The fastest way to make AI videos is to open a video tool, describe one shot in a sentence or two, and generate a clip of 4 to 10 seconds; with a photo, the photo becomes the first frame and you describe how it should move. Below are five tools side by side, the five-part prompt formula from Google's Veo guide, steps for text-to-video and photo-to-video, and how to build a longer film, follow YouTube and TikTok labeling rules, and fix a clip that misses.
5 ways to make AI videos
Most of the options with a free allowance come from Google, and one clip from any of them is short. As of September 2026:
| Method | Works on | Cost | Best for |
|---|---|---|---|
| Google Flow | A Google account, age 18 or older, in supported countries | Free daily allowance of 50 credits; Veo 3.1 Lite clips cost 10 each | Text or photo to video, start and end frames, extending clips |
| Gemini app | The Gemini app, signed in, 18 or older | A Google AI plan, from $4.99 a month in the US | Making and editing a 10-second clip by chatting |
| YouTube app | "Create video" for Shorts, where offered | Built into the app | Clips of up to 10 seconds for a Short |
| Google Photos | Photo to video in the Google Photos app | Capped per day; a Google AI plan raises the cap | Animating a photo you already have |
| Toybox Video Creator | Any browser, on a Pro plan | Currently 350 credits per clip, or 180 to 220 in Video Creator Lite | 16:9 or 9:16 clips from text or a photo, with start and end frames |
ChatGPT isn't on the list: OpenAI discontinued its Sora video app on April 26, 2026. The answer to whether ChatGPT can make videos covers what it can still do for a video project.
How to write an AI video prompt
A good video prompt reads like one line of a shot list: one camera setup, one subject, one action. Google's Veo 3.1 prompting guide builds prompts from five parts: cinematography, subject, action, context, and style and ambiance. In order:
- Cinematography. The shot and the camera: close-up, wide shot, slow pan, tracking shot, static camera.
- Subject. Who or what the viewer watches: "an older fisherman in a yellow raincoat".
- Action. One thing that happens in the clip: "coils a wet rope on the deck".
- Context. Where and when: "a small wooden boat in a gray harbor at dawn".
- Style and ambiance. The look and mood: "overcast light, muted colors, documentary style".
Put together, and ready to adapt for any tool:
Medium close-up, slow handheld camera. An older fisherman in a yellow raincoat coils a wet rope on the deck of a small wooden boat, then looks out at the water. A gray harbor at dawn with fog over the masts. Overcast light, muted colors, calm documentary style.Tools that make sound read cues in the same prompt: dialogue in quotation marks, and sound effects and ambience labeled as such. The guide to AI video with sound shows the cue format, and ready-made Veo 3 prompts give you scenes to adapt.
How to make an AI video from text in Google Flow
Google Flow is the no-cost place to start: anyone 18 or older in a supported country gets 50 Flow credits a day. These steps follow Google's Flow Help:
- Open Google Flow with your Google account.
- Open an existing project or create a new one.
- Describe the scene in the prompt box: the subject, what happens, the place, the light and the look.
- Click the model name and choose Video. The prompt box opens on an image model, so skipping this step gets you a picture instead of a clip.
- Set the shape, how many videos to make, the model and the clip length. Veo 3.1 Lite clips cost 10 credits each, so set one output while you test.
- Click Generate, then watch the result and adjust one part of the prompt at a time.
Without a Google AI plan, Flow may refuse video requests between about 2 PM and 5 PM UTC. Daily credits don't roll over, and the next 50 arrive a day after your first generation.
How to make a video with AI from a photo
Photo-to-video starts from your image and adds motion, so the clip opens on your photo instead of a scene the model invents. It's the better route for a product, a pet, a family picture or a flyer. In Google Flow:
- In a project, click the model name in the prompt box, then choose Video, then Frames.
- Drag your photo to "Add start frame". For a fixed last shot, drag a second image to "Add end frame".
- Describe the action or the move between the frames, then set your preferences and click Generate.
Google Photos has a one-tap version called Photo to video, which makes six-second portrait clips from pictures already in your library. The image to video AI page covers it, along with motion prompts for portraits, products, pets and drawings.
Whatever the tool, crop the photo to the video's shape first. A 4:3 phone photo in a 9:16 video leaves the tool to fill or cut the difference.
How to make a movie with AI
A movie made with AI is many short clips edited together, so the work is planning shots and keeping them consistent. One generation lasts 4 to 10 seconds in the tools compared here, so a longer film comes from joining or extending clips.
- Write a shot list. One line per clip: the shot size, who's in it, what happens, how long. A one-minute film is 8 to 15 clips.
- Design the characters once. Make or pick one image for each character and setting, and reuse it. Flow's "Ingredients" take reference images, and Google's Gemini API documentation says Veo 3.1 takes up to three reference images, such as a person, a character or a product, and uses them to keep that subject's look in the video.
- Chain the shots. Take a still from the last moment of one clip and use it as the start frame of the next. Start and end frames in Flow and in Toybox's Video Creator make the handoff predictable.
- Extend where the tool allows it. Flow can extend 8-second Veo 3.1 clips using Veo 3.1 Lite. In the Gemini API, each Veo 3.1 extension adds 7 seconds, 20 extensions is the limit, and the finished video can reach 148 seconds.
- Time the action inside a clip. Google's Veo 3.1 guide splits one 8-second prompt into timed beats, each starting with a bracket like [00:02-00:04].
- Edit it like a film. Assemble the clips on an editing timeline, trim the first and last half-second of each where motion starts and stops, then add music, voices, sound effects and titles.
How to make AI videos for YouTube and TikTok
Make vertical 9:16 clips for Shorts and TikTok and 16:9 for regular YouTube videos, then follow each platform's AI rules. Square and vertical uploads of up to three minutes count as Shorts on YouTube, and a clip of up to 10 seconds can be made inside the YouTube app with "Create video". The page on the AI video generator for YouTube Shorts covers Shorts specs, hooks and thumbnails.
- YouTube disclosure. Realistic AI-made or AI-altered content needs the "AI use" box set to "Yes" in YouTube Studio. Fantasy scenes and behind-the-scenes help like an AI draft of your script don't.
- YouTube monetization. Channels can't earn from AI videos that look mass-produced from a template with nothing of the creator's own added, or from slideshows and scrolling text with little narrative.
- TikTok labels. TikTok's Community Guidelines say creators must label AI-made or heavily edited videos that show realistic people or scenes, and TikTok may take down, limit or label a video that skips it.
- Music. YouTube says tracks from its Audio Library won't receive a copyright claim when used in Shorts. Anywhere else, use only music you're licensed to use.
How to make AI videos for free
To make a video with AI free of charge, you work within daily allowances, not unlimited use. As of September 2026, Google Flow's 50 daily credits cover five Veo 3.1 Lite clips or four 8-second Gemini Omni Flash clips at 720p (12 credits each). Google Photos makes six-second clips from your photos up to a daily limit. The YouTube app's "Create video" is built in where it's offered, outside the European Economic Area. Google AI Plus, at $4.99 a month in the US, adds 200 Flow credits a month and video generation in the Gemini app.
Tips for a clean result
- One action per clip. "She pours the coffee" is a clip. "She pours the coffee, answers the phone and leaves" is three.
- Say what stays the same. "Keep his face, jacket and the street behind him unchanged" guards the details that matter most.
- Keep the camera move simple. One move per shot: a slow push in, a pan, or a static camera.
- Leave text out. Put titles, prices and captions on in your editor, where they stay sharp.
- Match shapes. Start images, clips and the final video should all be 16:9 or all 9:16.
- Change one thing per retry. When a clip misses, edit one part of the prompt, so you learn what fixed it.
Troubleshooting
| Problem | Likely cause | Fix |
|---|---|---|
| The clip skips half of what you asked for | Too many actions for a few seconds | Split the prompt into one action per clip |
| A face or product changes during the clip | A large motion or camera move | Use a static camera, a shorter clip and a start frame |
| Words in the video are garbled | Text drawn by the model | Add text in your editor instead |
| The framing is cut off | The start image's shape doesn't match the video | Crop the photo to 16:9 or 9:16 first |
| Characters look different in every clip | Each clip was generated from scratch | Reuse the same start image or reference images |
| The generation is blocked | A safety filter, or a problem with the generated audio, stopped it | Rephrase the prompt, and take out real names, brands and logos |
| Flow won't make a video at certain hours | Peak-hour limits for accounts without a plan | Generate before 2 PM or after 5 PM UTC |
| The clip is silent or the sound is wrong | The tool doesn't make audio, or the prompt had no sound cue | Add sound cues (dialogue in quotes, SFX, ambience), or add sound in your editor |
Make an AI video in Toybox Video Creator
Toybox AI's Video Creator makes one clip at a time from a description, optionally starting on your photo and ending on a second image. It's powered by Google Veo 3.1, makes 720p clips, and needs a Pro plan. For lower-cost drafts, Video Creator Lite uses Veo 3.1 Lite and offers 720p or 1080p.
- Sign in, then open Video Creator. It's included with Pro and higher plans.
- Type the shot into "Describe your video". You have up to 2,000 characters: name the camera, subject, action, setting and style, plus any dialogue or sounds you want. Or tap the mic to speak your prompt (Chrome, Edge and Safari).
- Add a photo if the clip should open on one. Use "Start from a photo", then put a second image in the End frame slot if you want to fix the final frame too.
- Pick a Format and a Length. Wide (16:9) suits regular YouTube videos and Vertical (9:16) suits Shorts and TikTok. Lengths are 4, 6 or 8 sec, and an end frame sets the clip to 8.
- Rule things out in the description. Describe what the scene contains instead of listing bans, as Google's Veo 3.1 guide advises: "a quiet country road with no other cars" rather than "no vehicles".
- Press "Create video". Expect a wait of a few minutes per clip; if one fails, its credits are refunded automatically.
- Review the clip at full screen before you download the MP4. Unsaved results stay available for 24 hours. Save each clip you plan to edit.
Video Creator currently charges 350 credits for a clip of any length, and Video Creator Lite 180 credits at 720p or 220 at 1080p. Clips from both come with sound. There's no separate setting for it, so put the dialogue, effects and music you want into the description. See current prices.