OiiOii adds Viral Clone ad remixes with shot analysis and structure-preserving swaps

• egeberkina's workflow walkthrough shows OiiOii Viral Clone ingesting a source ad, breaking it into shot-level components, and treating the original pacing as a reusable scaffold rather than asking for a full manual remake.
• According to egeberkina's shot analysis post, the system extracts scene, character, dialogue, composition, camera movement, lighting, and shot purpose before generation starts.
• In egeberkina's customization step, the editable layer includes characters, products, clothing, environments, and other visual elements, while the original structure stays in place.
• egeberkina's preview step adds a review screen before credits are spent, and egeberkina's side-by-side variants shows one source ad turning into multiple branded remixes.
• A separate hasantoxr thread about OpenCreator pitches the same prize from another angle, consistent multi-shot ads built from one conversation, which makes the category fight pretty clear.

You can watch the source-upload test, skim the shot-analysis checklist, and compare the multi-variant output without leaving the evidence pool. The adjacent OpenCreator thread is useful because it names the pain point directly, consistent characters and worlds across shots, then packages it as an agent workflow with format, seed, and hook controls.

The core move is not text-to-video. It is video-to-structure. In egeberkina's upload step, the source ad gets analyzed shot by shot so the remake starts from an existing sequence instead of a blank canvas.

The extracted fields in egeberkina's checklist are specific enough to read like a production brief:

• Scene and environment
• Character
• Dialogue
• Shot composition
• Camera movement
• Lighting and color palette
• Shot purpose

That turns a good UGC ad into a dissected template. The interesting bit is the inclusion of shot purpose, because it suggests OiiOii is trying to preserve why a beat exists, not just what appears in frame.

Once the analysis is done, egeberkina's customization post says the editable controls sit on top of the preserved structure. The swaps include characters, products, clothing, environments, and individual visual elements.

Earlier in the thread, egeberkina's preset example describes picking an existing structure and replacing the original character. Later, the final test pushes that farther by changing the model, product, and styling while keeping the source sequence underneath.

For ad teams, that is the whole promise in one sentence: one proven structure, multiple brand-specific versions.

The most practical feature in the thread is the review layer. egeberkina's preview post says OiiOii shows what it understood from the source video before generation, so the user can catch errors before spending credits.

That preview also answers a trust problem common in video agents. If the system misread the product, speaker, or shot intent, the mistake appears upstream instead of after a full render.

The thread's strongest evidence is the output comparison. In egeberkina's side-by-side test, one skincare-style source ad becomes several different commercials with different products, branding, and styling.

The workflow shown across the thread is straightforward:

1. Upload a source ad.
2. Let OiiOii generate shot analysis.
3. Review the system's understanding.
4. Swap characters, products, and visual elements.
5. Render multiple variants from the same base structure.

The earlier skincare ad test matters here because it moves the demo away from a flashy cyberpunk example and into the more repetitive world of UGC commerce, where structure reuse is the real asset.

A separate thread from hasantoxr's OpenCreator post frames the broader market pitch more bluntly: most AI video demos look good until the second shot breaks character consistency.

In that thread, the claimed control stack is:

• Format selection: cinematic, social, commercial, or freeform
• Uploaded product or character
• A "Director Seed" for visual language
• A "Viral Hook" layered on top

Then hasantoxr's step-two post says the agent handles script, visuals, cuts, captions, and background music, while the output claim ties the result to access across VEO 3.1, Sora 2, Kling 3.0, and Seedance 2.0 in one subscription.

That does not confirm anything about OiiOii's own model stack, but it does add one new fact to the story: multiple video tools are converging on the same sell, keep the character stable, keep the pacing that already works, and compress the whole ad build into a single guided workflow.