Gemini Omni 1.2 Flash on Segmind: generate, edit, refine and upscale video with audio
Gemini Omni 1.2 Flash is Google's video family for stylised, data and explainer clips with native audio and legible on-screen text. Generate from a prompt, a starting image or reference images; edit, refine and upscale finished clips up to 15 seconds, drafting at 360p and finishing to 4K.
Gemini Omni 1.2 Flash is Google's video model for stylised, data and explainer work — motion graphics, claymation, 3D, low-poly and doodle rather than photoreal footage. Every clip carries native synchronized audio and on-screen text you can read at thumbnail size. This release raises the ceiling to 15-second clips and edits, adds a 360p draft mode you can upscale to 4K, and brings conversational editing of clips you already made. Every output carries a SynthID watermark and C2PA Content Credentials. On Segmind it ships as six modes: text to video, image to video, reference to video, video edit, upscale and refine.
Text to video
Gemini Omni 1.2 Flash Text to Video turns a written prompt into a 3-15 second clip with native synchronized audio. It is strongest on stylised, no-people work: data videos, animated infographics and explainers with legible on-screen text. Quote every on-screen word in single quotes, give the beats in order, and the model follows them. Pick a resolution from 360p to 4K and a 16:9 or 9:16 frame.
ModelGemini Omni 1.2 Flash Text to VideoText-to-video with native audio and legible on-screen text.Flat 2D vector motion-graphics infographic, locked-off camera, deep teal background, bold white sans-serif type. First a title bar slides in from the left reading 'CITY RAINFALL'. Then three rounded blue bars grow upward one after another, each labelled beneath it 'JAN', then 'MAY', then 'SEP'. Finally a yellow tag pops in at the top reading 'MM PER MONTH'. The background grid and the title bar stay still; only the bars and the tag animate. No people, no photorealism. Sound: a soft paper whoosh as the title slides in, a short rising marimba note as each bar grows, a light chime when the tag appears. Every word spelled exactly as given, no other text.
It renders a flat teal infographic whose title, three labelled bars and 'MM PER MONTH' tag animate in order over a matched whoosh, marimba and chime track.
Stop-motion claymation scene on a tiny clay kitchen table, single continuous shot, locked-off camera. First a round clay teapot tips forward and pours a ribbon of amber clay tea into a white clay cup. Then a clay sugar cube drops in and the surface ripples. Finally a clay lemon wedge tips onto the saucer. The table and background stay still; only the teapot, tea, sugar and lemon move. No people, no text, no photorealism. Sound: a soft gurgle as the tea pours, a gentle plop as the sugar drops, a tiny clink as the lemon lands. No dialogue.
A stop-motion clay teapot pours into a clay cup, a sugar cube drops and a lemon wedge lands, each with its own sound.
Clean 3D product-render video, vertical 9:16, plain white studio background, single continuous shot, locked-off camera. A matte-black quadcopter drone rotates slowly on a turntable at the center of the frame. Floating callout labels appear one at a time with thin leader lines pointing to the drone: first 'CARBON FRAME', then '4K GIMBAL', then '38-MIN FLIGHT'. The drone and background stay steady; only the drone's rotation and the labels animate. No people, no photorealism. Sound: a soft tech whoosh as each label appears, a low steady hum underneath. No dialogue. Every word spelled exactly as given, no other text.
A matte-black drone turns on a white turntable in a 9:16 frame while three callouts — 'CARBON FRAME', '4K GIMBAL', '38-MIN FLIGHT' — appear in turn.
Use text to video from the API
Send a prompt and the clip settings; the picture and native audio come back in one response.
# pip install "segmind>=1.1.0"
# export SEGMIND_API_KEY="YOUR_API_KEY"
import segmind
result = segmind.run(
"gemini-omni-1.2-flash-text-to-video",
prompt="Flat 2D vector motion-graphics infographic, locked-off camera, deep teal background, bold white sans-serif type. First a title bar slides in from the left reading 'CITY RAINFALL'. Then three rounded blue bars grow upward one after another, each labelled beneath it 'JAN', then 'MAY', then 'SEP'. Finally a yellow tag pops in at the top reading 'MM PER MONTH'. The background grid and the title bar stay still; only the bars and the tag animate. No people, no photorealism. Sound: a soft paper whoosh as the title slides in, a short rising marimba note as each bar grows, a light chime when the tag appears. Every word spelled exactly as given, no other text.",
duration=6,
resolution="720p",
aspect_ratio="16:9",
)
print(result.get("output"))# 1. Submit the request; the reply carries its request_id
REQUEST_ID=$(curl -s -X POST "https://api.segmind.com/v2/gemini-omni-1.2-flash-text-to-video" \
-H "x-api-key: YOUR_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"prompt": "Flat 2D vector motion-graphics infographic, locked-off camera, deep teal background, bold white sans-serif type. First a title bar slides in from the left reading '\''CITY RAINFALL'\''. Then three rounded blue bars grow upward one after another, each labelled beneath it '\''JAN'\'', then '\''MAY'\'', then '\''SEP'\''. Finally a yellow tag pops in at the top reading '\''MM PER MONTH'\''. The background grid and the title bar stay still; only the bars and the tag animate. No people, no photorealism. Sound: a soft paper whoosh as the title slides in, a short rising marimba note as each bar grows, a light chime when the tag appears. Every word spelled exactly as given, no other text.",
"duration": 6,
"resolution": "720p",
"aspect_ratio": "16:9"
}' | jq -r .request_id)
# 2. Check its status until it is COMPLETED
curl -s "https://api.segmind.com/v2/requests/$REQUEST_ID/status" \
-H "x-api-key: YOUR_API_KEY"
# 3. Fetch the result
curl -s "https://api.segmind.com/v2/requests/$REQUEST_ID" \
-H "x-api-key: YOUR_API_KEY"Image to video
Gemini Omni 1.2 Flash Image to Video animates a starting image into a 3-15 second clip with native synchronized audio. Give it a first frame and a prompt that lists the motion in order. Add a last frame when the ending matters, which caps the clip at 10 seconds. It keeps the same stylised strengths: infographics, claymation and low-poly scenes that stay legible, not photoreal footage.
ModelGemini Omni 1.2 Flash Image to VideoAnimate a still image into a 15-second video with audio.Locked-off camera, nothing in the background moves. The three blue bars grow upward one after another into tall columns of different heights - first the 'JAN' bar to a short height, then the 'MAY' bar to a medium height, then the 'SEP' bar to the tallest. After the bars finish growing, a yellow rounded tag slides in at the top-right reading 'MM PER MONTH'. The deep teal background, the faint grid and the 'CITY RAINFALL' title stay perfectly still; only the bars and the tag move. Flat 2D vector style, no people, no photorealism. Sound: a short rising marimba note as each bar grows, then a light chime when the tag appears. Every word spelled exactly as given, no other text.
The flat 2D vector 'CITY RAINFALL' infographic animates cleanly at 720p: the three bars grow in order and the 'MM PER MONTH' tag slides in on a chime.
Claymation stop-motion style. A chubby brown modelling-clay treasure chest sits on flat sand-coloured clay ground against a plain pale-mint background. First the lid creaks and tilts, then it swings fully open, finally a faceted yellow clay gem rises into view inside the chest and gives off a small glow. The clip ends on the open chest with the yellow gem showing. Locked-off camera; the ground and the pale-mint background stay still. Sound: a wooden creak as the lid opens, then a bright chime when the gem appears. No people, no dialogue, no text.
The clip opens on the supplied closed-chest frame and lands exactly on the open-chest-with-gem frame, the lid creaking before a chime as the gem glows.
Low-poly 3D isometric style. A small faceted wooden cabin with a stone chimney sits on a low-poly green hill under a flat pale-blue sky. First a yellow light switches on in the cabin window, then a thin ribbon of grey smoke rises from the chimney, finally a small lantern by the front door lights up. Locked-off isometric camera; the cabin, the hill and the sky stay still. Sound: a quiet evening wind throughout, a short click as the window light turns on, a faint crackle from the chimney, a gentle ding when the lantern lights. No people, no dialogue, no text.
A low-poly isometric cabin lights its window, puffs smoke from the chimney and lights a porch lantern in order, each beat with its own sound.
Use image to video from the API
Pass a first-frame image and a prompt; the response carries the animated clip with native audio.
# pip install "segmind>=1.1.0"
# export SEGMIND_API_KEY="YOUR_API_KEY"
import segmind
result = segmind.run(
"gemini-omni-1.2-flash-image-to-video",
first_frame_image="https://segmind-resources.s3.amazonaws.com/input/gemini-omni-1.2-flash-image-to-video-firstframe.jpg",
prompt="Locked-off camera, nothing in the background moves. The three blue bars grow upward one after another into tall columns of different heights - first the 'JAN' bar to a short height, then the 'MAY' bar to a medium height, then the 'SEP' bar to the tallest. After the bars finish growing, a yellow rounded tag slides in at the top-right reading 'MM PER MONTH'. The deep teal background, the faint grid and the 'CITY RAINFALL' title stay perfectly still; only the bars and the tag move. Flat 2D vector style, no people, no photorealism. Sound: a short rising marimba note as each bar grows, then a light chime when the tag appears. Every word spelled exactly as given, no other text.",
duration=5,
resolution="720p",
aspect_ratio="16:9",
)
print(result.get("output"))# 1. Submit the request; the reply carries its request_id
REQUEST_ID=$(curl -s -X POST "https://api.segmind.com/v2/gemini-omni-1.2-flash-image-to-video" \
-H "x-api-key: YOUR_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"first_frame_image": "https://segmind-resources.s3.amazonaws.com/input/gemini-omni-1.2-flash-image-to-video-firstframe.jpg",
"prompt": "Locked-off camera, nothing in the background moves. The three blue bars grow upward one after another into tall columns of different heights - first the '\''JAN'\'' bar to a short height, then the '\''MAY'\'' bar to a medium height, then the '\''SEP'\'' bar to the tallest. After the bars finish growing, a yellow rounded tag slides in at the top-right reading '\''MM PER MONTH'\''. The deep teal background, the faint grid and the '\''CITY RAINFALL'\'' title stay perfectly still; only the bars and the tag move. Flat 2D vector style, no people, no photorealism. Sound: a short rising marimba note as each bar grows, then a light chime when the tag appears. Every word spelled exactly as given, no other text.",
"duration": 5,
"resolution": "720p",
"aspect_ratio": "16:9"
}' | jq -r .request_id)
# 2. Check its status until it is COMPLETED
curl -s "https://api.segmind.com/v2/requests/$REQUEST_ID/status" \
-H "x-api-key: YOUR_API_KEY"
# 3. Fetch the result
curl -s "https://api.segmind.com/v2/requests/$REQUEST_ID" \
-H "x-api-key: YOUR_API_KEY"Reference to video
Gemini Omni 1.2 Flash Reference to Video carries 1 to 10 reference images into a brand-new scene, keeping products, characters and brand marks looking like the same objects. Bind each reference in the prompt with its tag — <IMAGE_REF_0>, <IMAGE_REF_1> — and name exactly what to keep. Clips run 3 to 10 seconds with native audio, at 360p up to 4K.
ModelGemini Omni 1.2 Flash Reference to VideoTurn reference images into consistent AI video with audio.3D product-feature studio scene, no people, locked-off camera. Place the mint-green portable speaker from <IMAGE_REF_0> on a matte-black rotating podium against a deep-navy backdrop; keep its mint-green body, brass carry-loop, perforated grille and two side buttons exactly. The circular mint-and-brass brand emblem from <IMAGE_REF_1> glows on the wall behind it, unchanged. First the podium turns a quarter rotation while a thin light sweep crosses the speaker. Then a white rounded callout label slides in from the right reading '360 SOUND'. Finally a small white tag pops at the base reading 'IPX7'. Only the podium, the light sweep and the labels move; the backdrop and emblem stay still. Sound: a low studio hum throughout, a short whoosh as each label slides in, a bright synth chime when the 'IPX7' tag appears. Every word spelled exactly as given, no other text.
Two references — a mint-green speaker and its brand emblem — carry into a new rotating-podium studio scene with legible '360 SOUND' and 'IPX7' callouts and a matched soundtrack.
Claymation stop-motion meadow, single continuous shot, no scene cuts, locked-off camera. The clay toy triceratops from <IMAGE_REF_0> waddles across a bright clay meadow dotted with tiny clay flowers, then stops and tilts its head toward the camera. Keep its rounded green clay body, orange neck frill, three cream horns and stubby legs exactly; the grassy clay ground and painted blue clay sky stay still, only the triceratops moves. Sound: light playful marimba throughout, soft clay footstep taps as it waddles, a short cheerful squeak as it tilts its head. No dialogue.
The exact claymation triceratops from a single reference still waddles and turns its head in a fresh clay meadow, its colour, frill and horns intact.
Low-poly 3D forest clearing at dusk, single continuous shot, no scene cuts, locked-off camera. The low-poly fox from <IMAGE_REF_0> sits on faceted grass beside the low-poly campfire from <IMAGE_REF_2>, while the low-poly owl from <IMAGE_REF_1> perches on a faceted branch just above them. The fox's ears twitch, the campfire's flame facets flicker, and the owl turns its head toward the fire. Keep the fox's orange facets, the owl's grey facets and amber eyes, and the campfire's orange flames exactly; the faceted trees and ground stay still. Sound: soft night crickets throughout, a gentle fire crackle, a short low hoot as the owl turns. No dialogue.
Three separate low-poly references — a fox, an owl and a campfire — share one dusk clearing, each matching its still, as the owl turns toward the fire.
Use reference to video from the API
Supply an array of reference images and a prompt that tags each one; the clip returns with native audio.
# pip install "segmind>=1.1.0"
# export SEGMIND_API_KEY="YOUR_API_KEY"
import segmind
result = segmind.run(
"gemini-omni-1.2-flash-reference-to-video",
reference_images=[
"https://segmind-resources.s3.amazonaws.com/input/gemini-omni-1.2-flash-reference-to-video-ref0-e64f3190.jpg",
"https://segmind-resources.s3.amazonaws.com/input/gemini-omni-1.2-flash-reference-to-video-ref1-d835922b.jpg",
],
prompt="3D product-feature studio scene, no people, locked-off camera. Place the mint-green portable speaker from <IMAGE_REF_0> on a matte-black rotating podium against a deep-navy backdrop; keep its mint-green body, brass carry-loop, perforated grille and two side buttons exactly. The circular mint-and-brass brand emblem from <IMAGE_REF_1> glows on the wall behind it, unchanged. First the podium turns a quarter rotation while a thin light sweep crosses the speaker. Then a white rounded callout label slides in from the right reading '360 SOUND'. Finally a small white tag pops at the base reading 'IPX7'. Only the podium, the light sweep and the labels move; the backdrop and emblem stay still. Sound: a low studio hum throughout, a short whoosh as each label slides in, a bright synth chime when the 'IPX7' tag appears. Every word spelled exactly as given, no other text.",
duration=5,
resolution="720p",
aspect_ratio="16:9",
seed=-1,
)
print(result.get("output"))# 1. Submit the request; the reply carries its request_id
REQUEST_ID=$(curl -s -X POST "https://api.segmind.com/v2/gemini-omni-1.2-flash-reference-to-video" \
-H "x-api-key: YOUR_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"reference_images": [
"https://segmind-resources.s3.amazonaws.com/input/gemini-omni-1.2-flash-reference-to-video-ref0-e64f3190.jpg",
"https://segmind-resources.s3.amazonaws.com/input/gemini-omni-1.2-flash-reference-to-video-ref1-d835922b.jpg"
],
"prompt": "3D product-feature studio scene, no people, locked-off camera. Place the mint-green portable speaker from <IMAGE_REF_0> on a matte-black rotating podium against a deep-navy backdrop; keep its mint-green body, brass carry-loop, perforated grille and two side buttons exactly. The circular mint-and-brass brand emblem from <IMAGE_REF_1> glows on the wall behind it, unchanged. First the podium turns a quarter rotation while a thin light sweep crosses the speaker. Then a white rounded callout label slides in from the right reading '\''360 SOUND'\''. Finally a small white tag pops at the base reading '\''IPX7'\''. Only the podium, the light sweep and the labels move; the backdrop and emblem stay still. Sound: a low studio hum throughout, a short whoosh as each label slides in, a bright synth chime when the '\''IPX7'\'' tag appears. Every word spelled exactly as given, no other text.",
"duration": 5,
"resolution": "720p",
"aspect_ratio": "16:9",
"seed": -1
}' | jq -r .request_id)
# 2. Check its status until it is COMPLETED
curl -s "https://api.segmind.com/v2/requests/$REQUEST_ID/status" \
-H "x-api-key: YOUR_API_KEY"
# 3. Fetch the result
curl -s "https://api.segmind.com/v2/requests/$REQUEST_ID" \
-H "x-api-key: YOUR_API_KEY"Video edit
Gemini Omni 1.2 Flash Video Edit changes a source video from a plain-language instruction — add, remove or restyle elements inside it — and keeps its native audio. The source can run up to 15 seconds, and the edited clip follows the source length. Optional reference images, up to 10, bring specific props or styles into the same edit. On-screen text stays legible through the change when you keep the output at 720p or higher.
ModelGemini Omni 1.2 Flash Video EditEdit source video: restyle, add or remove elements.Name the target medium and list what must not move, and every label stays correct through the restyle.
Quoting each new word — 'PLUIE EN VILLE', 'JANV', 'MAI', 'SEPT', 'MM PAR MOIS' — and ending with 'every word spelled exactly as given' keeps the labels right.
Describe the prop and where it goes, add 'Keep everything else the same', and pass the reference image in the same edit call.
Use video edit from the API
Send the source video and an edit instruction; the edited clip returns at your chosen resolution.
# pip install "segmind>=1.1.0"
# export SEGMIND_API_KEY="YOUR_API_KEY"
import segmind
result = segmind.run(
"gemini-omni-1.2-flash-video-edit",
video="https://segmind-resources.s3.amazonaws.com/input/gemini-omni-1.2-flash-video-edit-source-infographic.mp4",
prompt="Restyle this entire motion-graphics clip into a low-poly 3D look: the flat bars become faceted low-poly 3D blocks with visible triangular facets, standing on a low-poly 3D ground plane and casting soft shadows, and the title bar and the yellow tag become clean low-poly 3D panels. Keep the exact same layout, bar heights and timing, locked-off camera, nothing repositioned. Keep every on-screen word exactly: 'CITY RAINFALL', 'JAN', 'MAY', 'SEP', 'MM PER MONTH', every word spelled exactly as given, no other text. A soft low synth pad as the look shifts to 3D, a short rising marimba note as each bar settles, a light chime at the end.",
resolution="720p",
seed=-1,
)
print(result.get("output"))# 1. Submit the request; the reply carries its request_id
REQUEST_ID=$(curl -s -X POST "https://api.segmind.com/v2/gemini-omni-1.2-flash-video-edit" \
-H "x-api-key: YOUR_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"video": "https://segmind-resources.s3.amazonaws.com/input/gemini-omni-1.2-flash-video-edit-source-infographic.mp4",
"prompt": "Restyle this entire motion-graphics clip into a low-poly 3D look: the flat bars become faceted low-poly 3D blocks with visible triangular facets, standing on a low-poly 3D ground plane and casting soft shadows, and the title bar and the yellow tag become clean low-poly 3D panels. Keep the exact same layout, bar heights and timing, locked-off camera, nothing repositioned. Keep every on-screen word exactly: '\''CITY RAINFALL'\'', '\''JAN'\'', '\''MAY'\'', '\''SEP'\'', '\''MM PER MONTH'\'', every word spelled exactly as given, no other text. A soft low synth pad as the look shifts to 3D, a short rising marimba note as each bar settles, a light chime at the end.",
"resolution": "720p",
"seed": -1
}' | jq -r .request_id)
# 2. Check its status until it is COMPLETED
curl -s "https://api.segmind.com/v2/requests/$REQUEST_ID/status" \
-H "x-api-key: YOUR_API_KEY"
# 3. Fetch the result
curl -s "https://api.segmind.com/v2/requests/$REQUEST_ID" \
-H "x-api-key: YOUR_API_KEY"Upscale
Gemini Omni 1.2 Flash Upscale raises a finished clip to a higher resolution and keeps its native audio. Feed it a clip of 15 seconds or less at 360p, 720p or 1080p; resolution is the only control, and the source drives the result. A 360p source reaches 720p or 1080p, a 720p source reaches 1080p or 4K, and a 1080p source reaches 4K.
ModelGemini Omni 1.2 Flash UpscaleUpscale video clips to 720p, 1080p or 4K.1080p is the everyday delivery target — draft at 360p, approve the frames and text, then finish.
Upscaling sharpens what is already there but will not redraw a dropped label.
Keep the source 9:16 so the portrait frame and native audio carry through; 720p is plenty for a social reel.
Use upscale from the API
Pass a finished clip and a target resolution; the upscaled clip keeps the original length and audio.
# pip install "segmind>=1.1.0"
# export SEGMIND_API_KEY="YOUR_API_KEY"
import segmind
result = segmind.run(
"gemini-omni-1.2-flash-upscale",
video="https://segmind-resources.s3.amazonaws.com/input/gemini-omni-1.2-flash-upscale-source-360p.mp4",
resolution="1080p",
)
print(result.get("output"))# 1. Submit the request; the reply carries its request_id
REQUEST_ID=$(curl -s -X POST "https://api.segmind.com/v2/gemini-omni-1.2-flash-upscale" \
-H "x-api-key: YOUR_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"video": "https://segmind-resources.s3.amazonaws.com/input/gemini-omni-1.2-flash-upscale-source-360p.mp4",
"resolution": "1080p"
}' | jq -r .request_id)
# 2. Check its status until it is COMPLETED
curl -s "https://api.segmind.com/v2/requests/$REQUEST_ID/status" \
-H "x-api-key: YOUR_API_KEY"
# 3. Fetch the result
curl -s "https://api.segmind.com/v2/requests/$REQUEST_ID" \
-H "x-api-key: YOUR_API_KEY"Refine
Gemini Omni 1.2 Flash Refine takes a clip you generated in the last 24 hours and applies a follow-up prompt that changes one thing and keeps the scene. Name what to change and what to keep, and quote any on-screen words exactly. The clip length follows the source and the native audio comes along; the follow-up does not guarantee the source resolution.
ModelGemini Omni 1.2 Flash RefineRefine a generated video with a follow-up prompt.The follow-up recolours the bars and rewrites the title bar, and nothing else in the scene moves.
Use refine from the API
Refine needs the interaction_id of a clip you just generated, read from its API response; the example leaves it blank, so paste your own.
# pip install "segmind>=1.1.0"
# export SEGMIND_API_KEY="YOUR_API_KEY"
import segmind
result = segmind.run(
"gemini-omni-1.2-flash-refine",
previous_interaction_id="",
prompt="Keep the same flat 2D vector motion-graphics infographic: deep teal background, bold white sans-serif type, locked-off camera, and the exact bar layout and grid. Change only two things - recolour the three bars from blue to warm amber, and rewrite the top title bar to read 'CITY SUNSHINE'. Every word spelled exactly as given, no other text. Add a soft chime the moment the title changes. Do not move the camera, the background grid, the labels or the bars.",
)
print(result.get("output"))# 1. Submit the request; the reply carries its request_id
REQUEST_ID=$(curl -s -X POST "https://api.segmind.com/v2/gemini-omni-1.2-flash-refine" \
-H "x-api-key: YOUR_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"previous_interaction_id": "",
"prompt": "Keep the same flat 2D vector motion-graphics infographic: deep teal background, bold white sans-serif type, locked-off camera, and the exact bar layout and grid. Change only two things - recolour the three bars from blue to warm amber, and rewrite the top title bar to read '\''CITY SUNSHINE'\''. Every word spelled exactly as given, no other text. Add a soft chime the moment the title changes. Do not move the camera, the background grid, the labels or the bars."
}' | jq -r .request_id)
# 2. Check its status until it is COMPLETED
curl -s "https://api.segmind.com/v2/requests/$REQUEST_ID/status" \
-H "x-api-key: YOUR_API_KEY"
# 3. Fetch the result
curl -s "https://api.segmind.com/v2/requests/$REQUEST_ID" \
-H "x-api-key: YOUR_API_KEY"Side by side
The family shares one subject across its modes: a flat teal rainfall infographic.
- Text to video
- Video edit
- Refine
Shared settings
Defaults are a 5-second 720p clip in 16:9; override them per mode.
| Setting | Value | Modes | When to use it |
|---|---|---|---|
| Prompt | Text instruction | Text, image, reference, edit, refine | Describe the scene, motion, on-screen text and sound; quote words to spell them |
| First frame | Starting image | Image to video | The still the clip animates from |
| Last frame | Optional ending image | Image to video | Pin the ending; caps the clip at 10 seconds |
| Reference images | 1-10 images | Reference (required), edit (optional) | Carry products, characters or props into the scene |
| Source video | Clip up to 15 seconds | Video edit, upscale | The clip to edit or upscale |
| Interaction id | Id from a clip in the last 24 hours | Refine | Point the follow-up at a clip you just made |
| Duration | 3-15s (3-10s for reference) | Text, image, reference | Length; edit, upscale and refine follow the source |
| Resolution | 360p, 720p, 1080p, 4K | Text, image, reference, edit; upscale 720p and up | Draft at 360p; use 720p or higher for legible text |
| Aspect ratio | 16:9 or 9:16 | Text, image, reference | Landscape or vertical; edit and upscale follow the source |
| Seed | -1 random | Text, image, reference, edit | Pin an integer to reproduce a take |
Good to know
- Every output carries a SynthID watermark and C2PA Content Credentials for provenance.
- A dense prompt can drop about one instruction in four, so check the frames and text before you upscale.
- Legible on-screen text needs 720p or higher; draft at 360p, then upscale the take you like.
- Photoreal people, acting, lip-sync, cinematic brand films and physics-heavy action are weaker and can trip the safety filters. English is fully supported; other languages work but vary.
- Reference to video caps at 10 seconds, and a first-plus-last-frame clip caps at 10 seconds. Upscale refuses sources over 15 seconds or outside 360p, 720p and 1080p, and a 360p source cannot reach 4K directly — step up through 720p or 1080p.
- Refine needs the interaction_id of a clip you generated in the last 24 hours on the same account; the id is single-use, so the stored example leaves it blank.
Resources
- Try Gemini Omni 1.2 Flash Text to Video on Segmind · API reference
- Try Gemini Omni 1.2 Flash Image to Video on Segmind · API reference
- Try Gemini Omni 1.2 Flash Reference to Video on Segmind · API reference
- Try Gemini Omni 1.2 Flash Video Edit on Segmind · API reference
- Try Gemini Omni 1.2 Flash Upscale on Segmind · API reference
- Try Gemini Omni 1.2 Flash Refine on Segmind · API reference
- Generate and edit videos with Gemini Omni Flash — prompt structure, on-screen text, audio prompting, timing beats and editing prompts
- Gemini Omni Flash model page — model overview, supported input and output types, and output-video limits
- #gemini-omni-1-2-flash
- #model-launch
- #video-generation
- #native-audio
