Skip to main content
Visonic AI Visonic AI Stories & Notes

· Ari Surana · Product  · 3 min read

Introducing Visonic AD 2: Ultra, Prime and Micro

Choose the level of processing and review that fits each title, from flagship productions to large archives.

Choose the level of processing and review that fits each title, from flagship productions to large archives.

Since its inception, Visonic AI has been breaking new ground. With Visonic AD 1, we created the first commercially available AI audio description model to achieve human-level quality. It proved that automated audio description didn’t have to be a mechanical compliance exercise but could be a truly immersive, narrative-driven experience.

Today, with Visonic AD 2 we’re taking our biggest leap forward yet. A new family of audio-description models built on a simple premise: a great description doesn’t just list what’s in the frame - it understands the story.

Whether you’re delivering flagship cinema or clearing a massive content archive, AD 2 matches your exact production priorities with zero compromise on continuity.

Prime and Micro are available today. Ultra is coming soon.

Model at a Glance

FeatureUltra (Coming Soon)PrimeMicro
ForStudio-grade precision for flagship releasesDefault for quality & production efficiencyCost-efficient for high-volume content
Best-Fit ContentFeature films, premium drama, flagship cinemaEpisodic TV, documentaries, corporate mediaArchives, social video, internal catalogues
Context & ContinuityHighest (spatial detail & editorial nuances)Deep narrative tracking & character continuityCore long-form context & plot tracking
QA OverheadNear-zero review requiredSignificantly reduced review timesMinimal review needed

Purpose-Built for Your Production Requirements

With AD 2, model selection is an active editorial and operational strategy.

By matching the model to the specific demands of your title, production teams can precisely align three core variables:

  • Title Complexity: Scale model depth based on plot density, scene transitions, and narrative subtlety.

  • Review & QA Overhead: Match model precision to your team’s review capacity, significantly shortening the time needed for final sign-off.

  • Production Economics: Allocate your audio-description budget dynamically across your entire catalogue rather than applying a single flat rate to vastly different types of content.

What Changed in AD 2

AD 2 focuses on the errors that appear when an automated system loses track of the title: a character misidentified after a long absence, a returning location treated as new, fast action skipped, or an object described without understanding its later importance.

AD 2 addresses these continuity problems directly in the first generated draft, before editorial QA even begins.

The Model Has to Follow the Title

Visual captioning describes what is present in a frame. Audio description also has to judge why that frame matters to the story.

A useful system must carry context across a film or episode, recognise setup and payoff, understand which character is acting, respect the soundtrack, and fit language into the spaces where it can be heard. A sentence can be factually accurate and still be wrong for the story.

AD 2 maintains context across a complete title.

Why Prime Matters

Most active productions need a dependable workhorse model that balances narrative accuracy with operational efficiency. Prime is engineered as that default choice. It handles complex visual storytelling such as tracking fast action, character continuity, and subtle plot cues so broadcast and episodic teams get consistent, high-grade output without inflating QA time.

Why Micro Matters

Most videos will never have a flagship production budget. Without a practical cost per title, deep archives, creator content, and internal media remain undescribed. Micro eliminates that bottleneck by lowering unit economics while retaining the long-form context fundamental to the AD 2 family.

Moving the Industry Forward

Visonic didn’t enter the market after automated audio description became viable. Visonic made it viable. With AD 2, we are reaffirming our commitment to advancing long-form video understanding for media accessibility across the entire industry.

We invite creators, broadcasters, and production teams to experience the AD 2 model family on their own content catalogues.

Compare current usage pricing or review the complete Visonic AD 2 model family.

Try Visonic Audio Description

Upload a video, choose the model and language, then review the timed script and audio in the Visonic app.

Related Posts

View All Posts »