Start Here: What Google Flow Actually Is
Understand what Flow does, what it does not do and how this course is organised.
Google Flow is an AI creative studio
Google Flow is an AI filmmaking and creative tool for generating images, video clips, scenes and stories. It brings several Google generative models into one project-based workspace.
What you can make
Images
Create original images, edit existing images and prepare visual references for video.
Video clips
Generate clips from text, images, ingredients, start frames and start-plus-end frames.
Scenes
Arrange clips in Scenebuilder, change their order, trim them, preview the sequence and download the result.
Reusable creative assets
Create characters, use your personal avatar, organise project assets and build custom Flow Tools.
Before you begin
- Check access. You must be at least 18, in a supported region and using an eligible Google AI or qualifying Workspace plan.
- Use a supported device. A desktop Chromium browser gives you the strongest editing experience. Mobile availability and features can differ.
- Open a project. Keep the assets for one production inside the same project so they are easier to find and reuse.
- Check the active model. Available features, clip length, resolution and credit cost depend on the model selected.
What this course teaches
- How to create and edit images inside Flow
- How Ingredients and visual references work
- How to use start and end frames
- How to create reusable Characters and use your own Avatar
- How to generate dialogue with consistent voice references
- How to choose between Veo 3.1 and Gemini Omni options
- How to plan consistent multi-shot videos without promising impossible perfection
- How to edit, extend and assemble clips
- How the real Google Flow Agent and custom Tools work
Projects, Models & Credits
Set up a clean project and understand why model selection matters before generating.
Start with one organised project
A Flow project holds the images, videos, characters, scenes, uploads and collections used for one creative production. Keeping related work together makes references easier to find and reduces accidental mix-ups.
Create and organise a project
- Open Google Flow and select New project.
- Give the project a specific name, such as Coffee Launch Vertical Video.
- Upload or generate the main visual references before creating video.
- Use Collections to group character references, location images, product images and approved clips.
- Rename useful assets so they are easy to find with search or the @ picker.
Know what is stored
Assets
Images, videos, characters, scenes and uploads stored within the project.
Collections
Folder-like groups used to organise related assets. Collections can sit inside other collections.
History
Previous versions and prompts remain available when you edit an image or video.
Why the active model matters
Flow uses different models for different jobs. A feature that works with one model may be unavailable with another. Check the generation settings before every important run.
Credits and failed generations
Credit allowances depend on your Google plan and can change. Flow displays the active model and cost before generation. Google states that failed generations are not charged and refunded credits may take a moment to appear.
Useful project settings
- Grid or batch view for arranging generations
- Sound on hover
- Return silent videos
- Show tile details
- Clear the prompt after submission
Prompting for Google Flow
Write clear prompts that describe the shot without burying the model in contradictions.
Build the shot in layers
A useful Flow video prompt tells the model what is in the shot, what changes during the clip and how the camera should show it.
Subject
Who or what the viewer sees.
Action
The main movement or event in the clip.
Environment
Where it happens and what must remain visible.
Camera
Shot size, angle and intentional movement.
Lighting & style
Time of day, mood, realism and colour treatment.
Sound
Dialogue, ambience and important sound effects when supported.
A practical prompt order
- Reference any Character, Avatar, Ingredient or Frame first.
- Describe the setting and the subject's starting position.
- Describe one clear action with a natural beginning and ending.
- Add the camera framing and movement.
- Add lighting, visual style and sound.
- Finish with the few exclusions that genuinely matter.
Example video prompt
Use the supplied woman and coffee-machine images as visual references. A bright modern kitchen in soft morning light. The woman stands beside the kitchen counter with the coffee machine clearly visible. She pours coffee beans into the machine, closes the lid and looks toward the camera with a small, satisfied smile. Medium shot at eye level. Slow, steady push-in. Natural body movement and realistic hand motion. Warm cinematic colour, gentle kitchen ambience and the soft sound of beans pouring. Keep the same woman, clothing, kitchen layout and coffee machine throughout this clip. Do not add other people or extra products.
Keep one clip manageable
Trying to cram several locations, an outfit change, multiple camera moves and a long speech into one short generation makes failures more likely. Split the idea into separate shots and assemble them later.
When a prompt fails
- Remove conflicting directions.
- Reduce the number of actions.
- Replace a weak reference image.
- State where the subject begins and ends.
- Use a start frame when composition matters.
- Generate a fresh clip instead of endlessly adding instructions to a broken one.
Create & Edit Images in Flow
Build standalone images and stronger reference assets without leaving your Flow project.
Images are part of the video workflow
Flow can generate and edit images directly inside a project. You can keep them as finished assets or use them as Ingredients, start frames and end frames for later video generations.
Create an image with the standard prompt box
- Open a project and click the model name in the prompt box.
- Select Image.
- Choose the image model, aspect ratio and number of outputs.
- Write a detailed image prompt and add references if needed.
- Select Generate.
Example reference-image prompt
Create a clean full-body character reference image. One adult woman standing naturally, facing almost directly toward the camera. Her face, hair, hands, complete outfit and shoes are clearly visible. Plain light-grey studio background with even soft lighting and no hard shadows. Photorealistic detail, natural skin texture, realistic body proportions and no beauty filter. No props, no text, no other people and nothing covering the face or clothing.
Edit an existing image
Agent edit
Turn on the Flow Agent, drag in one or more images and describe the change. The edit is added to the asset's history stack.
Select and edit
Select a specific part of an image and prompt Flow to add, remove or change that area.
Crop
Open an image, choose Crop and select the crop you need.
Draw and text
Use the drawing tools to mark an area, draw a rectangle or add text before generating the requested change.
Keep useful versions
Editing does not automatically destroy the original. Open the History panel to view earlier versions and their prompts. Save any version you want to reuse back to the project.
Preparing images for video references
- Keep the subject or product clear and well lit.
- Use a plain or segmented background when possible.
- Avoid extra people or objects that Flow may treat as important.
- Keep the visual style compatible across all reference images.
- Use separate references for a person, product and environment when each one matters.
Ingredients & Visual References
Use uploaded or generated media to guide the people, products, locations and style in a clip.
What an Ingredient is
An Ingredient is visual media added to a generation as a reference. It can guide the appearance of a person, product, object, location or style from one clip to another.
Add Ingredients to a video prompt
- Open the generation settings and choose Video, then Ingredients.
- Drag media from your device or project into the prompt box, use Add, or type @ to find a named project asset.
- Add the reference for each element that matters.
- Describe how each reference should appear in the clip.
- Confirm the model supports Ingredients before generating.
A clean reference set
Person
Clear face and clothing, minimal obstruction and no unrelated people.
Product
Full shape visible, simple background and enough detail to recognise it.
Location
Room layout, furniture, windows and lighting direction visible.
Example Ingredient prompt
Use the supplied woman as the person reference, the supplied black coffee machine as the product reference and the supplied kitchen as the environment reference. The woman stands at the left side of the counter beside the coffee machine. She presses the top button and watches the machine begin brewing. Keep the kitchen layout, morning light, clothing colours and product shape consistent with the supplied references. Medium-wide eye-level shot with a slow push toward the product. Natural movement and realistic scale. Do not add extra people, appliances or product packaging.
Important limitations
- A reference guides generation but does not paste the original pixels into the video.
- Small writing, logos and packaging details may change.
- Too many references can create conflicts.
- Text instructions should support the visual references, not contradict them.
- Ingredient support and available clip lengths vary by model.
Frames to Video
Control how a clip begins and, when supported, where it ends.
Use frames when composition matters
A frame gives the model a visual starting point. A start-plus-end frame generation also gives it a target ending image and asks the model to create the movement between them.
Start frame
Animate a still image or begin a clip from a chosen composition.
Start + end frames
Create a transition between two supplied images. Both frames should be visually compatible.
Create a Frames video
- Open the generation settings and select Video, then Frames.
- Add an image to the start-frame position.
- If the active model supports it, add an end frame.
- Describe the action or transition that should occur between the frames.
- Choose orientation, length, outputs and the compatible model.
- Generate and inspect the complete clip.
Start-frame prompt
Begin from the supplied start frame. The woman slowly turns from the coffee machine toward the camera while holding the cup steadily in both hands. The camera remains at eye level and makes a gentle push forward. Keep the room, clothing, lighting and cup appearance close to the supplied frame. Natural body movement, realistic hands and no new people.
Start-plus-end prompt
Start from the first supplied frame and move naturally toward the supplied end frame. The camera glides from the wide kitchen view into the closer product composition while the woman reaches toward the machine. Use smooth realistic movement with no sudden jump in pose, scale, lighting or camera direction.
Save a useful frame from a video
- Open the video and pause on the frame you want.
- Hover over the paused frame and select Save frame.
- Use the saved image later as an Ingredient, start frame or end frame.
Create Reusable Characters
Bundle visual and voice references into one named character for repeated use.
When to use a Character
A Flow Character bundles specific visual and audio references into a reusable project asset. It is useful for a fictional presenter, spokesperson or recurring person who appears across several generations.
Create a Character
- Open or create a Flow project.
- On the left, select Characters, then create a new Character.
- Describe the Character clearly.
- Upload or generate one or two images showing the Character's appearance. At least one image is required.
- Select Generate.
- Add the Character's name and choose a voice or create a custom voice.
- Add optional Character information, then select Done.
Use the Character
In the prompt box, type @ followed by the Character's name, then explain what the Character is doing in the shot.
@Riley stands behind a bright studio desk with a tablet resting beside them. Riley looks into the camera and says, “Start with one clear shot. You can build the rest from there.” Medium close-up at eye level. Soft aqua and magenta studio lighting, calm delivery and natural facial movement.
Choose stronger visual references
- Use one or two clear images that agree about the face, hair, age and clothing.
- Avoid multiple people, blocked faces and extreme filters.
- Show enough of the outfit if clothing consistency matters.
- Do not mix images that show conflicting hairstyles, ages or body proportions.
Character or Avatar?
| Use a Character when | Use your Avatar when |
|---|---|
| You are creating a fictional or reusable project character. | You want Flow to use your verified visual and audio likeness. |
| The asset may need its own selected voice. | You want to reference yourself by typing @me. |
Create & Use Your Avatar
Set up your verified visual and audio likeness and reference it safely with @me.
Your Avatar represents you
Flow's Avatar feature creates a highly accurate visual and audio likeness of you. It is different from uploading a casual portrait as an Ingredient.
Create your Avatar
- Open Flow on a computer and enter a project.
- Select your profile picture or initial, then choose Create avatar.
- Select Get started.
- Scan the QR code with a phone or tablet.
- Follow the recording and verification instructions on the mobile device.
Use your Avatar
Type @me in the Flow prompt box and describe the scene, action, camera and delivery.
@me stands in a modern creative studio with a dark background, purple lighting and bright aqua accents. I look directly at the camera and say, “This started as one unfinished idea. Now it is ready to share.” Medium shot at eye level. Slow camera push-in, natural movement, clear speech and a small confident smile.
Privacy and sharing limits
- Your Avatar can be used across your own Flow projects.
- Avatar data cannot be shared through public project links or Flow Tools.
- Flow applies safety and likeness checks to Avatar generations.
- Deleting a generated video does not delete the Avatar itself. Manage the Avatar separately from your profile menu.
Before publishing an Avatar clip
Voices, Dialogue & Audio
Add supported voice references and write dialogue that fits inside a short generated clip.
Voice references depend on the model
Current Flow instructions place voice references under Gemini Omni Flash, Video and Ingredients. Voice references are not available for every type of generation.
Add a preset voice
- Select Gemini Omni Flash in the generation settings.
- Select Video, then Ingredients.
- Select Add, then Voices.
- Preview and choose a voice.
- Reference it in the prompt with the name Flow provides, such as @Voice: Andrew.
Create a custom voice
Choose a base preset voice, name the new voice and describe the performance change you want, such as a slightly raspy quality, a particular accent or a calmer delivery. Preview it, then save it.
Dialogue prompt
Use @Riley as the Character reference and @Voice: Andrew as the voice reference. Riley looks directly at the camera and says, “Choose the shot you need first. Then build the next one from what actually worked.” Warm, clear delivery at a relaxed pace. Natural lip movement and subtle facial expression. Medium close-up with soft studio ambience. No background music over the spoken words.
Write for the available time
- Keep dialogue short enough for the selected 4, 6, 8 or 10-second clip.
- Use one speaker when working with a single-speaker voice reference.
- Place the exact spoken words in quotation marks.
- Describe delivery separately from the dialogue.
- Generate separate shots for a longer conversation.
If audio generation fails
Flow may reject a result when the generated audio quality is too low. Google states that the credits are refunded. Try the prompt again, shorten the dialogue or simplify the sound directions.
Create Videos & Choose a Model
Match the generation method to a model that currently supports it.
Start with the job, then choose the model
Do not choose a model because its name sounds strongest. Choose it because it supports the generation type you need.
| Model | Current strengths | Important limits |
|---|---|---|
| Veo 3.1 Lite | Text, first frame, first + last frames, Ingredients and Extend | Ingredients and Extend use 8-second clips. No video-to-video editing. |
| Veo 3.1 Fast | Text, first frame, first + last frames and Ingredients | Ingredients use 8-second clips. No Extend or video-to-video editing. |
| Veo 3.1 Quality | Text, first frame and first + last frames | No Ingredients, Extend or video-to-video editing. |
| Gemini Omni Flash 1.1 | Text, Frames, Ingredients, video editing, custom voices and clips up to 10 seconds | Extend is not currently supported. Some features vary by region. |
Current clip lengths
- Veo 3.1 text and frame generations: 4, 6 or 8 seconds.
- Veo 3.1 Ingredient generations: 8 seconds.
- Gemini Omni Flash: 4, 6, 8 or 10 seconds.
Create a standard video
- Open a project and describe the scene in the prompt box.
- Open the model settings and select Video.
- Choose Text, Frames or Ingredients.
- Select orientation, number of outputs, model and generation length.
- Add the required references.
- Generate and review the result before building the next shot.
Model check before generating
GENERATION TYPE: Text / Frames / Ingredients / Video Edit NEEDED LENGTH: 4s / 6s / 8s / 10s ORIENTATION: Landscape / Portrait VOICE REFERENCE NEEDED: Yes / No EXTEND NEEDED: Yes / No VIDEO EDIT NEEDED: Yes / No SELECTED MODEL: CURRENT CREDIT COST SHOWN IN FLOW:
Build Consistent Multi-Shot Videos
Use references, frames and a simple continuity plan while accepting that generation can still drift.
Consistency comes from a repeatable process
No prompt can guarantee perfect continuity. The practical goal is to reduce avoidable changes and catch problems before they reach the final sequence.
The four-part consistency workflow
- Prepare references. Choose approved images for the Character, product, environment and style.
- Record continuity. Write down the clothing, location, lighting, important objects and story position.
- Generate one shot at a time. Keep each clip focused on one main action.
- Use successful outputs. Save useful video frames and reuse them where a visual hand-off helps.
Mini continuity sheet
PROJECT: FORMAT AND ORIENTATION: CHARACTER Name/reference: Face and hair: Clothing: Voice: LOCATION Room or outdoor setting: Key furniture or landmarks: Lighting direction and time: PRODUCT OR IMPORTANT OBJECT Reference asset: Size and position: Details that must be checked manually: SHOT 1 Starting position: Action: Ending position: SHOT 2 Starting position: Action: Ending position:
Use the right continuity method
| Need | Useful Flow feature |
|---|---|
| Recurring fictional person | Reusable Character |
| Your own verified likeness | Avatar with @me |
| Product, prop, location or style guidance | Ingredients |
| Continue from a chosen composition | Saved frame used as a start frame |
| Move between two planned compositions | Start + end frames |
| Continue an existing Veo clip | Extend with a compatible Veo model |
Review every clip for drift
- Face, age, hair and body proportions
- Clothing colours and accessories
- Product shape, writing and scale
- Furniture positions and background objects
- Lighting direction and time of day
- Starting and ending story position
Extend & Edit Video Clips
Continue Veo footage or edit a selected part of a video with a compatible model.
Extend and video editing are different
Extend
Adds new footage to the end of an existing Veo-generated clip. Describe how the action should continue.
Video editing
Uses Gemini Omni Flash to change a selected segment of an uploaded or generated video.
Extend a Veo clip
- Open a Veo-generated video in the project.
- Select Extend.
- Describe what should happen next.
- Generate and review the continuation.
Continue from the existing clip. The woman finishes placing the cup beneath the coffee machine, presses the button once and rests her hand beside the cup. Keep the same camera direction, kitchen, clothing, lighting and product appearance. Use natural hand movement and do not introduce new people or objects.
Edit a video with Gemini Omni Flash
- Set the video model to Gemini Omni Flash.
- Upload or choose the video you want to edit.
- For an upload, use a supported format and keep it within the limits shown by Flow.
- Select up to a 10-second segment in the trimming window.
- Describe the change and optionally add Ingredients.
- Generate, then use up to three follow-up editing turns while the edit context remains available.
Current upload limits
- Maximum uploaded video size: 1 GB
- Maximum upload length: 60 seconds
- Videos over 30 seconds must be trimmed to 30 seconds in Flow
- The selected edit segment can be up to 10 seconds
- Supported formats listed by Google: MOV, MP4, AVI and WMV
Earlier versions remain in the History panel. Save a past version back to the project if you want to use it in Scenebuilder.
Assemble Clips in Scenebuilder
Arrange, trim, preview and download a sequence of approved clips.
Scenebuilder joins your clips
Scenebuilder is Flow's built-in sequence area. Use it after the individual clips are approved.
What Scenebuilder currently does
- Arranges multiple clips in one sequence
- Changes the clip order
- Trims the beginning and end with handles
- Previews the full sequence
- Downloads the scene
Add clips to a scene
- Hover over an approved video clip.
- Open the More menu and select Add to Scene.
- Repeat for the remaining clips.
- Open the scene and drag clips into the correct order.
- Trim unwanted frames from the start or end of each clip.
- Preview the sequence with sound.
- Download the completed scene.
Scene review checklist
A practical three-shot structure
Shot 1: Set up
Show the person, product or location clearly.
Shot 2: Change
Show the main action, demonstration or problem.
Shot 3: Payoff
Show the result, reaction or final message.
Use the Real Google Flow Agent
Plan, generate, edit and organise creative media through a conversation inside Flow.
The Flow Agent is a creative collaborator
The Google Flow Agent helps you brainstorm ideas, refine prompts, plan storyboards, create mood boards, generate media, edit selected assets and organise project content.
Turn the Agent on
- Open a Flow project on web or PC.
- Select Agent in the prompt box.
- Expand the side panel if you want the full conversation view.
- Describe what you need or drag project media into the prompt box.
Useful jobs for the Flow Agent
Plan
Turn a rough idea into a storyboard, shot list, mood board or prompt sequence.
Generate
Create images or videos and request multiple variations.
Edit
Change selected images or videos through natural-language instructions.
Organise
Rename selected files, group media in a Collection or delete unused assets.
Agent planning prompt
Help me plan a three-shot vertical product video. The product is the coffee machine already saved in this project. The video needs a clear setup, demonstration and finished result. Keep the same adult presenter, kitchen and morning lighting across all three shots. Before generating anything, give me the shot list, the reference assets needed and a separate prompt for each clip.
Control credit use
Agent chat queries currently have a daily quota but do not use Flow credits. Media the Agent generates does use credits. In Agent settings, choose whether it must always confirm before generating or can generate without stopping.
Sessions and instructions
- Agent conversations are stored as project-specific Sessions.
- You can create, rename and delete Sessions.
- Agent Instructions can hold project-wide creative guidance and a reference image.
- Deleting a Session removes its chat history but does not delete media already generated.
Custom Tools, Final Project & Resources
Build reusable creative mini-apps, complete a final project and keep the official help links.
Custom Tools are reusable creative mini-apps
Flow can build a Tool from a natural-language description. Flow writes the code and creates the interface. Tools can help repeat a creative task, such as applying an image treatment or gathering structured inputs for a generation.
Create a Tool
- Open a Flow project on web or PC.
- Select Tools, then Create Tool.
- Describe the inputs, controls and output you want.
- Generate the Tool and review its interface.
- Select Save when it works as expected.
Create a reusable product-shot planning tool. Ask the user for the product type, audience, platform, orientation, visual style and the one action the product should perform. Return three short shot ideas: setup, demonstration and payoff. For each shot, provide a Google Flow-ready prompt with subject, action, environment, camera, lighting and sound. Do not generate media until the user reviews the three shot ideas.
Share and remix carefully
- You can favourite a Tool for global use or pin it inside a project.
- You can remix a shared Tool into an editable personal copy.
- A shared link exposes the Tool's code, name and thumbnail to anyone with the link.
- A shared link keeps the version that existed when the link was created.
- Stopping Tool sharing currently requires deleting the Tool.
Final course project
- Create one organised project and Collection structure.
- Prepare a presenter or Character, a location reference and one product or key object.
- Plan a three-shot story with setup, change and payoff.
- Generate the clips with the appropriate combination of Ingredients and Frames.
- Save useful frames and correct any continuity problems.
- Assemble the approved clips in Scenebuilder.
- Preview, download and review the finished video before publishing.
Official Google resources
- Get started with Google Flow
- Create videos in Google Flow
- Edit videos and build scenes
- Create and edit images
- Manage projects, assets, collections and Characters
- Models and supported features
- Use the Google Flow Agent
- Create and use your Avatar
- Create and manage Flow Tools