Flow

tvWatch Flow TV
help_outlinedFlow Help Centremore_vert

Flow Help Centre

What is Flow?

Built with and for creatives, Flow is the only AI filmmaking tool custom-designed for Google Deepmind's most advanced models – Veo, Imagen and Gemini. With Flow, storytellers can create cinematic clips, seamlessly transition them into scenes and have enough consistency to tell a story. Simply prompt how you speak, craft your narratives and easily organise all of your videos and prompts in one place.

Available exclusively for Google AI subscribers

Flow is available exclusively for Google AI Pro and Ultra subscribers. Ultra subscribers also get first access to our latest models and features like Ingredients to video. Learn more here.

Best experienced on desktop, Chromium browsers

Flow is currently best experienced on desktop with a Chromium-based browser, like Google Chrome. We'll be optimising the experience for other browsers and mobile surfaces shortly. Please bear with us!
  • To get you started, you can find a quick tutorial here, or can continue reading below.

    Thinking of a scene? Create a project for it.

    This is where your clips will live. Come back to a project to find all your prompts and videos.

    Generate clips using Text to video, Frames to video or Ingredients to video.

    • Text to video = Generate using a text prompt
    • Frames to video = Upload or generate images to use as starting and/or ending frames
    • Ingredients to video = Upload or generate images to use as subject or style references

    Decide what happens next in Scenebuilder.

    Once you find a clip that you like, you can add it to Scenebuilder to easily create the next shot(s) with the power of Gemini behind the scenes to help with consistency

    • Extend = Continue the action seamlessly. Flow takes the last 24 frames of your generation into account to smoothly continue the shot.
    • Jump to = Transition to a new shot while preserving context from the last frame. This allows you to put a character in a new situation for example, while keeping the look and feel of the previous shot.

    Tip: The scenebuilder resets when you leave your project, but you can always find all your clips in the clips view. We'll add a scenebuilder-saved state soon.

    Share your masterpiece

    When you're done generating your scene, download it! Additionally, upscale your clips to 1080p or download as GIFs for easy sharing.

    Tip: Not done yet? All clips and prompts will be saved as a project for you to return to later.
  • Flow provides a range of features powered by various Veo models under the hood, in addition to Imagen 4 and Gemini, that you can see below
    Table showing Flow's capabilities, mainly that the Fast model (powered by Veo 2) is capable of almost everything but is bested at a few things by the Quality model (powered by Veo 2) and Highest Quality model (powered by Veo 3).

    * Requires first frame or first and last frame

    ** For best results, use Quality or Highest Quality.

  • See details on how to use AI credits, how much they are worth and more in the Google One Help Centre.
  • As described above, we offer three tiers of models (Fast, Quality, Highest Quality) but some features only exist for certain models.

    If we detect that you're trying to use a feature that is not available in your selection, we'll default you to a compatible model. You can check the model and related credit costs in settings before confirming.

  • Here are three things to remember to get the best out of Flow

    Prepare your ingredients

    Clean ingredients make it easier for the model to capture what you care about. When you're using Ingredients to video and add a subject or product reference, it can help if they are on a plain or segmented background.

    Avoid conflicting guidance

    Your text prompts should reference the frames or ingredients that you're pointing to. In Ingredients to video specifically, you can point to either a style or up to three subjects, locations or objects.

    Make sure that:

    • Location, style references avoid extra subjects (unless desired)
    • You have a consistent look across ingredients to help with blending

    Don't hesitate to ask Gemini for help too!

    Iterate on your prompt with Gemini! As a Google AI Pro subscriber, you get access to the most capable Gemini models. Visit gemini.google.com and send Gemini your prompt, an image or even video and ask it to write you a new prompt, riff on ideas and more.

    It's early days for Flow. If you get blocked or find a mistake anywhere – please do share feedback with the flag button in the three-dot menu so that we can iterate on it. We value your input!
  • Veo 3 is Google Deepmind's latest video generation model, with improved visual quality, prompt adherence and audio co-generation. Veo 3 supports:

    • Text to video, with speech and environmental sound.
    • First frame to video, with environmental sound.

    Note that Veo 3 is not yet compatible with first + last frame, camera controls, extend or Ingredients to video capabilities. We will default you back to Veo 2 models when you want to use these features.

  • You can now add sound effects, background noise and – in some cases – speech. Simply weave your guidance into the prompt itself. e.g. 'The sound of an ice cream van can be heard in the background', 'the captain turns towards the ocean and says "we set sail at daybreak!"'

    Please note that this is a new experimental feature, with certain caveats

    • It's only available with Veo 3.
    • Speech is only available in Text to video mode
    • Audio might not generate in certain cases.

    If you have any feedback, please use the flag button (in the three-dot menu) so that we can iterate on it.

  • All outputs generated in Flow using Veo and Imagen include invisible SynthID watermarks. SynthID identifies AI-generated content by embedding digital watermarks directly into AI-generated content.

    As an additional step to help people identify AI-generated content, videos generated by Google One Pro users will contain a visible watermark indicating the content was made with Veo. Videos generated by Google One Ultra users will not contain this visible watermark.

  • We're working to bring Flow to as many people as possible. See below for a list of countries that the site is currently available to users aged 18+ in.

    American Samoa, Angola, Antigua and Barbuda, Argentina, Australia, Bahamas, Belize, Benin, Bolivia, Botswana, Brazil, Burkina Faso, Cabo Verde, Cambodia, Cameroon, Canada, Chile, Côte d'Ivoire, Colombia, Costa Rica, Dominican Republic, Ecuador, El Salvador, Fiji, Gabon, Ghana, Guam, Guatemala, Honduras, Jamaica, Japan, Kenya, Laos, Malaysia, Mali, Mauritius, Mexico, Mozambique, Namibia, Nepal, New Zealand, Nicaragua, Niger, Nigeria, Northern Mariana Islands, Pakistan, Palau, Panama, Papua New Guinea, Paraguay, Peru, Philippines, Puerto Rico, Rwanda, Senegal, Seychelles, Sierra Leone, Singapore, South Africa, South Korea, Sri Lanka, Tanzania, Tonga, Trinidad and Tobago, Türkiye, US Virgin Islands, Uganda, United Kingdom, United States, Uruguay, Venezuela, Zambia and Zimbabwe.

  • Flow currently only supports English prompts, with more coming soon.
  • If you were a previous early access user of VideoFX and your history is enabled, your generations are in My library. Click on your profile icon in the top right, then select My library from the drop-down.
  • We want the Flow experience to be fun and safe, therefore, we prohibit users from knowingly generating harmful, illegal or inappropriate content. For example, Flow features precautions to protect minors and certain queries might not be generated. For more details on categories of content and our policies, please see our Generative AI Prohibited Use Policy.

    If you're facing difficulties or getting blocked in ways that were unexpected, please do send us feedback by flagging output or the in-app feedback (three-dot menu at the top right) and help make our product better!

  • Flow TV is an ever-growing showcase of clips, channels and content generated with Google's Veo generative model. Use the remote control at the bottom of the screen to browse Channels and see the prompt behind the clip.

    Hit the Home button to visit the channel guide or watch short films created using Flow, alongside other tools and techniques. All generations were created with Veo 2.

    Flow TV does not collect your data. For details on Google policies, see our Privacy Policy and Terms of Service.

    Flow TV visuals were generated by filmmakers and creatives using a variety of techniques. Output will look different with every generation.

  • If you want to share your work or connect with like-minded creatives, join the #Flow channel in our Discord.
  • For information about your user data and more, please take a look at labs.google/fx's FAQs.
Thank you for trying Flow!

Flow 6/3/25 Update

~ Thank you for using Flow ~
Here’s a few things we’ve fixed and updated, and some best practices to get the most out of the tool. Please continue to leave us feedback anytime in the three dot menu!

Veo 3

The excitement for our latest model, Veo 3, has been overwhelming!
Please remember that:
  • Veo 3 currently only supports text-to-video and first-frame to Video.
  • We’re working hard on bringing extend, first + last frame, fast mode and more to Veo 3 soon!
  • The full Flow experience is available on Veo 2, which is why it remains the default setting.
  • If you select a feature not yet supported by Veo 3, we’ll let you know we’re moving you to Veo 2 to complete your request.

Audio availability

Audio is still an experimental feature, and available on Veo 3 only. We’re working on improving coverage but your generated videos might not always have sound. Things to know:
  • Speech does better with longer transcripts.
  • Speech is muted on minors and Frame to Video (we’re working on it).
  • Speech can trigger subtitles (we’re working on it).

Bugs & Improvements

  • Thumbnails: fixed issues around previewing videos and missing thumbnails.
  • Scenebuilder: added sound support for Veo 3 clips, Veo 3 “Jump To” now available, fixed broken multi-clip behaviors.
  • Upscaling: Upscales to 1080p are free. Fixed sound on Veo 3 upscales.
  • Downloading: Fixed reliability of downloading outputs from scene builder.
  • Image uploads: now supports WebP and HEIC file formats, fixed chance rejections.

Other

  • Flow access has expanded to 70+ countries
  • Veo 3 generations are now available in Pro tier.