# Gemini Omni 1.1 Flash (`gemini-omni-1.1-flash`) Technical Reference `gemini-omni-1.1-flash` is Google's generally available (GA) multimodal foundation model for fast video generation, multi-reference synthesis, frame interpolation, and stateful conversational video editing via the **Interactions API**. --- ## 1. Technical Specifications | Property | Value | | :--- | :--- | | **Model Code** | `gemini-omni-1.1-flash` (replaces deprecated `gemini-omni-flash-preview`) | | **Supported Inputs** | Text, Image (`image/jpeg`, `image/png`, `image/webp`), Audio, Video (`video/mp4`) | | **Supported Outputs** | Video (`video/mp4` with synchronized stereo audio) | | **Input Token Limit** | `1,048,576` tokens | | **Output Clip Duration** | `3s` to `10s` per generation (`duration`: `"3s"`, `"4s"`, `"5s"`, `"6s"`, `"7s"`, `"8s"`, `"9s"`, `"10s"`, default `"10s"`) | | **Video Extension** | `+10s` per extension turn, up to `40s` total duration (uses the final `10s` of the prior clip as context) | | **Resolutions** | `360p`, `720p` (default), `1080p`, `4k` | | **Aspect Ratios** | `16:9` (default), `9:16` | | **Frame Rate** | `24 FPS` | | **Delivery Modes** | Inline base64 (`delivery: "inline"`, default) or Files API URI (`delivery: "uri"`) | | **Watermarking** | Invisible SynthID watermark embedded in video frames and audio | > [!NOTE] > **Regional Availability for Video Uploads**: Uploading input videos for video editing or video extension is not available in the European Economic Area (EEA), Switzerland, the United Kingdom, and certain US states. --- ## 2. `gemini-api video` CLI Recipes ### Text-to-Video with Timecoded Cuts & Dialogue ```bash gemini-api video \ '[0-4s] Medium shot of a robotics engineer in a brightly lit lab inspecting a bipedal robot arm. She says in a crisp voice: "Actuators calibrated." [4-10s] Cut to close-up as the metallic fingers pick up a delicate glass sphere with soft servo whirring sounds.' \ --resolution "1080p" \ --aspect-ratio "16:9" \ --out "robotics_lab.mp4" ``` ### Uploading Large Media via `gemini-api files upload` For large source videos or reusable reference assets, upload once via the Files API and pass the returned `files/` URI to `-i`: ```bash # Upload reference video and character sheet once REF_VIDEO=$(gemini-api files upload "dancer_ref_3s.mp4") CHAR_IMAGE=$(gemini-api files upload "character_concept.png") gemini-api video \ "[# References @Image1 @Video1] The character shown in performs the dance choreography from on a rooftop at sunset, upbeat funk bassline. Use Image1 as a character reference and Video1 as a motion reference." \ -i "$CHAR_IMAGE" \ -i "$REF_VIDEO" \ --resolution "1080p" \ --out "rooftop_dance.mp4" ``` ### Multi-Turn Extension Up to 40 Seconds ```bash # Segment 1 (0s - 10s) ID_1=$(gemini-api video \ "FPV drone flight entering a bioluminescent crystal cavern, echoing water drops and ethereal synth pad" \ --resolution "1080p" \ --out "cavern_10s.mp4" \ --jq '.response_id') # Segment 2 (10s - 20s) ID_2=$(gemini-api video \ "Extend this video: the drone dives through a narrow crystal arch into an underground subterranean lake as the synth music swells." \ --previous-interaction-id "$ID_1" \ --out "cavern_20s.mp4" \ --jq '.response_id') # Segment 3 (20s - 30s) ID_3=$(gemini-api video \ "Extend this video: after 2s the camera skims across the mirror-like water surface toward a glowing monolith." \ --previous-interaction-id "$ID_2" \ --out "cavern_30s.mp4" \ --jq '.response_id') ``` --- ## 3. SDK Patterns Across Languages ### Python: First-and-Last-Frame Interpolation ```python import base64 from pathlib import Path from google import genai client = genai.Client() first_b64 = base64.b64encode(Path("first_frame.jpg").read_bytes()).decode("utf-8") last_b64 = base64.b64encode(Path("last_frame.jpg").read_bytes()).decode("utf-8") interaction = client.interactions.create( model="gemini-omni-1.1-flash", input=[ {"type": "image", "data": first_b64, "mime_type": "image/jpeg"}, {"type": "image", "data": last_b64, "mime_type": "image/jpeg"}, { "type": "text", "text": " A smooth time-lapse transition from daytime construction to an illuminated skyscraper at night.", }, ], response_format={ "type": "video", "resolution": "1080p", "aspect_ratio": "16:9", }, ) if interaction.output_video and interaction.output_video.data: Path("interpolation.mp4").write_bytes(base64.b64decode(interaction.output_video.data)) ``` ### Python: URI Delivery Mode (`delivery: "uri"`) for Large Outputs When generating high-bitrate `1080p` or `4k` videos in backend services, request `delivery: "uri"` to stream the resulting file via `client.files.download` instead of transferring base64 in the JSON body: ```python import time from pathlib import Path from google import genai client = genai.Client() interaction = client.interactions.create( model="gemini-omni-1.1-flash", input="An aerial tracking shot over glacial fjords at golden hour.", response_format={ "type": "video", "resolution": "4k", "aspect_ratio": "16:9", "delivery": "uri", }, ) uri = interaction.output_video.uri file_name = "files/" + uri.split("/")[-1] while True: file_info = client.files.get(name=file_name) if file_info.state.name == "ACTIVE": break if file_info.state.name == "FAILED": raise RuntimeError(f"Video processing failed for {file_name}") time.sleep(5) video_bytes = client.files.download(file=uri) Path("fjords_4k.mp4").write_bytes(video_bytes) ``` ### TypeScript (`@google/genai`): Uploaded Video Editing & Extension ```typescript import { GoogleGenAI } from '@google/genai'; import * as fs from 'fs'; const ai = new GoogleGenAI({}); let videoFile = await ai.files.upload({ file: 'source_clip.mp4', mimeType: 'video/mp4', }); while (videoFile.state === 'PROCESSING') { await new Promise((r) => setTimeout(r, 5000)); videoFile = await ai.files.get({ name: videoFile.name! }); } const interaction = await ai.interactions.create({ model: 'gemini-omni-1.1-flash', input: [ { type: 'video', uri: videoFile.uri! }, { type: 'text', text: '[# Sources @Video1] Replace the clear blue sky with an aurora borealis and add gentle arctic wind sounds. Keep everything else the same.', }, ], response_format: { type: 'video', resolution: '1080p', }, }); if (interaction.output_video?.data) { fs.writeFileSync('aurora_edit.mp4', Buffer.from(interaction.output_video.data, 'base64')); } ``` ### Go (`google.golang.org/genai`): Text-to-Video Generation ```go package main import ( "context" "encoding/base64" "log" "os" "google.golang.org/genai" "google.golang.org/genai/interactions/models/interactions" "google.golang.org/genai/interactions/models/operations" ) func main() { ctx := context.Background() client, err := genai.NewClient(ctx, nil) if err != nil { log.Fatal(err) } videoFormat := interactions.VideoResponseFormat{ Resolution: interactions.ResolutionOneThousandAndEightyp.ToPointer(), AspectRatio: interactions.VideoResponseFormatAspectRatioSixteen9.ToPointer(), } res, err := client.Interactions.Create(ctx, operations.CreateInteractionRequest{ Body: operations.NewCreateInteractionRequestBody(interactions.CreateModelInteraction{ Model: interactions.Model("gemini-omni-1.1-flash"), Input: interactions.NewInteractionsInput("A drone shot of a mountain landscape at sunrise."), ResponseFormat: genai.Ptr(interactions.NewCreateModelInteractionResponseFormat( interactions.NewResponseFormat(videoFormat), )), }), }) if err != nil { log.Fatal(err) } if res.Interaction.OutputVideo != nil && res.Interaction.OutputVideo.Data != nil { videoBytes, err := base64.StdEncoding.DecodeString(*res.Interaction.OutputVideo.Data) if err != nil { log.Fatal(err) } if err := os.WriteFile("sunrise.mp4", videoBytes, 0644); err != nil { log.Fatal(err) } } } ```