Flow AI - Google Labs AI Video and Image Studio
Flow is Google's AI creative studio, built on Veo 3.1, Gemini Omni Flash, and Nano Banana. It generates video with sound, generates and edits images, keeps the same character across separate shots, and assembles the result into a sequence. It is free to use, with 50 credits a day on any Google account.
Flowaid Ashley · Last reviewed 2026-09-03
What is Flow AI
Flow is Google's AI creative studio, released through Google Labs in May 2025. It generates video with natively generated sound on Veo 3.1, edits existing footage with Gemini Omni Flash, and since February 2026 it generates and edits images too, after Whisk and ImageFX were folded into it. Clips run 4, 6, or 8 seconds on Veo, up to 10 on Omni Flash.
The part that separates Flow from a plain prompt box is that things persist. Save a character, an object, or an avatar of yourself, and call it into any shot with @name, which is the reliable way to keep one face across a whole sequence. Scenebuilder then arranges those clips, trims them, and exports the scene. Add camera controls, a Gemini agent that edits on instruction, and custom tools, and it behaves less like a generator and more like a small studio. It is free to start, with 50 credits a day.
How Flow AI works
Flow works shot by shot. You generate up to eight seconds at a time and extend, rather than trying to prompt an entire sequence at once, which is why continuity holds across a scene.
- 1
Write the shot
Describe one shot, not a whole film. Name the subject, the action, the setting, and the light. Veo 3.1 reads camera language, so "slow dolly in, shallow depth of field, golden hour" changes the output more than adjectives like beautiful or cinematic do.
- 2
Or start from frames
Frames to Video takes a first frame, or a first and a last, and generates the motion between them. When you already know the composition, handing Flow the endpoints beats describing them.
- 3
Reuse your character
Save a character, an object, or your own avatar, then call it with @name in any prompt. Ingredients to Video is what keeps the same face across separate shots, and it is the step most people skip.
- 4
Assemble in Scenebuilder
Arrange the clips you kept, reorder them, trim the heads and tails, preview the sequence, and download it. Extend can add more footage to a Veo clip first, though only on the Lite model.
Features of Flow AI
Four capabilities that separate Flow from a plain text to video generator. Ingredients to Video is the one that matters most, because keeping a character consistent across shots is the hard part of AI video.
Ingredients to Video
Save a character or an object and reuse it across separate shots by name. Google calls this the most reliable way to keep the same face and the same props consistent from one shot to the next, and it is the feature that makes a sequence hang together.
Frames to Video
Supply a first frame, or a first and a last frame, and Flow generates the motion between them. Fixing the endpoints yourself removes most of the guesswork, which is why it usually needs fewer attempts than describing a shot in words.
Native audio
Veo 3.1 generates ambient sound, effects, and dialogue in the same pass as the picture. Put a spoken line in quotation marks in your prompt and it gets spoken rather than described, timed to the action instead of laid over it.
Scenebuilder
Arrange your finished clips into a sequence, reorder them, trim the heads and tails with handles, preview the whole thing, and download the scene. It is where a pile of generations becomes something you can actually watch.
What you can create with Flow AI
Flow produces short video with sound, plus stills since Whisk and ImageFX merged into it. What changes between these users is how much of the shot they plan before generating.
Filmmakers and directors
Previsualise a scene before it costs anything to shoot. Block a sequence with camera controls, keep the same cast across shots with saved ingredients, then assemble it in Scenebuilder and hand a director of photography something concrete instead of a mood board and a description.
Marketers and content creators
Produce short ad spots, product cutaways, and social video without a crew or a licence fee. Native audio means the clip arrives with an ambient bed already matched to the action, not silent footage waiting on a sound pass.
People new to AI video
Start with Frames to Video rather than a text prompt. Supplying a first and last frame removes most of the guesswork, and it teaches you how camera language changes a generation faster than writing longer prompts does.
How to use Flow AI
Your first clip takes a couple of minutes. Flow gives you 50 free credits a day and the full feature set, so the only thing to manage is which model you spend those credits on.
- 1Open Flow on Google Labs and sign in with any Google account.
- 2Pick your model. Veo 3.1 Lite is 10 credits, Fast is 20, Quality is 100, Omni Flash from 4.
- 3Describe one shot, or supply frames, or call a saved character with @name.
- 4Generate. You get 4, 6, or 8 seconds on Veo, with sound already in it.
- 5Assemble the clips you kept in Scenebuilder, then trim and download the sequence.
You can open it here directly on the official Google page. For the full walkthrough, read our how to use Flow AI page. It covers shot prompt structure, the camera terms Veo actually responds to, how to keep a character consistent across a Scenebuilder extension, and where credits get wasted.
Why people use Flow AI
Three things Flow does that most text to video tools still do not.
Consistency
Most AI video tools reinvent the character every generation. Flow lets you save a character, or your own avatar, and call it into any prompt with @name. That is what turns a set of clips into a scene with one person in it.
Sound in the same pass
Veo 3.1 writes the audio with the picture. Ambient noise, effects, and dialogue land timed to the action, which removes the sound editing step that normally follows AI generated footage.
Free to actually try
Fifty credits a day on any Google account, with the full feature set unlocked rather than a crippled trial. That is five Lite generations daily, which is enough to learn whether the tool fits your work before paying anything.
Flow AI examples
Shot types that work well in Veo 3.1. Camera language does more for a generation than style adjectives do.
- Slow dolly in on a subject, golden hour, shallow depth of field.
- Product cutaway with a locked off camera and ambient room tone.
- Frames to Video transition between two still compositions you already shot.
- Dialogue close up where Veo generates the voice with the picture.
- Three shots of the same character, held consistent with a saved ingredient.
- Handheld tracking shot following a character through a doorway.
Check our growing Flow AI examples collection for full shot prompts you can copy, with the camera and lighting terms that produced each result.
Flow AI FAQ
What Flow costs, which feature does what, and the two that people most often mix up.
Is Flow AI free to use?+
Yes. Any signed-in Google account gets 50 Flow credits a day plus 100 starter credits, and the free tier is not a crippled trial: it includes Veo 3.1, Gemini Omni Flash, Nano Banana Pro, the Agent, Custom Tools, Ingredients to Video, Scenebuilder, Characters, Avatars, and 2K image upscaling. Paid plans add monthly credits on top of the same daily 50, starting at $4.99 for Google AI Plus.
How many credits does a generation cost?+
On Veo 3.1: Lite is 10 credits, Fast is 20, Quality is 100. Ultra subscribers pay half on Lite and Fast. Gemini Omni Flash is cheaper, from 4 credits for a 360p four second clip up to 15 for 10 seconds at 720p. So the free 50 a day is five Lite clips, two Fast clips, or a dozen short Omni Flash ones.
How do I keep the same character across different shots?+
Use Ingredients to Video, which Google describes as the most reliable way to reuse the same character and objects from one shot to the next. Save a character, or create an avatar of yourself, then reference it in any prompt by typing @ and the name, or @me for your avatar. This is a different feature from Extend, and it is the one that actually solves continuity.
What is Scenebuilder?+
Scenebuilder is where you assemble finished clips. You arrange them in a sequence, reorder them, trim the start and end of each with handles, preview the whole thing, and download the scene. It is an assembly tool, not the feature that keeps a character consistent, which is a common misunderstanding.
How long can a Flow video be?+
Veo 3.1 generates 4, 6, or 8 second clips. Gemini Omni Flash reaches 10 seconds. Extend appends more footage to an existing Veo clip, but only on the Lite model and only at 8 seconds, which catches people out when they are working on Fast. Longer pieces are built by generating several clips and assembling them in Scenebuilder.
Does Flow generate images as well as video?+
Yes, since February 2026, when Whisk and ImageFX were folded into Flow. Image generation and editing run on Nano Banana, free on the Nano Banana 2 Lite default, with Nano Banana Pro for Ultra. You can generate a still and use it immediately as a frame or an ingredient for a video without leaving the tool.
Does Flow generate audio with the video?+
Yes. Veo 3.1 generates ambient sound, effects, and dialogue in the same pass as the picture. Put a spoken line in quotation marks in the prompt and it gets spoken rather than described. Because the audio comes from the same generation, it matches the action instead of drifting against it.
Can Flow edit video I already have?+
Yes, through Gemini Omni Flash. Upload a clip of up to 60 seconds and a gigabyte, select a 10 second segment, and describe the change in text, refining across up to three conversational turns. Veo models cannot do video-to-video editing; only Omni Flash can.
Ready to try it
If you have been curious, the fastest way to understand Flow AI is to open the official page and try a prompt. Bring a photo you want to restyle, or start from a blank prompt. Either works on the first try, and the second generation usually shows you what to steer next. Ten minutes of playing with it will teach you more than any reading you could do here.