You stop adding sound in a second pass. Gemini Omni Flash renders picture and audio in one call. Text, a still or references go in.
Free account · ⚡50 on signup · No card
● Made with Omni Flash
Audio rendered with the picture. Every take, one call.
Gemini Omni Flash
“Make a photorealistic cinematic video of a man free-running across Italian rooftops at golden hour.”
10sGemini Omni Flash
“Make a cinematic video of a woman in cyberpunk city.”
10sGemini Omni Flash
“Make an editorial video of a watch.”
● Release notes
Chapter 01 · The mix
Audio renders with the picture. No foley, no switch.
Chapter 02 · Doors in
Text, an image or references. Same model, same call.
Chapter 03 · The cast
Cast, product and palette carry across the take.
Chapter 04 · The engine
Google model. Three input modes in one call.
● Nine by sixteen
The same model returns vertical when you pick the frame.
Gemini Omni Flash
Gemini Omni Flash
Gemini Omni Flash
Ask for the frame in the prompt and pick 9:16 in the composer. Omni Flash returns vertical with sound. Same four length stops either way.
A four-second clip: a lighthouse beam sweeps once across a stormy sea at night, waves hit the rocks, rain in the beam
Generated on OpenClips · omni-flash · September 2026
A street musician plays trumpet under a bridge at night, passing cars echo under the arch, warm sodium light on wet stone, medium shot
Generated on OpenClips · omni-flash · September 2026
Wide shot of a steam train crossing a viaduct in a green valley, the whistle blows and echoes, smoke drifts, slow pan following the train
Generated on OpenClips · omni-flash · September 2026
A woman in her late twenties talks straight into her phone camera in a bright kitchen, handheld selfie framing, natural window light, she smiles and gestures mid-sentence, casual creator energy, no on-screen captions
Generated on OpenClips · omni-flash · September 2026
Cinematic product shot: a matte black wireless earbud case turns slowly on a dark reflective surface, a single warm rim light sweeps across it, macro lens, shallow depth of field, slow push-in, plain unbranded case
Generated on OpenClips · omni-flash · September 2026
● How to
Three steps from blank box to a take with sound. The page works logged out. Sign-in gates the render only.
Make a photorealistic cinematic video of a man free-running across Italian rooftops at golden hour.
Make a cinematic video of a woman in cyberpunk city.
Make an editorial video of a watch.
01 ·
≈ 2 min
Press a theme to load a verbatim sheet prompt. Or describe the scene and the sound you want to hear.
02 ·
≈ 1 min
Up to 6 stills. The model holds them across the take. Pick the frame and the length.
03 ·
~1–2 min
Pick the stop. Confirm the ⚡ quote. Render, download, remix.
● Three ways in
Write the scene. The take comes back with sound. Or start from a community prompt in our AI video examples.
Start from the frame you already have. The still becomes a take. Full rates sit on the pricing page.
Attach up to 6 references. Cast, product and palette hold across the take.
● Steer it with sound
Ambience, foley and dialogue render with the picture. Press any card for sound.
● Use cases
Ten seconds with sound covers an ad, a hook, a beat. One call each.
9:16 nativeFeed-ready vertical with dialogue attached.
Sound onThe pitch renders with the picture. No voiceover session.
One callMacro detail and the sound it makes, one pass.
10s ceilingA bedroom becomes orbit without a cut.
● Every format
Landscape and vertical from the same model. Press any tile for sound.
Cinematic00:00:08:00Gemini Omni Flash
Watch with sound
Story ad00:00:10:00Gemini Omni Flash
Watch with sound
Macro00:00:10:00Gemini Omni Flash
Watch with sound
World-building00:00:08:00Gemini Omni Flash
Watch with sound
One take00:00:10:00Gemini Omni Flash
Watch with sound
Lifestyle00:00:10:00Gemini Omni Flash
Watch with sound
Epic scale00:00:08:00Gemini Omni Flash
Watch with sound
Kitchen demo00:00:10:00Gemini Omni Flash
Watch with sound
Atmosphere00:00:08:00Gemini Omni Flash
Watch with sound
Macro00:00:10:00Gemini Omni Flash
Watch with sound
Journey00:00:10:00Gemini Omni Flash
Watch with sound
Product pitch00:00:10:00Gemini Omni Flash
Watch with sound
Cinematic00:00:08:00Gemini Omni Flash
Watch with sound
Desk setup00:00:10:00Gemini Omni Flash
Watch with sound
Dialogue00:00:10:00Gemini Omni Flash
Watch with sound● The sheet
Omni Flash for mixed input and sound you cannot switch off. Veo 3.1 for cinematic polish.
| Spec | Gemini Omni Flash | Veo 3.1 |
|---|---|---|
| Status | Gemini Omni FlashLive in the composer | Veo 3.1Live in the composer |
| Input | Gemini Omni FlashText, image or reference | Veo 3.1Text and reference images |
| Length | Gemini Omni Flash4, 6, 8 or 10s | Veo 3.14, 6 or 8s |
| Audio | Gemini Omni FlashAlways on | Veo 3.1Opt in per render |
| Resolution | Gemini Omni Flash360p, 720p, 1080p or 4k | Veo 3.1720p or 1080p |
| References | Gemini Omni FlashUp to 6 | Veo 3.1Up to 3 |
| Frames | Gemini Omni Flash16:9 · 9:16 | Veo 3.116:9 · 9:16 |
| From | Gemini Omni Flash⚡17 | Veo 3.1⚡326 |
● Spec sheet
Figures reflect OpenClips defaults, September 2026.
Compare every model that renders sound.Compare every model that takes several references.
● What the takes cover
The jobs Omni Flash lands. And the ones it wins.
● Quoted first
4s · omni-flash
50
from $0.31
The cheapest way to test a beat.
8s · omni-flash
100
from $0.62
Where the composer opens.
10s · omni-flash
125
from $0.78
The top of the ladder. Sound included.
● OpenClips MCP
No console and no syntax. Your agent briefs Omni Flash through the OpenClips MCP server. It queues the render and hands back the cut.
https://mcp.openclips.aiStep 01
One sentence is the brief. Subject, vibe, platform.
Step 02
Frame, length, references and the ⚡ quote.
Step 03
Download it, remix it, or ship it.
● Director's notes
Four habits that separate a clean take from a re-roll.
Name the dialogue, the room tone, the effects. Audio renders either way.
Say what it sounds like
One still per character or product. Six slots per render.
One still per subject
Prove the beat on the cheapest stop. Then buy the length.
Test short, buy long
Pick 9:16 before you render, not after. There is no reframe pass.
Pick the frame first
● From the blog
Model GuidesThree flagship video models, three personalities. Which one to roll.
Read it →
Prompt CraftSix beats that turn one line of text into a trailer-grade clip.
Read it →
Prompt CraftWhich frame wins where, and how to compose for it before you render.
Read it →● Questions
Omni Flash is free to try. A new account gets ⚡50 on signup with no card. Video renders draw from that wallet, and an Omni Flash take costs more than the grant, so top up when the deliverable matters. Packs start at $10/800⚡. Image models stretch the grant furthest.
Yes, and there is no switch to turn it off. Audio is not a second pass on Omni Flash. It renders with the picture in the same call. Dialogue, ambience and effects come back on the timeline, ready for the editor.
Four stops: 4, 6, 8 or 10 seconds. The composer opens on 8. It is a ladder, not a free length, so a take never runs past 10 seconds. A 10-second render is ⚡125 and a 4-second one is ⚡50.
Text, an image or references. Omni Flash is the any-to-any surface in the Google column. Type the brief, drop a still to animate, or attach up to 6 references to hold a cast. All three land on the same composer chip. Audio references are not an input.
Yes. Google embeds SynthID into the video its models generate. SynthID is a digital watermark that is imperceptible to humans. It is designed to survive cropping, filters and compression. OpenClips passes the render through untouched, so the mark stays with the file. You can check any clip in Gemini or the SynthID Detector.
Veo 3.1 for cinematic polish at ⚡326 (from $2.02). Omni Flash for mixed input and sound that is always on at ⚡17 (from $0.11). Both are Google models and share the composer, wallet and library.
● The receipts
Every link goes to the vendor's own page.
Google embeds SynthID into the video its models generate. SynthID is a digital watermark that is imperceptible to humans.
https://deepmind.google/models/synthid/It is designed to survive cropping, filters and compression.
https://deepmind.google/models/synthid/Sound included. Four stops, priced before you press.
Generate with Omni FlashOr roll with a sibling model
Omni Flash takes a prompt, a still or a set of references and returns video with sound. Audio is not a second pass. It renders in the same call, and there is no switch to turn it off.
Length is a ladder, not a slider: 4, 6, 8 or 10 seconds. The composer prices the stop before you render. Anything longer belongs to another model on the roster.
Reach for Veo 3.1 when the deliverable needs cinematic polish. Reach for Omni Flash when the brief starts from mixed input and the sound has to be there.
● Where to go next
