As a one-time aspiring filmmaker, I still like to play around with video editing software, but more recently, I've been experimenting with AI services that allow you to create your own videos.
Google's Flow can generate videos using text prompts or uploaded images. You describe every detail about the scene, camera movements, and angles, and add sound effects, background noise, and even speech. Flow then creates a video with multiple clips; select the one you like, then build, extend, and fine-tune it from there. Here's how I use Flow to create movie magic.
What You Need to Use Flow
To use Google Flow, you need a subscription to the $20-per-month Google AI Pro plan or the $250-per-month Google AI Ultra plan. I have an AI Pro subscription, which provides text-to-video, frame-to-video, scene builder, sound effect, dialogue, and limited camera control features. The Ultra plan introduces fewer restrictions to these features, plus advanced camera controls, more consistent video quality, and more credits, which are used to perform tasks.
To create your videos, Google allows you to use its Veo 2 model or the more advanced Veo 3, which supports sound effects and dialogue. Flow breaks down the models into four versions: Veo 2 Quality, Veo 2 Fast, Veo 3 Quality, and Veo 3 Fast. The two Quality models generate higher-quality videos but take longer to run and use up more credits. The two Fast models are of lower quality but don't take as long to run and use up fewer credits.
Set the Model and Number of Videos
You should be able to create your video from any desktop browser, but Google advises you to use a Chromium-based one, such as Chrome or Edge. Log in to the Flow website with your Google account and click New project.
Click the Settings icon to the right of the prompt. Here, you can change the number of videos generated and which model to use. The default number of two is a good choice. For the model, I typically use one of the Fast ones to limit the number of credits used and then turn to a Quality model only for a video I want to publish or share. Unless I plan to add sound effects or dialogue, I also typically opt for Veo 2 Fast over Veo 3 Fast as it chews up fewer credits.
Create a Video From a Prompt
Since I plan to describe the video using text, I keep the option set to Text to Video. I then type a description of the video I want to generate. I'll sometimes use an AI, such as ChatGPT, Copilot, or Gemini, to create a prompt for me based on details that I provide. After adding my description, I click the arrow to submit it. For this one, I submit the following description:
"Create a photorealistic and cinematic video capturing a tense, epic aerial battle high above New York City at midday. The shot is a dynamic tracking medium close-up following a determined female superhero in a sleek, dark blue and silver suit as she dodges an energy blast. Shimmering golden energy crackles around her clenched fists. Her adversary is a massive, insectoid robot with a glinting metallic carapace and multiple glowing red optical sensors, which transforms its limbs into plasma cannons. The bright, high-contrast sunlight casts sharp shadows, emphasizing the action-packed and dangerous confrontation against a backdrop of towering skyscrapers."
Flow then creates the number of videos I specified. I can click the Play button for each video to watch it. Hovering over any video displays a floating toolbar. From there, I can click the Download button to save the video in its original size, in an upscaled format, or as an animated GIF. Clicking the Fullscreen button lets me watch the video full screen. Clicking the three-dot icon and selecting Delete allows me to get rid of the video.
If you want to generate a video with sound effects and dialogue, you'll need to select New project at the main screen and choose one of the Veo 3 options. While describing the scene I want, I have to specify the sounds and dialogue I want to include. For the sounds, I can just write something like "soft jazz music is playing in the background." For the dialogue I have to identify which character should be speaking and put the words in quotes, such as "The man says to the woman: "Will you marry me?" The woman then says to the man: "No, I won't."
For this one, I submitted the following description:
"A giant, monstrous creature with glistening scales and menacing eyes slowly emerges from the dark, churning ocean, its immense form towering over a tiny, vulnerable boat carrying a man and a woman who stare up in disbelief and terror. The sound of powerful, crashing waves against the boat is audible, accompanied by a deep, ominous musical score that builds in intensity as the behemoth continues its ascent from the depths. As the monster reaches its full, terrifying height, the man, with a tremor in his voice, looks at the woman and says, "I think we need a bigger boat," to which she replies, with a sense of hopeless resignation, "I don't think that will help."
Flow then generates my video. As before, I can play each one, view any of them full screen, download them, or delete the one I don't want to keep.
Create a Video From an Image
Beyond writing a text description, you can use still images to create your video, either by generating them within Flow or uploading them. To do this, I returned to the main screen, chose New Project, picked the model, and then changed the option above the prompt from Text to Video to Frames to Video. I clicked the plus (+) icon at the bottom and chose to upload a photo from my computer.
After adding the first image, I clicked the plus (+) icon again and added a second one. The first one will be the starting point for the video, and the second will be the ending point. In this case, I chose two photos of me sitting in the captain's chair of the Enterprise at a Star Trek set recreation. I then add a description of the video I want generated. Glow won't let you use Veo 3 with images, so I have to make this a silent movie using Veo 2.
Since Flow is relying heavily on the still images, I can keep the description brief and simple. I entered and submitted the description as follows:
"A medium shot of a confident and determined male captain in his late 30s, sitting in the iconic command chair on the bridge of the Federation starship USS Enterprise. He is wearing a classic gold Star Trek captain's uniform from the Original Series era, reminiscent of the one worn by Captain Kirk. The camera slowly pushes in on the captain's face, capturing his focused gaze."
Manage Your Videos
Finally, you can return to the main screen to manage all your generated videos. From here, you can click any video to view, rename, or delete it. There's more you can do in Flow, including choosing a camera angle when you use images for your video and using the scene builder to combine multiple scenes in one video. But the information I shared here should serve as a good primer for getting started with the tool.
Can Google Be Your Cinematographer?
And how did Google Flow fare in generating the videos I requested? The videos themselves all looked great, very cinematic and stylish. But AI being imperfect, glitches did crop up. Flow didn't always follow every detail in the prompt. The mouths of characters didn't always move when speaking. And certain transitions in the videos weren't always as smooth as expected. With time and tweaking, I could probably fix those errors, but as a first attempt in each case, Flow turned out to be a fairly adept filmmaker.
