Omni is an editor, not a filmmaker. Use it accordingly.

Most video model reviews ask the same question: how pretty is the output? By that measure, Omni, Google’s new multimodal video model, will disappoint you. Prompt it cold for a moody campaign film and you'll get something pretty competent, but a little generic.

We took a different approach in our testing. We ran Omni through an internal evaluation across eight production categories, scored by two human annotators (one an AI PM with multimodal eval experience, one a creative workflow specialist from video production), against the highest-volume workflows we see in FLORA: advertising, fashion campaigns, e-commerce product visualization, and VFX compositing. Omni lost the same way every time and won the same way every time. Weak when generating footage from nothing. Strong, sometimes really strong, when changing footage that already exists.

Four places it earned a spot in our workflows.

Multi-element composition straight to video

This was the workflow result that surprised us most. We connected three separate inputs on the canvas: a model portrait, a product shot of the sneaker, and an outdoor scene with a stone pedestal. Instead of compositing them into a single image first, we prompted Omni to generate the video directly from all three. One node, one step. Both annotators scored it 4 out of 5. The comparison model needed the intermediate composite and still lost the shoe's design details in motion.

What Omni got right:

• Held identity across all three references, so the model, the shoe, and the environment each survived the jump to video intact Made real compositional judgment calls: framing, how the shoe sits in the scene, where the subject stands relative to the pedestal

• Collapsed a two-node workflow into one without the usual cost in consistency

• Omni lets you defer those decisions to a single generation while keeping element-level control through references. Campaign variations, product-in-context footage, and editorial shots from a product still and a location plate all get one step shorter.

Best used when you have strong reference assets and want the model to handle the scene assembly, not just the motion.

Kinetic typography and motion graphics inside FLORA

This one surprised us a bit. We prompted a kinetic typography sizzle with six words and six distinct animations, expecting the usual garbled type or a background full of hallucinated shapes. Omni gave each word its own animation with a strong rhythm and no background drift. Both annotators called it near production-ready, and it was Omni's highest-scoring category in our eval at 4.1 out of 5. The comparison model produced nearly identical animations across all six words on the same prompt. Six animations versus the same animation six times. On a sneaker prompt, Omni synced text appearance to camera motion and held the type together through the whole move.

Animated type is one of the most requested formats we see: brand manifesto videos, launch sizzles, social ads with overlaid copy. It usually means After Effects time or a motion designer's queue. Omni compresses that into a single node, which we love at FLORA.

• Distinct per-word animation styles in a single generation

• Text appearance synced to camera movement

• Type stays legible and correctly rendered through motion

One caveat: the advantage is motion-specific. Static type treatments are still a gap, so this isn't the model for a locked-off end card with dense copy.

Best used in workflows where type is the hero and it needs to move.

Camera direction and product hero shots inside FLORA

Every brand has some version of this problem: one good product photo, no budget for a shoot, and a product page that needs motion. Omni won our camera direction category outright, and the win came mostly from this exact scenario: a slow slide across a perfume bottle still life that scored 4 and 4.5 from our two annotators, with real foreground and background separation. The parallax reads as depth, not as a pan across a flat image, which is exactly the failure mode that makes most image-to-video product work look cheap.

• Convincing depth separation between foreground and background elements

• Works from a single still, no video capture required

• Stronger depth handling than the comparison model, which produced flat output with no foreground and background distinction

• 
  

Users should plan around two known issues. First, Omni has a reproducible speed hallucination: on prompts requiring constant camera speed, outputs sometimes accelerate mid-clip. Both annotators caught it independently across multiple prompts. Until it's fixed, generate a few takes and expect to select, or build the constraint into your Technique at the prompt level. Second, sustained single-take orbits are still out of reach. Neither model in our head-to-head test could hold one, so you should plan cuts into anything orbital.

Best used in workflows turning static product photography into short motion assets for PDP, paid social, or marketplace listings.

Effects compositing inside FLORA

Ask Omni to add rain to a window and it understands the glass is the compositing surface. Streaks land on the exterior with convincing physics while the interior stays clean. The comparison model's rain bled through onto interior surfaces on the same prompt. That boundary awareness, knowing where an effect should apply and where it shouldn't, is the difference between an effect that integrates and one that sits on top of the footage like a sticker.

• Correctly isolates the surface an effect belongs to

• Preserves the rest of the frame while applying the change

Use Omni for surface-bound effects and atmospherics on existing footage, not full physical simulation.

Best used in workflows adding weather, reflections, or surface treatments to shot footage without a VFX pass.

Omni isn’t all hits

Knowing where Omni breaks down is as useful as knowing where it excels. These patterns surfaced consistently across our testing, not as edge cases.

• Generating from scratch. On open cinematic prompts, Omni trends generic and stock-like, missing prompted lighting and texture detail. Other models are stronger for the hero footage itself.

• Image-to-video adherence. Without explicit, granular scene constraints, Omni frequently drifts from or ignores reference images. If your workflow starts from a reference frame, over-specify the scene or route to a different model.

• Multi-scene consistency. Across multi-beat campaign prompts, subject look and lighting logic wander. Not the model for a three-scene narrative in one generation.

• Physics. Water mass, gravity, and explosion dynamics were unconvincing, with debris trajectories too symmetrical and camera shake too polite for the intensity prompted.

• Audio. Generated audio felt out of scene context on every prompt we scored it on. Treat Omni output as picture only and handle sound elsewhere in your workflow.

• Static text. The typography strength is motion-specific. Dense static copy is not reliable.

  
  

The routing takeaway

Inside FLORA, the answer to "which video model is best" is always "for which node." Omni's node is the edit: swap the garment, animate the type, move the camera across the still, put rain on the glass. Generate your hero footage with a model built for cinematic mood, then hand it to Omni for the transformation pass. One model for the shot, one for the change. That's the pairing worth building your next Technique around.

Omni prompt pack

Try the model with these prompts below.

V2V garment replacement: best for virtual try-on, outfit variation, product swaps in existing footage

Replace the subject's [current garment] with [new garment: material, color, fit]. Preserve original body motion, hand positions, and hair movement. Match the scene's existing lighting on the new garment. Keep fabric physics natural throughout.

Kinetic typography: best for manifesto videos, launch sizzles, social ads with overlaid copy

[Copy line] builds word by word in [typeface style] over a [background]. Each word enters with a hard snap and 2-frame overshoot, then settles. Slow camera push-in throughout. Final word scales past frame edges and cuts to [end frame]. Rhythm synced to [tempo].

Product hero from a still: best for turning product photography into motion assets

Animate this product photo. Slow [orbital / push-in / tilt] camera move. [Light behavior: caustics, sweep, flicker] moves across the product surface. Subtle atmospheric particles in the backlight. The product stays static and sharp. Loopable, [duration].

Get the project here: https://app.flora.ai/view/js7dt9gqamcphmcjm9p99m3v0x8ae8gv