Editing videos can be really fun, but if you want to do it seriously, check out this prompting method.
Your best prompts are the ones you'd never bother typing.
The detailed ones. The ones with examples and edge cases. Wispr Flow lets you speak them instead — clean, structured, ready to paste into any AI tool. Free on Mac, Windows, and iPhone.
✅ Before You Start
A Google account
One of the following access paths:
Google AI Plus, Pro, or Ultra subscription (for Gemini app and Google Flow)
A YouTube channel with access to YouTube Shorts or YouTube Create app (free)
For avatar creation: a clear photo or short video of your face
For reference based generation: your source images, audio files, or video clips ready to upload
SynthID awareness: every Omni output carries an invisible watermark identifying it as AI generated
🚀 How to Access It
Gemini Omni Flash is available on three surfaces as of May 19, 2026.
Via Gemini App (Paid)
Requires Google AI Plus, Pro, or Ultra subscription.
1. Go to gemini.google.com
2. Sign in with your Google account
3. Open a new chat
4. Look for the video generation option in the input bar
5. Upload your inputs or type your promptVia Google Flow (Paid)
Requires Google AI Plus, Pro, or Ultra subscription. Best for multi step creative projects.
1. Go to flow.google.com
2. Sign in
3. Start a new project
4. Use the Flow Agent for brainstorming, batch editing, or multi variation generation
5. Switch to Omni Flash for video outputVia YouTube Shorts (Free)
No subscription required. Rolling out this week to all YouTube Shorts and YouTube Create app users.
1. Open the YouTube app or YouTube Create app
2. Tap Create
3. Look for the AI video generation option powered by Gemini Omni
4. Enter your prompt or upload a reference
5. Generate and post directly to Shorts⚠️ API access for developers and enterprise customers is not yet available. Google confirmed it will follow "in the coming weeks" after the May 19 launch.
🛠️ Setup Step by Step
First Generation (Gemini App)
Step 1: Open gemini.google.com and sign in
Step 2: Start a new conversation
Step 3: In the input bar, select the video or media attachment option
Step 4: Upload your reference image, audio, or video clip (or skip and use text only)
Step 5: Type your generation prompt (see Prompts section below)
Step 6: Review the 10 second output clip
Step 7: Type a follow up instruction to iterate without starting over
Step 8: Download or share when satisfiedSetting Up Your Avatar
Step 1: In the Gemini app, navigate to avatar settings
Step 2: Upload a clear photo or short video of your face
Step 3: Follow the on screen steps to create your digital likeness
Step 4: In any video prompt, type "include me presenting this" to use your avatar⚠️ Clips are capped at 10 seconds at launch. This is a deployment decision, not a model limitation. Google DeepMind product management director Nicole Brichtova confirmed the cap is designed to manage compute demand during the rollout phase.
💬 Prompts to Use It
Scene Generation from a Photo
Here is a photo of [location or subject]. Animate this into a 10 second video.
Add natural motion: wind in the trees, light changing, subtle camera movement forward.
Ground the physics in reality. No surreal effects.Brand Video from Mixed Inputs
I am attaching a product photo, a brand color reference image, and a voiceover audio file.
Generate a 10 second product video that uses all three as references.
The tone should be clean and premium. Use the voiceover as the audio track.Iterative Editing
Good start. Now do the following:
Change the lighting to late afternoon golden hour.
Slow the final 3 seconds down by 50 percent.
Keep everything else consistent with the previous version.YouTube Shorts Content
Generate a 10 second hook video for a YouTube Short about [topic].
Open with a visual that stops the scroll. Text safe zone on lower third.
Upbeat pacing. No dialogue needed. End on a strong visual frame.Avatar Presentation
Create a 10 second clip of my avatar presenting the following:
[Your script here, max 2 sentences for 10 seconds]
Background: [describe or upload a reference]
Tone: confident and direct.Reference Synthesis
I am uploading 3 short clips. Each has a quality I want to combine:
Clip 1: the camera movement style
Clip 2: the color grading
Clip 3: the pacing and cut rhythm
Generate a new 10 second scene that pulls all three together. Subject: [describe your scene]📊 Commands Table
Action | How to Do It |
|---|---|
Generate from text only | Type prompt in Gemini chat input |
Generate from image reference | Upload image then add prompt |
Generate from audio reference | Upload audio file then describe the visual |
Mix multiple inputs | Upload all files then describe how to use each |
Edit a previous output | Type follow up instructions referencing the clip |
Use your avatar | Include "use my avatar" or "include me presenting" in prompt |
Check SynthID watermark | Open clip in Gemini app and use the verify tool |
Access Google Flow Agent | Open flow.google.com and start a project |
Batch generate variations | Use Flow Agent with multi variation generation feature |
Free access via Shorts | Use YouTube Create app, no account upgrade needed |
📅 Daily Workflow
This is how a content creator realistically uses Omni day to day.
Morning: content planning Open Google Flow. Use the Flow Agent to brainstorm 5 video concepts for the week. Ask it to generate 3 variations of each concept brief. Pick the strongest.
Midday: generation Take your best concept into Gemini Omni Flash. Upload any reference assets you have (product photos, brand images, audio). Run your first generation. Iterate 2 to 3 times in the same conversation using follow up prompts.
Afternoon: Shorts output Export your clip. Post directly to YouTube Shorts. If you want multiple variations for A/B testing, go back to Flow and use batch editing to generate 3 versions of the same brief.
On the go Use the YouTube Create app for quick generations. Free tier. No setup needed. Good for reactive content around trending topics.
🔥 Tips
Mix your inputs intentionally. The model is strongest when you give it a visual reference AND a text brief. Vague text only prompts produce generic results. A photo plus a clear direction produces something specific.
Treat it like a conversation, not a search. Do not try to write the perfect prompt on the first try. Generate something rough, then talk it into shape with follow up instructions. Three short follow ups beat one long prompt.
Use your avatar for talking head content. If you create content that requires a presenter on screen, set up your avatar once and reuse it across every project. Saves hours of filming.
Keep your SynthID exports. Every Omni output is watermarked. If a platform ever challenges AI generated content, you can verify the origin through the Gemini app.
Flow Agent is underrated. Most people will skip straight to generation. The Flow Agent for brainstorming and batch editing is where the real time saving is. Use it before you touch Omni.
10 seconds is enough for Shorts hooks. The cap is not a problem for the use case it was built for. A 10 second visual hook at the start of a Short is exactly the format that stops the scroll. Build your content strategy around it.
🔧 Troubleshooting
"Video generation option not showing in Gemini app" Your subscription tier may not be updated yet. Check that you are on Google AI Plus, Pro, or Ultra. If your subscription is correct, the rollout may not have reached your account yet. Wait 24 to 48 hours and check again.
"YouTube Create app not showing AI video option" The free rollout to YouTube Shorts and YouTube Create began the week of May 19, 2026. If you do not see it yet, update the app to the latest version and check again over the following days.
"My reference image is not being used correctly" Be explicit in your prompt about how you want each input used. Instead of just uploading and prompting, write: "Use the attached image as the scene background. Do not change the composition. Add motion and lighting only."
"The clip cuts off before I wanted" All clips are capped at 10 seconds at launch. This is a hard limit on the current deployment, not a prompt issue. A higher tier Omni Pro model with longer output is planned but has no confirmed release date yet.
"Avatar does not look accurate" Re upload with a clearer, well lit frontal photo. Avoid sunglasses, heavy shadows, or side angle shots. The model needs a clean frontal reference to generate an accurate likeness.
"Audio editing of my existing clip is not working" This feature is intentionally disabled at launch. Google is holding back audio editing of existing video due to deepfake risk concerns. There is no workaround. It will be released at a later date.
"API calls are returning errors" Developer API access is not yet generally available. It was confirmed to be coming "in the coming weeks" after May 19. Check Google AI Studio for updates on API availability.
⚡ Quick Reference
ACCESS PATHS
Free: YouTube Shorts app / YouTube Create app
Paid: gemini.google.com (Plus, Pro, Ultra)
Creator tool: flow.google.com (Plus, Pro, Ultra)
API: Coming soon via Google AI Studio
CURRENT LIMITS
Max clip length: 10 seconds (deployment cap, not model limit)
Audio editing: Disabled at launch (deepfake precaution)
API access: Not yet available (coming weeks)
Omni Pro: Announced, no release date confirmed
INPUTS SUPPORTED
Text prompt ✓
Image upload ✓
Audio upload ✓
Video upload ✓
Mixed inputs ✓ (all at once in one prompt)
OUTPUT
Format: High resolution video with audio
Watermark: SynthID (invisible, verifiable in Gemini app)
SURFACES
Gemini app ✓ gemini.google.com
Google Flow ✓ flow.google.com
YouTube Shorts ✓ free, via app
YouTube Create ✓ free, via app
API Coming soonThe best tools spread person to person. Be that person.
By The AI Leverage - Learn and master AI daily