Key takeaways
- Gemini Omni 1.2 Flash (
gemini-omni-1.2-flash) is Google's video generation and editing model, generally available since November 5, 2026. - Clips run 3 to 15 seconds in one-second steps, up from 10 seconds on Omni 1.1. First-and-last-frame clips run 3 to 10 seconds.
- Output comes in 360p, 720p, 1080p or 4K at 24 FPS, in 16:9 or 9:16. Google produces 1080p and 4K by upscaling.
- On PoYo you pay per output second: 5.46 credits at 360p, 16.38 at 720p, 24.58 at 1080p and 49.14 at 4K ($1 = 200 credits), from pay-as-you-go credits with no subscription.
- One model ID covers four modes. You choose text-to-video, first/last frames, reference images and videos, or video editing just by which fields you send.
- One API key, one simple flow. Submit a task, then poll its status or receive a webhook, the same way you call every other image and video model on PoYo.
Gemini Omni 1.2 Flash at a glance
| Item | Gemini Omni 1.2 Flash |
|---|---|
| Developer | Google DeepMind |
| Model ID | gemini-omni-1.2-flash |
| Status | Generally available, November 5, 2026 |
| Tasks | Text-to-video, image-to-video (first/last frames), reference-to-video, video editing, extension |
| Clip length | 3–15 s in one-second steps; 3–10 s with first and last frames |
| Resolution | 360p, 720p, 1080p, 4K (1080p and 4K are upscaled) |
| Frame rate | 24 FPS |
| Aspect ratio | 16:9 (default) or 9:16 |
| Video input for editing | Up to 15 s |
| Output | Video with native audio |
| Provenance | SynthID watermark on every video |
| Language | English fully supported |
This guide focuses on what 1.2 changes and how to call it. For the 1.1 model, its fixed-price tiers and its migration from the preview, see our Gemini Omni 1.1 Flash API guide.
What changed from Gemini Omni 1.1 Flash
| Gemini Omni 1.1 Flash | Gemini Omni 1.2 Flash | |
|---|---|---|
| Release | August 27, 2026 | November 5, 2026 |
| Clip length | 3–10 s | 3–15 s, any whole second |
| First and last frames | Supported | 3–10 s |
| Video input for editing | Up to 10 s | Up to 15 s |
| Extension (Gemini API) | Up to 10 s of input video | 3–10 s per turn, 10 s lookback, up to 30 s total |
| PoYo billing | Per generation, 4 / 6 / 8 / 10 s presets | Per output second, 3–15 s |
Google also reports better reference adherence, consistency and multi-turn editing, plus more natural speech, clearer lip-sync and multi-speaker conditioning. In the side-by-side comparison chart Google published for Omni 1.2 Flash, raters preferred 1.2 over 1.1 in every category shown:
| Task | Metric | Omni 1.2 | Omni 1.1 |
|---|---|---|---|
| Reference-to-video | Overall preference | 63% | 37% |
| Reference-to-video | Multi-reference identity consistency | 63% | 37% |
| Image-to-video | Overall preference | 59% | 41% |
| Image-to-video | Reference identity consistency | 60% | 40% |
| Text-to-video | Visual consistency | 62% | 38% |
| Text-to-video | Prompt adherence | 55% | 45% |
| Editing | Instruction editing | 57% | 43% |
| Editing | Edit instruction adherence | 54% | 46% |
Google notes that fast sports action, rapid script changes and complex physics are still areas of active work, so test those shots before you scale them.
Gemini Omni 1.2 Flash pricing on PoYo
PoYo bills Omni 1.2 Flash per output second from pay-as-you-go credits, with no subscription. Reference images and reference videos are included in the per-second rate.
| Resolution | Credits per second | USD per second |
|---|---|---|
| 360p | 5.46 | $0.0273 |
| 720p | 16.38 | $0.0819 |
| 1080p | 24.58 | $0.1229 |
| 4K | 49.14 | $0.2457 |
Generation bills the duration you choose. Video editing bills the length of the source video, rounded up to whole seconds.
Cost per clip
Credits per clip, with USD rounded to the cent:
| Duration | 360p | 720p | 1080p | 4K |
|---|---|---|---|---|
| 3 s | 16.38 ($0.08) | 49.14 ($0.25) | 73.74 ($0.37) | 147.42 ($0.74) |
| 5 s | 27.30 ($0.14) | 81.90 ($0.41) | 122.90 ($0.61) | 245.70 ($1.23) |
| 8 s (default) | 43.68 ($0.22) | 131.04 ($0.66) | 196.64 ($0.98) | 393.12 ($1.97) |
| 10 s | 54.60 ($0.27) | 163.80 ($0.82) | 245.80 ($1.23) | 491.40 ($2.46) |
| 15 s | 81.90 ($0.41) | 245.70 ($1.23) | 368.70 ($1.84) | 737.10 ($3.69) |
Worked examples
| Scenario | Calculation | Cost |
|---|---|---|
| Draft, then deliver: 20 drafts at 360p × 8 s, then 2 finals at 1080p × 8 s | 20×43.68 + 2×196.64 | 1,266.88 credits ($6.33) |
| Ad set in 9:16: one 6 s, one 10 s and one 15 s cut at 720p | (6+10+15)×16.38 | 507.78 credits ($2.54) |
| Edit a 12.4 s clip at 720p (billed as 13 s) | 13×16.38 | 212.94 credits ($1.06) |
The first row is the pattern that saves the most: settle prompt, timing and motion at 360p, where a draft costs a third of 720p, and pay for 1080p or 4K only on the takes you keep. For Google's own free and subscription routes, such as Google Vids, Flow and the Gemini app, see Is Gemini Omni free?
Why call Gemini Omni 1.2 Flash through PoYo
- One API key for every model. Omni 1.2 Flash sits next to PoYo's other video, image and language models, so you can switch or combine models without new accounts.
- Pay-as-you-go credits. Buy credits at $1 = 200 credits and pay per output second, with no subscription or monthly plan.
- All four modes under one model ID. Text-to-video, first/last frames, references and editing share
gemini-omni-1.2-flash. - A simple async API. One submit endpoint, one status endpoint and optional webhooks, the same for every model.
- Playground first. Test prompts, frames, references and edits in the browser with a live credit estimate, then copy the exact JSON request.
Four input modes, one model ID
PoYo picks the mode from the fields in each request, checking them in this order:
| Mode | Send | Duration | Notes |
|---|---|---|---|
| Video editing | video_url + prompt | Follows the source clip (up to 15 s) | Send prompt and an optional resolution with video_url |
| Reference generation | reference_image_urls and/or reference_video_urls | 3–15 s | Up to 10 images and 3 videos (each up to 15 s); images and videos can be combined |
| First/last frames | image_urls with 1 or 2 images | 3–15 s; up to 10 s with a last frame | One image sets the start; two set start and end, in order |
| Text to video | prompt only | 3–15 s | Add resolution, duration, aspect_ratio as needed |
Shared fields:
| Field | Values | Default |
|---|---|---|
prompt | Required; up to 20,000 characters | — |
duration | Integer 3–15 | 8 |
resolution | 360p, 720p, 1080p, 4k | 720p |
aspect_ratio | 16:9, 9:16 | 16:9 |
callback_url | Your webhook URL | — |
All media URLs are public HTTP or HTTPS links. Send image_urls on its own, and the two reference fields together or one at a time. Edited clips keep the framing of the source video.
How to call the Gemini Omni 1.2 Flash API on PoYo
The flow is asynchronous: submit a task, receive a task_id, then poll the status endpoint or wait for your webhook.
1. Submit a task
curl -X POST https://api.poyo.ai/api/generate/submit \
-H "Authorization: Bearer $POYO_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "gemini-omni-1.2-flash",
"callback_url": "https://your-domain.com/callback",
"input": {
"prompt": "[0-4s] Wide shot of a lighthouse on a cliff at dusk, waves crashing below. [4-9s] The camera pushes in as the lamp turns on and sweeps across the sea. [9-12s] Close-up of the keeper closing a logbook. Wind and distant gulls, no dialogue.",
"duration": 12,
"resolution": "720p",
"aspect_ratio": "16:9"
}
}'
The response returns the task ID:
{
"code": 200,
"message": "success",
"data": {
"task_id": "task-unified-1757165031-uyujaw3d",
"status": "not_started",
"created_time": "2026-11-14T10:30:00"
}
}
Always check code in the body: some validation errors return HTTP 200 with "code": 400.
2. Poll the task status
curl https://api.poyo.ai/api/generate/status/task-unified-1757165031-uyujaw3d \
-H "Authorization: Bearer $POYO_API_KEY"
data.status moves through not_started, running, and then finished or failed. A finished task returns the video URL and the credits charged:
{
"code": 200,
"data": {
"task_id": "task-unified-1757165031-uyujaw3d",
"status": "finished",
"credits_amount": 196.56,
"files": [
{
"file_url": "https://storage.poyo.ai/generated/video-abc123.mp4",
"file_type": "video"
}
],
"created_time": "2026-11-14T10:30:00",
"progress": 100,
"error_message": null
}
}
A 12-second 720p clip is 12×16.38 = 196.56 credits.
3. Full Python example
import os
import time
import requests
BASE = "https://api.poyo.ai"
HEADERS = {
"Authorization": f"Bearer {os.environ['POYO_API_KEY']}",
"Content-Type": "application/json",
}
body = {
"model": "gemini-omni-1.2-flash",
"input": {
"prompt": "A ceramic mug of coffee on a sunlit wooden table. Steam rises as the camera slowly orbits. Soft cafe ambience, no dialogue.",
"duration": 6,
"resolution": "360p",
"aspect_ratio": "9:16",
},
}
submit = requests.post(f"{BASE}/api/generate/submit", json=body, headers=HEADERS, timeout=30).json()
if submit.get("code") != 200:
raise RuntimeError(submit)
task_id = submit["data"]["task_id"]
while True:
status = requests.get(f"{BASE}/api/generate/status/{task_id}", headers=HEADERS, timeout=30).json()["data"]
if status["status"] == "finished":
print(status["files"][0]["file_url"], status["credits_amount"])
break
if status["status"] == "failed":
raise RuntimeError(status.get("error_message"))
time.sleep(10)
This 6-second 360p draft costs 32.76 credits. When the shot works, resubmit the same prompt at 1080p or 4k.
Request bodies for the other modes
First and last frames (two images, so duration stays at 10 or below):
{
"model": "gemini-omni-1.2-flash",
"input": {
"prompt": "One continuous camera move from the shop exterior through the door to the diner at the counter. Keep the signage and the character unchanged.",
"image_urls": [
"https://example.com/first-frame.webp",
"https://example.com/last-frame.webp"
],
"duration": 8,
"resolution": "1080p",
"aspect_ratio": "16:9"
}
}
A first frame of a miniature ramen shop exterior and a last frame of the diner at the counter, joined by one continuous camera push with Gemini Omni 1.2 Flash.
Reference images and videos:
{
"model": "gemini-omni-1.2-flash",
"input": {
"prompt": "The woman from the first image walks toward the camera wearing the jacket from the second image, matching the walking rhythm of the reference video.",
"reference_image_urls": [
"https://example.com/model.webp",
"https://example.com/jacket.webp"
],
"reference_video_urls": ["https://example.com/walk.mp4"],
"duration": 10,
"resolution": "720p",
"aspect_ratio": "9:16"
}
}
One model photo plus front and back shots of each outfit, turned into a 15-second virtual try-on sequence that keeps the same person.
Video editing (send prompt, video_url and an optional resolution; length and framing follow the source):
{
"model": "gemini-omni-1.2-flash",
"input": {
"prompt": "Change the season to winter with light snowfall. Keep everything else the same.",
"video_url": "https://example.com/source.mp4",
"resolution": "720p"
}
}
To refine an edit, send the returned video back as the next video_url with a new instruction.
One clip of a knight and a queen, edited turn by turn from an oak forest to a stable, a market fair and a stone castle while the characters and framing stay intact.
Prompt tips for Gemini Omni 1.2 Flash
A 15-second clip leaves room for several shots in one take, like this four-shot macro explainer filmed inside a beehive:
A 15-second, four-shot macro explainer inside a beehive with animated callouts, generated with Gemini Omni 1.2 Flash.
Google's prompting guidance, applied to the longer 1.2 clips:
- Use timecodes for 15-second clips. Syntax like
[0-4s],[4-9s]lets you place several beats or shots in one clip and control when each happens. - Describe the audio. Omni generates sound with the video. Say what you want to hear: dialogue, ambience, sound effects or music style.
- Keep negatives short. Google recommends simple phrases such as "No dialogue" to remove unwanted elements.
- Keep edit prompts simple. State the one change, then add "Keep everything else the same" to protect the rest of the shot.
- Spell out on-screen text. If a sign or caption must appear, write the exact words in the prompt.
- Say "one continuous shot" when you mean it. Omni may add cuts to make a clip feel like a short story; ask for a single shot if you need one.
- Write prompts in English for the most reliable results. Google lists English as fully supported.
Choosing duration and resolution
| Goal | Start with | Why |
|---|---|---|
| Prompt exploration | 360p, 4–6 s | 5.46 credits per second, fastest feedback |
| Social clip | 720p, 6–10 s, 9:16 | Good quality-to-cost balance |
| Multi-shot story or ad | 720p, 15 s | Room for 3–4 timecoded beats in one take |
| Controlled transition | First/last frames, ≤10 s | Fixes where the shot starts and lands |
| Product or character consistency | Reference images + 720p | Identity and products from separate files |
| Final delivery | 1080p or 4K after approval | Upscaling sharpens a good take; approve motion first |
Moving from 1.1 to 1.2
- Model ID: change
gemini-omni-1.1-flashtogemini-omni-1.2-flash. - Durations: 1.1 on PoYo takes 4, 6, 8 or 10 seconds; 1.2 takes any whole second from 3 to 15, with a default of 8.
- Field names: check your request against the 1.2 fields:
image_urlsfor frames,reference_image_urlsandreference_video_urlsfor references, andvideo_urlfor editing. - Budget: 1.1 on PoYo bills a flat price per generation; 1.2 bills per second, so a 3-second clip costs less than an 8-second one.
- Timeline: Google has announced that, from December 3, 2026, Gemini API requests to
gemini-omni-1.1-flashrun on 1.2, so 1.2 is the version to build on.
Both model IDs are available on PoYo. Re-run a small regression set (faces, products, fast motion, on-screen text, dialogue) on 1.2 before you switch production traffic, and compare cost per approved clip rather than cost per request.
FAQ
What is Gemini Omni 1.2 Flash?
It is Google's multimodal video model for generating and editing short clips with native audio. It accepts text, images and video, and produces 3- to 15-second videos at up to 4K. Google released it as generally available on November 5, 2026.
Is there a Gemini Omni Flash API?
Yes. Google offers gemini-omni-1.2-flash through the Gemini API and Google Cloud. On PoYo you call the same model with POST https://api.poyo.ai/api/generate/submit and "model": "gemini-omni-1.2-flash", then read the result from GET /api/generate/status/<task_id>.
How much does the Gemini Omni 1.2 Flash API cost?
On PoYo it costs 5.46 credits per second at 360p, 16.38 at 720p, 24.58 at 1080p and 49.14 at 4K, so an 8-second 720p clip is 131.04 credits ($0.66). You pay from pay-as-you-go credits, with no subscription.
How long can Gemini Omni 1.2 Flash videos be?
Any whole number of seconds from 3 to 15. With both a first and a last frame, the limit is 10 seconds. An edited video follows the length of its source clip, up to 15 seconds.
What is the difference between Gemini Omni 1.1 Flash and 1.2 Flash?
1.2 extends clips from 10 to 15 seconds, edits source videos of up to 15 seconds, bills per second on PoYo, and led 1.1 in every category of Google's published comparison chart, including 63% overall preference for reference-to-video.
Does Gemini Omni 1.2 Flash support 4K?
Yes. You can request 360p, 720p, 1080p or 4K. Google creates 1080p and 4K by upscaling, so check motion and detail at a lower resolution first.
Is the Gemini Omni 1.2 Flash API free?
The API is pay-as-you-go: Google offers Omni 1.2 Flash on the Gemini API's paid tier, and on PoYo you pay per second from credits. For trying Omni at no cost, Google offers routes in its own apps, such as Google Vids and daily Flow credits; our guide Is Gemini Omni free? compares them all.
Try Gemini Omni 1.2 Flash on PoYo
Open the Gemini Omni 1.2 Flash model page to test prompts, frames, references and edits in the Playground with a live credit estimate, then copy the JSON request into your code. If your pipeline still uses fixed 4–10 second tiers, the Gemini Omni 1.1 Flash model page is available as well. For other video models, see our best AI video generators of 2026.