and the five questions to ask before you pay for human review.

Read the article

Case studies

A page torn from a blank calendar beside a row of identical presenter screens

Training · Voiceover

Internal training courses

Voiceover for an LSP course library · recurring talent casting · updated segments only · native-speaker QA.

Read the case study

One master fader feeding a fan of ribbons out to a semicircle of regional stages

Brand · Up to 22 languages

Brand variants

On-screen graphics localization · subtitles · After Effects sources retained · Frame.io review cycle.

Read the case study

A machine-fed draft page rewritten by hand at the desk below the chute

Instructional · 5 languages

Instructional video localization

Voice-over · time-coded captions · on-screen text replacement · Frame.io review · 15–20 videos in 2–3 weeks.

Read the case study

A blank screen conducted toward ten microphones, one beam of light each

Commercial · 10 languages

Global product launch

Native VO casting · timed scripts to picture · subtitles for on-screen talent · 10 languages.

Read the case study