Cross-Modal Ad Creative for Display, Video and Audio

Imagine a media brief that says, “This is one campaign. The platform will distribute budget across display, online video, streaming TV and audio.” A production team can easily hear a different instruction: “Make one hero film, then export it into several sizes.”

Those instructions are not equivalent. A platform can unify buying, audiences and optimization. It cannot make consumers receive the same information when one placement removes motion, another removes clicking, and another removes the screen entirely. If a brand procures only one master, automation can scale the campaign—and scale its information gaps with it.

The more useful production unit is not a hero film plus resizes. It is a creative system designed to survive across senses.

What changed—and what did not

On September 29, 2026, Amazon Ads announced Display, Video, Audio Plus (DVA+), a workflow that can run streaming TV, online video, display and audio within one campaign. Amazon says the system handles targeting, inventory, format allocation, bidding and continuous optimization. The launch was a closed beta in the US, with open beta expected in late October 2026, international expansion in Q4 and broader availability targeted for early 2027.

The boundary matters. Amazon's launch materials explicitly describe AI-generated creative for display, online video and streaming TV at launch; audio expansion is described as a post-launch Q4 step. A unified campaign exists, but that does not prove creative production has become identical across all four media types.

The procurement change is still structural. Instead of commissioning a film and asking every channel to adapt it later, brands should first define what each sensory condition must communicate, then reverse-engineer the capture, recording and design work.

One campaign contains three different sensory contracts

Display must work while still and silent

Display has little narrative time and no voiceover to rescue an unclear frame. It must identify the product or brand, establish one useful distinction and point toward a destination. If the central benefit appears only after eight seconds of a film, a banner crop has not inherited the argument.

Video and streaming TV carry mechanism and proof

Moving image can show sequence, causality, use and transformation. But the viewing environment changes the ending. Amazon's Streaming TV and Prime Video specifications describe full-screen, non-skippable ads and state that non-interactive streaming TV runs in non-clickable environments, so it should not contain click-oriented CTA elements. A “click now” ending that works on the web can become a false instruction in the living room.

Audio must close the loop without a screen

Amazon's audio specifications require the brand or product and its main features or benefits to be clear, recommend one action per creative, and prohibit click language for smart-speaker experiences. Audio can also follow another ad, so the script should not assume the listener already knows the context.

One capture plan can share assets across these conditions. The information architecture cannot be copied without testing. Our earlier framework for creative resilience across placements addressed ratios, interfaces and viewing contexts. Cross-modal planning goes one level deeper: it reallocates the message when a medium removes sight, sound or clickability.

The ONCE Cross-Modal Message Load Map

Before approving scripts or quotes, place every claim into a Message Load Map. The columns are not channel names. They are the sensory conditions the customer actually has. The rows are the persuasion jobs the campaign must complete.

Layer 1: the invariant core

Every version should preserve four things: product or brand identity, one primary promise, the boundary of its evidence, and a truthful next action. The expression can change; the facts cannot. If video demonstrates “installs in 30 seconds,” audio should not inflate that into “installs instantly,” and the product page should not specify something different.

This layer belongs in a governed Product Truth Kit: approved specifications, allowed claims, prohibited wording, market differences and supporting evidence. Otherwise each new format becomes another opportunity for facts to drift.

Layer 2: what the silent image must carry

Assign product recognition, visual distinction and a destination cue to display. Ask: with sound off and motion frozen, can someone understand what this is and why it deserves attention in two seconds? If not, plan a dedicated pack shot, detail, comparison or composition. Do not rely on a random frame from the hero edit.

Layer 3: what motion must carry

Give mechanism, sequence, use context and emotional movement to video. The shot list should distinguish proof shots from atmosphere. It should also include an ending that remains meaningful when a television placement cannot be clicked.

Layer 4: what sound must carry

An audio script must state identity, context, one benefit and one action without visual assistance. If a proposition cannot be understood through language or sound, either translate it into an audible outcome or acknowledge that it is a poor fit for audio. Narrating an invisible picture is not adaptation.

Layer 5: what the landing experience must carry

Specifications, variants, compliance detail and complex comparisons belong on the product or landing page. The ad creates qualified interest and accurate expectations; the page completes verification. Trying to make every format “complete” usually produces an overloaded 15-second spot and an unreadable banner.

Price approved message–modality combinations, not files

A campaign may have one concept yet contain several combinations that require separate writing, capture, rights and approval. ONCE recommends comparing scope with this internal metric:

Cost per approved message–modality combination = (concept development + capture or recording + adaptation + rights + QA + version control) / approved claim × format combinations.

This is not an industry benchmark or a promised price. It is a way to expose work hidden by the phrase “one master.” Does audio need a rewritten script and a voice session? Does display require a dedicated product angle? Does the TV ending need a non-click CTA? Do music, talent voice and image rights cover every medium and market?

Pricing by export count rewards folders full of low-value variants. Pricing approved combinations better reflects the usable capability the client is buying.

Choose among adaptation, pickups and separate executions

A. Light adaptation

Use it when the product is simple, the campaign has one claim, clean visual and audio stems already exist, and the market scope is narrow. The concept, principal photography and much of post-production can be shared, but silent comprehension, screen-free comprehension and the non-click ending still need separate checks.

B. Targeted pickups and recording

Use this when the concept survives but one modality lacks proof. Typical additions include a dedicated pack shot, product detail, an audio-specific script and record, or a different television end card. It costs more than resizing and far less than rebuilding the campaign.

C. Separate execution

Use it when the proposition depends on the very thing the medium removes. A visually complex demonstration may not support meaningful audio. A reveal driven by sound may collapse in silent display. Forcing one master across them lets the weakest translation cap the entire campaign.

A two-week pilot

In week one, choose one product and one primary promise. Complete the Message Load Map, then make one display, one video/TV and one audio execution, each with a single job. Review them with three tests: silent, eyes closed, and non-clickable.

In week two, record production time, approval rounds, rights limitations and launch status for each version. DVA+ reporting can break performance down by format and geography, so align those reports with explicit version IDs. First learn which message survives which sensory condition; then decide whether automated budget should expand.

The decision is simple. When a media platform places several formats inside one campaign, brands should not compress production into one file. Long-term efficiency comes from sharing an approved concept and a governed fact base while giving each sensory condition enough independent expression. Unified buying makes the platform smarter. A cross-modal creative system keeps the brand from losing its voice, its picture or its path to action inside that automation.