Use Gemini Omni 1.2 Flash through EvoLink’s unified video API for 3–15s video generation, editing, extension and upscaling. Choose a task-specific model ID and estimate token costs before integrating.
Pick a route, write a prompt, and see the estimated cost before you spend anything.
Create a video
Every route runs on the same async video endpoint.Results are kept for 24 hours — download anything you want to keep.
Gemini Omni 1.2 Flash is billed by tokens, not by the second. The final charge always uses the tokens the model actually returned.
One 3s 360p sample
About 135 tests with $10 in credits.
Draft at 360p for a 720p or 1080p upscale, or draft at 720p for 4K. Upscale only the take you want to keep. Compare the total in the calculator, including any input and other output tokens.
Good for a first validation.
Good for prompt iteration.
Good before production integration.
Text-to-video estimates include the reserve for thinking tokens. Video Upscale is billed only for video output, by the length of the input video, with no input or thinking charge. Input images and input video are billed at the input rate.
Every request settles per token across three SKUs — video output, input and thinking — against the token counts the upstream actually returns. Video Upscale uses only the video output SKU.
| Token type | What it covers | Price | Billing |
|---|---|---|---|
| Video outputPer 1K tokens | The generated video with its synchronized audio. 720p produces 5,792 tokens per second. | $0.011 / 1K tokens-10% 0.7344 cr / 1K tokens$0.012 official | Output tokens |
| InputPer 1K tokens | Your prompt text, input images and input video, combined. Each image counts as 1,100 tokens; an input video is billed in full at its own resolution (up to 5,792 tokens per second). | $0.0014 / 1K tokens-10% 0.0918 cr / 1K tokens$0.0015 official | Input tokens |
| Thinking tokensPer 1K tokens | The model's internal reasoning. Typically a few hundred tokens per request. | $0.0081 / 1K tokens-10% 0.5508 cr / 1K tokens$0.0090 official | Output tokens |
Charges are settled per token against the tokens the upstream actually returned, so the final amount can differ slightly from the estimate. Logged-in users see the rates for their own pricing group. Video Upscale bills video output only: 2,896 tokens per second for 360p to 720p or 720p to 1080p, 5,792 for 360p to 1080p, and 11,584 for any upscale to 4K.
gemini-omni-1.2-flash-text-to-videoNative audio included. 360p and 1080p and 4K scale the output tokens by 1/3, 1.5 and 3 against 720p; final charges use actual token usage.
Gemini Omni 1.2 Flash is Google’s generally available multimodal video model, released on November 5, 2026, with the stable model ID gemini-omni-1.2-flash. EvoLink exposes six task-specific routes on one endpoint: text-to-video, image-to-video, reference-to-video, video edit, video extend, and video upscale. A generation creates 3–15 seconds with native synchronized speech, music, and sound effects; an extension appends 3–10 seconds; an upscale lifts a finished 360p, 720p, or 1080p clip to a higher resolution. Output tiers are 360p and 720p, plus upscaled 1080p and 4K.
Picking the right route matters more than sending more references. The same field can mean different things across routes — image order carries meaning in image-to-video and none in reference-to-video.
gemini-omni-1.2-flash-text-to-videoGenerate a 3–15 second clip from a prompt alone, with native synchronized audio, at 360p through 4K.
gemini-omni-1.2-flash-image-to-videoOne image animates it as the first frame for 3–15 seconds. Two images are read in order as the first and last frame, and the model interpolates between them in up to 10 seconds.
gemini-omni-1.2-flash-reference-to-videoGuide identity, style, or composition with up to 10 reference images plus up to 3 optional reference videos — only one video when the clip runs longer than 10 seconds. Order carries no meaning here.
gemini-omni-1.2-flash-video-editRewrite objects, style, or lighting in one input video of up to 15 seconds. Length and aspect ratio follow the input; quality: auto preserves its resolution, or you can select an output tier.
gemini-omni-1.2-flash-video-extendAppend 3–10 seconds to an input video of up to 30 seconds, continuing the motion and camera language. duration is the number of seconds added, not the total length.
gemini-omni-1.2-flash-video-upscaleLift one 360p, 720p, or 1080p clip of up to 15 seconds to 720p, 1080p, or 4K without a prompt; the content stays the same while detail increases. The input must be exactly 640×360, 1280×720, or 1920×1080 (or the vertical equivalent). Draft cheaply, then upscale only the keeper: a 360p draft plus an upscale to 1080p costs about 11% less than generating 1080p directly.
Iterate at 360p, confirm at 720p, deliver at 1080p or 4K — directly or by upscaling the draft you picked. Output tokens scale 1/3, 1x, 1.5x, and 3x against 720p.
Speech, music, and sound effects are generated with the video at no extra cost. There is no toggle and no audio input.
Two ordered images define the start and end of the shot, for clips of up to 10 seconds.
Manage long-running jobs with a task ID, polling, and HTTPS callbacks. Result links stay valid for 24 hours.
Use the EvoLink API for product integration and batch jobs, or call it from Codex, Claude, or Gemini for fast creative and development workflows. Both paths share the same EvoLink API key, balance, model routes, and task history.
Best for: product backends, batch jobs, automated workflows
Call EvoLink’s unified video API from your server and control the model ID, parameters, task queue, callbacks, and result storage.
Best for: creative and development tasks in Codex, Claude, and Gemini
Give the Agent your output goal, assets, and acceptance criteria. It can choose the route, assemble the request, track the task, and return the result without requiring you to hand-code every step.
This example shows the shortest runnable flow: create a task, save the returned task ID, then poll or wait for an HTTPS callback. Open the API tab for the complete parameter and response reference.
curl -X POST https://api.evolink.ai/v1/videos/generations \
-H "Authorization: Bearer $EVOLINK_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "gemini-omni-1.2-flash-text-to-video",
"prompt": "A premium product film with controlled camera motion",
"duration": 8,
"quality": "720p",
"aspect_ratio": "16:9",
"callback_url": "https://your-domain.com/webhooks/video"
}'
# Save the returned task ID, then query:
curl https://api.evolink.ai/v1/tasks/{task_id} \
-H "Authorization: Bearer $EVOLINK_API_KEY"Check every value against the parameter cards above: ranges, allowed options, asset counts, and file types must match the selected route.
Check the Authorization bearer token and confirm the available balance and task billing in the console.
Review likeness rights, brand assets, readable text, sensitive content, and input-asset compliance.
A Gemini Omni 1.2 Flash task that reaches a final failed state is not charged. Keep the task ID, inspect the final state, and identify whether the cause is validation, review, or execution before retrying.
callback_url must be a public HTTPS URL, not localhost or a private network address. Always keep the task ID as a polling fallback.
Use the same brief, the same input assets, and the same acceptance criteria to compare usable-shot rate, reruns, and total project cost. The right answer depends on shot length and how much resolution you actually need.
| Decision factor | Gemini Omni 1.2 Flash | Gemini Omni 1.1 Flash | Seedance 2.5 |
|---|---|---|---|
| Best suited to | Shots up to 15 seconds, reference-driven consistency, and drafting at low resolution before upscaling the keeper | Existing 1.1 integrations already tuned for 3–10 second shots | Long shots, many references, and audio-driven pacing |
| Duration | 3–15 seconds (first and last frame up to 10); Video Extend appends 3–10 more | 3–10 seconds; Video Extend appends 3–10 more | 4–30 seconds |
| Resolution | 360p / 720p / upscaled 1080p / upscaled 4K, plus standalone upscale of an existing clip | 360p / 720p / upscaled 1080p / upscaled 4K | 480p / 720p / 1080p |
| Inputs | Text, first and last frame, up to 10 reference images plus up to 3 reference videos (one when longer than 10s), or one video to edit, extend, or upscale | Text, first and last frame, up to 10 reference images plus up to 3 reference videos, or one video to edit or extend | Text, images, video and audio references, up to 50 assets |
| Audio | Native, always on, no audio input accepted | Native, always on, no audio input accepted | Native, toggleable, accepts audio references |
| Cost decision | Video output is $12 per 1M tokens at the official rate (EvoLink $10.80), about 31% below 1.1; draft at 360p and upscale only the keeper | Video output is $17.50 per 1M tokens at the official rate | Per-second pricing; compare cost per usable shot on longer clips |
The hard part of a production video API integration is not one request. It is managing model choice, cost, references, asynchronous jobs, callbacks, failures, and fallback routes after launch. EvoLink brings these into one video API and console so teams can test first, integrate second, and route by usable-shot rate and total project cost.
All six Gemini Omni 1.2 Flash routes share EvoLink authentication, balance, and task infrastructure. Your server stores one EVOLINK_API_KEY and selects the model ID that matches each input.
When you also need Seedance, MiniMax H3, Veo, or another video model, you do not have to maintain separate provider accounts, keys, balances, and authentication code. Keep the same task and billing framework and change routes by duration, resolution, and use case.
Out-of-range durations are rejected with a clear message instead of being silently clamped upstream, so you never pay for a length you did not ask for. Video Upscale inputs are checked against the exact supported sizes before submission, and a reference-to-video request longer than 10 seconds with more than one reference video is refused up front — otherwise the upstream fails it seconds later with an unhelpful "Internal error".
Video Extend is the sharper edge: its duration means "seconds appended", and omitting it makes the upstream append 10 seconds by default. EvoLink always sends an explicit duration.
Output tokens scale linearly with resolution, so a 15-second clip is 86,880 video tokens at 720p and 260,640 at 4K — a 3x difference on the largest line of the bill.
Video Upscale changes the math: generate drafts at 360p or 720p, then upscale only the one you keep. The Playground estimate updates as you change the route, resolution, and input video, so the decision happens before the spend, not after.
Save the task ID and verify the final status, failure reason, and billing record in the console instead of treating “request created” as “video generated.” Use the final billing record as the source of truth for the charge outcome.
Video generation is asynchronous. Production systems should use polling or a public HTTPS callback, handle missing or duplicate callbacks and timeouts, and retry only after identifying a validation, review, or execution failure.
Compare Google video routes under one EvoLink account and API key, then choose by shot duration, resolution, and cost.
Gemini Omni 1.1 Flash
The previous generation: 3–10s shots at 360p to 4K with first-and-last-frame control, video edit, and extend. Still live for existing integrations.
View API
Gemini Omni Flash
The original Gemini Omni Flash: 720p only, four routes, no video extend. Still live for existing integrations.
View API
Veo 3.1
Google’s cinematic video model with native audio, for longer-form and higher-fidelity shots.
View APIWhen Gemini Omni 1.2 Flash is not the default, use the same brief to compare resolution, shot duration, audio, motion stability, and total project cost.
Seedance 2.5
For 4–30 second shots with up to 50 multimodal references; compare usable-shot rate on longer clips.
View API
MiniMax H3
For detailed 2K clips of 4–15 seconds; compare its usable-shot rate with the same brief.
View API
Kling 3.0
For creative generation, action, and multi-shot video; a production fallback outside the Gemini family.
View API
Tencent Video Upscale
For clips outside Omni’s upscale sizes, longer than 15 seconds, or made by other models, use Tencent Video Upscale: up to 60 seconds, 720p to 4K, seven enhancement modes.
View APIHigh-intent questions about availability, model IDs, pricing, limits, model selection, and troubleshooting.
Yes. Gemini Omni 1.2 Flash became generally available on November 5, 2026, and EvoLink exposes six routes: text-to-video, image-to-video, reference-to-video, video edit, video extend, and video upscale. Validate the output in Playground first, then integrate with the same EvoLink API key.
gemini-omni-1.2-flash-text-to-video, gemini-omni-1.2-flash-image-to-video, gemini-omni-1.2-flash-reference-to-video, gemini-omni-1.2-flash-video-edit, gemini-omni-1.2-flash-video-extend, and gemini-omni-1.2-flash-video-upscale. Pick the route that matches your input — text-to-video rejects media URLs; edit, extend, and upscale each require exactly one input video, and upscale needs no prompt.
Billing separates video output, input (prompt text, images and video), and other output such as thinking. Check the current rates and full estimate in the pricing section; final charges use the returned token usage. Video Upscale bills video output only. A per-second video-output estimate is not the full cost of a generation or edit.
Yes, with Video Extend: each extension appends 3–10 seconds to an input video of up to 30 seconds, and the result contains the input plus the new seconds. Inputs longer than 30 seconds are rejected. Only the appended seconds are billed as video output; the input video is billed as input tokens.
A generation creates 3 to 15 seconds (up to 10 with a first and last frame, or with more than one reference video); an extension appends 3 to 10 seconds; everything is 24fps. 360p and 720p are direct outputs; 1080p and 4K are upscaled. Omitting the resolution gives 720p. Out-of-range values are rejected instead of being silently clamped.
Order carries meaning: the first image is the first frame and the second is the last frame, and with two images the clip can run at most 10 seconds. Sending one image animates it as the first frame. A last frame cannot be sent alone — position in the request is the only role signal.
It is the number of seconds appended to your input video, not the total output length. A 3-second input with duration 3 returns about 6 seconds, and only the appended part is charged as video output. EvoLink always sends an explicit duration, because omitting it makes the upstream append 10 seconds by default.
It generates speech, music, and sound effects with the video. Usage is settled with the returned video tokens; there is no separate audio toggle. It does not accept audio files as input.
Video Upscale only accepts exact sizes: 640×360, 1280×720, or 1920×1080, or the vertical 360×640, 720×1280, and 1080×1920. Even a 1278×720 clip is rejected, and many other models produce sizes outside this list (for example 864×496). Clips generated by Gemini Omni at 360p, 720p, or 1080p always match. The target must also be higher than the input: 360p to 720p or 1080p, 720p to 1080p or 4K, 1080p to 4K.
Content review occasionally flags harmless footage — for example as a photorealistic video of a minor or as containing a prominent person — even when no person appears in the clip. The same footage often passes as a video edit or extend input. Retry with a different clip, and check the task’s final status and billing record in the console.