Status at a glance
The 1.1 column records the current working model. The 1.2 column is a forecast based on the existing product direction, not a claim that the predicted controls or quality gains have already been tested.
| Dimension | Gemini Omni 1.1 Flash | Gemini Omni 1.2 Flash (forecast) |
|---|---|---|
| Model ID | gemini-omni-1.1-flash | Expected: gemini-omni-1.2-flash |
| Current site workflow | Text/image, reference video, and video edit modes | Expected to retain the same three modes with a newer generation backend |
| Inputs | Text, image, and video | Expected: text, image, audio, and video references |
| Tasks | Text-to-video, image-to-video, reference-to-video, edit, and extend | Expected: generation, reference-to-video, editing, extension, and frame control |
| Output controls | 3-10 second output with landscape/portrait and resolution controls | Expected: landscape/portrait output with 720p, 1080p, and 4K delivery options |
Two side-by-side comparison tests
Each case fixes the task, input assets, duration, framing, success criteria, and attempt count. Prompts follow the Universal Video Prompt Skill: observable end states, explicit locks, and no unverified model-specific syntax.
Product motion between fixed endpoints
Can the model connect two supplied product frames without changing the object or missing the final composition?
Fixed conditions
- - 8 seconds
- - 16:9
- - same two images
- - one attempt per model
Success criteria
- The opening matches the supplied first frame.
- The camera completes one smooth clockwise arc without a cut.
- The product shape, label placement, and material remain stable.
- The final composition visibly reaches the supplied last frame.
View common brief and prompt
Common brief
A matte black wireless speaker sits on a dark stone plinth in a quiet studio. Connect the supplied opening and closing frames with one smooth clockwise camera arc. Preserve the speaker geometry, grille pattern, controls, and lighting direction.
Shared comparison prompt
Create one continuous 8-second cinematic product shot using the supplied first and last frames. Begin exactly from the first frame. Move the camera in a slow, smooth clockwise 30-degree arc while the matte black wireless speaker remains fixed on the dark stone plinth. Preserve the speaker geometry, grille pattern, button positions, label placement, surface material, background, and key-light direction. End visibly on the supplied last-frame composition. No cuts, no extra objects, no text, no logo changes, and no deformation.
Character identity with transferred motion
Can the model transfer a simple walking motion while preserving the supplied character and wardrobe?
Fixed conditions
- - 8 seconds
- - 16:9
- - same image and motion clip
- - one attempt per model
Success criteria
- The character face, hair, jacket, and shoes remain recognizable.
- The walk follows the pace and body rhythm of the motion reference.
- Hands, legs, and body count remain anatomically coherent.
- The camera and background remain stable for the full shot.
View common brief and prompt
Common brief
Use one shared character image for identity and one short walking clip for motion only. Place the character on a quiet wet city street at blue hour. Keep a fixed front-facing camera and a full-body composition.
Shared comparison prompt
Create one continuous 8-second front-facing full-body shot. Use the supplied character image only for identity, face, hair, navy jacket, gray trousers, and white shoes. Use the supplied video only for walking pace, arm swing, and body rhythm; do not copy its person, clothing, or background. The character walks toward a fixed camera on a quiet wet city street at blue hour. Reflections move naturally under the shoes. Keep the face, outfit, body proportions, background layout, and camera position stable. No cuts, no duplicate person, no wardrobe drift, no extra limbs, and no text.
Expected 1.2 direction
What Gemini Omni 1.2 Flash is expected to add
The next workflow is expected to keep the core strengths of 1.1 while making multimodal generation, frame control, editing, extension, and final delivery easier to manage in one place.
Unified multimodal references
Combine text, images, audio, and video references in one brief instead of rebuilding the project across separate tools.
Stronger frame guidance
Use opening and closing images to define the visual endpoints while the model generates the motion between them.
Reference-led motion
Transfer movement and timing from a reference clip while preserving the subject, wardrobe, product, or style supplied elsewhere.
Focused video editing
Describe a targeted change while keeping the source camera path, scene layout, subject identity, and unaffected materials stable.
Scene extension
Continue a source shot beyond its current ending while maintaining direction, action, lighting, and visual continuity.
Flexible delivery controls
Prepare landscape or portrait output with expected 720p, 1080p, and 4K delivery options for different production stages.
Task-by-task comparison
How the two versions fit real video workflows
A useful model comparison starts with the job to be completed. The sections below separate what can be done with Gemini Omni 1.1 Flash now from the workflow improvements expected in Gemini Omni 1.2 Flash, then translate each difference into a practical production decision.
Text and image to video
Gemini Omni 1.1 Flash today
Gemini Omni 1.1 Flash already provides a practical starting point for short text-to-video and image-to-video work. A prompt can define the subject, action, camera, environment, lighting, and sound, while an uploaded image gives the shot a visual anchor. This makes 1.1 useful for concept clips, social posts, product motion tests, and early story beats that need a working output now.
Gemini Omni 1.2 Flash direction
Gemini Omni 1.2 Flash is expected to preserve that direct workflow while making the handoff between text, images, and other references more unified. The meaningful improvement would not simply be a newer model number: it would be fewer restarts when the creator moves from an idea to a guided shot, plus more reliable use of the same subject and composition across revisions.
Practical takeaway: Choose 1.1 for immediate generation. Prepare richer briefs and reusable reference packages for 1.2.
First and last frame control
Gemini Omni 1.1 Flash today
A first-and-last-frame task asks the model to respect two fixed visual endpoints and create believable motion between them. The challenge is not only reaching the final frame. The object must keep its geometry, materials, labels, lighting direction, and spatial relationship to the camera throughout the transition. Our first paired test uses identical images, duration, framing, and prompt constraints to make those failures visible.
Gemini Omni 1.2 Flash direction
The expected 1.2 workflow places greater emphasis on frame-guided generation as part of the same workspace. For production teams, this could make shot planning more predictable because opening and closing compositions can be approved before motion is generated. The comparison should still judge the entire path, including acceleration, camera continuity, object stability, and whether the final composition is reached without a last-second jump.
Practical takeaway: Evaluate the middle of the clip as carefully as the endpoints; a matching final frame alone is not a complete success.
Reference-led character and motion work
Gemini Omni 1.1 Flash today
Reference-led generation separates what should be copied from what should remain original. An image may define identity and wardrobe, while a short video contributes only walking pace, body rhythm, or camera movement. A strong 1.1 prompt therefore assigns one role to each asset and explicitly prevents the model from importing the source performer, clothing, background, or unwanted props from the motion clip.
Gemini Omni 1.2 Flash direction
Gemini Omni 1.2 Flash is expected to make this multimodal role assignment more natural. The hoped-for workflow is one in which character, style, motion, sound, and scene references can be combined without losing track of their separate purposes. The second paired case tests exactly that boundary: identity should come from the image, movement from the video, and the environment from the written brief.
Practical takeaway: Use asset-specific instructions. Do not assume the model knows whether a reference supplies identity, style, motion, or composition.
Video editing and scene extension
Gemini Omni 1.1 Flash today
Gemini Omni 1.1 Flash supports a useful edit-first workflow: upload a source clip, describe the focused change, and lock everything that should remain untouched. The same discipline applies to extension. The prompt should describe what happens after the existing ending while preserving direction of travel, camera language, lighting, subject identity, and environmental continuity. Broad prompts often invite unnecessary rewrites of the whole shot.
Gemini Omni 1.2 Flash direction
For 1.2, editing and extension are expected to sit more comfortably beside generation and reference-led creation. That matters when a project moves through several passes: generate a base shot, adjust one material or background element, then continue the action. The practical test will be whether each pass preserves prior decisions instead of introducing identity drift, geometry changes, flicker, or a new camera path.
Practical takeaway: Describe the smallest acceptable edit and list the visual properties that must stay locked across every frame.
Resolution, aspect ratio, and delivery
Gemini Omni 1.1 Flash today
The current 1.1 workspace exposes duration, aspect ratio, and resolution controls so the requested output is defined before generation. These controls help teams avoid evaluating a landscape concept when the real deliverable is vertical, or spending unnecessary time on a high-resolution pass before motion and composition are approved. A short draft is often the right first checkpoint even when the final target is more demanding.
Gemini Omni 1.2 Flash direction
Gemini Omni 1.2 Flash is expected to support landscape and portrait production with delivery choices that extend through 720p, 1080p, and 4K. Higher resolution should be treated as a delivery option, not proof of stronger motion, identity, or prompt adherence. The best workflow validates creative behavior first, then moves the selected shot to the resolution and format required by the final channel.
Practical takeaway: Compare models at matched settings, then judge resolution separately from composition, motion quality, and instruction following.
Upgrade workflow
Plan a fair move from 1.1 to 1.2
Keep the same creative brief
Reuse the subject, action, camera, environment, timing, and negative constraints that already work in Gemini Omni 1.1 Flash.
Lock shared inputs and settings
Use the same reference assets, duration, aspect ratio, resolution target, and attempt count for both versions.
Compare complete outputs
Review identity, motion, frame accuracy, visual continuity, audio, and instruction adherence across the entire clip.
Move production after validation
Switch a workflow only when the newer result meets your quality bar and the required controls are available in the workspace.
Use-case recommendations
Choose by production need, not version number
The best choice depends on deadlines, reference complexity, review standards, and how much workflow change a team can absorb. These recommendations turn the comparison into an action plan for common creator and production scenarios.
Creators shipping today
Start with Gemini Omni 1.1 Flash
Use 1.1 when the project cannot wait for a newer backend. It already covers text and image generation, reference-video workflows, focused editing, and extension. Keep prompts modular and save the exact assets and settings so successful shots can be rerun or compared later.
Brand and product teams
Build a controlled reference package
Prepare approved identity images, product angles, material references, motion clips, opening and closing frames, and a list of locked details. This package improves current 1.1 prompting and gives the team a fair baseline for assessing 1.2 without changing the brief between versions.
Studios planning repeatable pipelines
Test workflow reliability, not one hero result
Run several matched attempts for the tasks that matter most, then record adherence, failure modes, revision count, output handling, and delivery time. The right model is the one that meets the production threshold consistently, not simply the one with the most dramatic single sample.
Teams preparing for Gemini Omni 1.2 Flash
Separate expected features from measured gains
Use this page to prepare prompts and evaluation criteria, but keep quality, speed, pricing, and success-rate claims open until the integrated 1.2 workflow can be tested. When access arrives, replace forecasts with recorded settings, complete outputs, and repeatable observations.
Which version fits your workflow?
Choose 1.1 now
You need a working production path
Gemini Omni 1.1 Flash is the verified choice for current text, image, reference-video, editing, extension, interpolation, and upscaling workflows. Its API model ID and controls are documented and available.
Wait for 1.2
Your decision depends on confirmed changes
Wait when you need verified 1.2 pricing, limits, model ID, controls, or measured quality differences. Re-run the two controlled cases after access is available; do not infer a quality winner from version numbering.