The best AI video generator in September 2026 depends on what you start from and what you can spend. Independent blind votes now put Google's Gemini Omni Flash first for text to video and MiniMax's H3 models first for image to video, with ByteDance's Seedance 2.0 in the top five on both boards. The cheapest daily clips come from Google Flow, and the longest single clips from Kling 3.0 and Seedance. The rankings come first, then a pick for each starting point (a text prompt, a photo, no budget, a phone, a YouTube channel), each with its limits next to its price.
The best AI video generators at a glance
Prices as of September 2026.
| Tool | Model | Free option | Starting paid price | One clip | Good for |
|---|---|---|---|---|---|
| Gemini app | Gemini Omni | None; video needs a Google AI plan | Google AI Plus, $4.99 a month | 10 seconds with audio | Editing a clip by chatting |
| Google Flow | Veo 3.1 and Gemini Omni Flash | 50 credits a day | Google AI Plus, $4.99 a month for 200 extra credits | 4, 6 or 8 seconds (Veo); up to 10 (Omni) | Free daily clips with sound |
| Kling | Kling 3.0 | Free Basic plan with daily credits | Standard, 660 credits a month: $6.99 the first month, then $8.80 | 3 to 15 seconds, up to native 4K | Longer scenes and phone apps |
| Dreamina | Seedance 2.0 | Free daily credits | Basic, $15 a month (first month discounted) | Up to 15 seconds | Scenes built from your photos, clips and audio |
| Runway | Gen-4.5, plus Kling 3.0 and Seedance 2.0 | 125 credits once | Standard, $15 a month or $12 billed yearly | 2 to 10 seconds at 720p (Gen-4.5) | Several models and video editing in one plan |
| Midjourney | Midjourney video | No free trial on the website | Basic subscription, 20% off billed yearly | 5 seconds, extendable to 21 | Animating Midjourney images |
| Wan 2.2 | Wan 2.2 (open source) | Free under Apache 2.0 | $0 | 720p at 24 fps | Running video generation on your own computer |
| Toybox AI Video Creator | Google Veo 3.1 | Sign up free and get 20 credits, not enough for a clip | Pro plan, currently $9.99 a month for 1,000 credits; 350 credits per clip | 4, 6 or 8 seconds at 720p | Video in the same account as images and flyers |
How the top video models rank in blind votes
Artificial Analysis runs a Video Arena where people vote between two clips without knowing which model made them, and it turns those votes into Elo scores. The two boards below rate each clip with its audio. For a verdict from outside the vendors, it's the closest public measure. As of September 28, 2026:
| Rank | Text to video (with audio) | Elo | Image to video (with audio) | Elo |
|---|---|---|---|---|
| 1 | Gemini Omni Flash (Google) | 1233 | MiniMax H3 Max, post-trained by fal | 1194 |
| 2 | Wan 3.0 (Alibaba) | 1229 | MiniMax H3 | 1181 |
| 3 | MiniMax H3 Max, post-trained by fal | 1227 | Gemini Omni Flash (Google) | 1178 |
| 4 | MiniMax H3 | 1220 | Seedance 2.0, 720p (ByteDance) | 1176 |
| 5 | Seedance 2.0, 720p (ByteDance) | 1210 | HiDream-O1-Video | 1174 |
Further down the text-to-video board, Kling 3.0 at 1080p scores 1095, Veo 3.1 scores 1088, and the December version of OpenAI's Sora 2 scores 1083. Three things to keep in mind when reading it. Scores carry a margin of about 6 to 10 points either way, so models a few points apart are tied. The board rates models, not apps, and the same model can cost very different amounts depending on where you run it. And a high rank doesn't cover features the vote ignores, such as clip length, languages or how many reference photos you can add.
Best text to video AI
The best text to video AI on current votes is Gemini Omni Flash, narrowly, since Wan 3.0 sits 4 points behind within the margin of error. You can use Omni Flash two ways. In the Gemini app, which Google says now uses Gemini Omni instead of Veo, you describe a scene and get a 10-second video with native audio, then keep changing it in the same chat. In Google Flow, Gemini Omni Flash clips run 4 to 10 seconds for 7 to 15 credits at 720p, or about half that at 360p for drafts.
For the best text to video AI generator on longer prompts, look at Kling 3.0 and Seedance 2.0. Kling makes 3 to 15 seconds per generation, can cut between shots inside one clip, and voices dialogue in Chinese, English, Japanese, Korean and Spanish. Seedance 2.0 also makes multi-shot clips of up to 15 seconds, with stereo sound, and ByteDance's Seedance 2.5 stretches a single generation to 30 seconds. Google's Veo 3.1, at 1088, is behind on votes but comes with a free daily allowance in Flow, and the explainer on what Veo 3 is covers what it does well.
Best AI image to video generator
The best AI image to video generator on the votes is MiniMax's H3 family, followed closely by Gemini Omni Flash and Seedance 2.0. MiniMax publishes H3's weights on Hugging Face, and its model card points to the Hailuo web app for trying it online. It makes video with stereo sound at up to 2K and 15 seconds.
For most people the practical choice comes down to how many photos you have:
- One photo, or a start and an end photo. Google Flow's "Frames" with Veo 3.1 costs 10 credits a clip with Lite. Kling 3.0 takes start and end frames and can lock a subject so it holds its look for up to 15 seconds.
- Several photos in one scene. Seedance 2.0 on Dreamina combines as many as 9 images with video and audio references in a single prompt.
- Images you already made in Midjourney. Midjourney animates them into 5-second clips at 480p, or 720p from its Standard plan.
Whatever you use, crop the photo to the video's shape first (9:16 or 16:9). The image to video guide covers the steps.
Best AI video generator free, and with no sign up
The best AI video generator free of charge, as of September 2026, is the one with a published daily allowance you can plan around, and that is Google Flow. Flow refills every Google account with 50 credits each day, and at 10 credits per Veo 3.1 Lite clip, that's five short clips with sound. Unused daily credits don't carry over, and in Google's afternoon rush, from about 2 to 5 PM UTC, accounts without a plan may not be able to make videos. Other free routes:
- Dreamina. Free daily credits for Seedance 2.0.
- Runway Free. A one-time 125 credits for testing some of Runway's models, not including Gen-4.5.
- Kling Basic. A free plan with daily credits; its clips carry Kling's watermark, and video extension and commercial use are reserved for paid plans.
- Wan 2.2. Free, open-source software under the Apache 2.0 license.
Finding a free AI video generator with no sign-up runs into the same wall everywhere: hosted tools attach credits to an account, and Flow's allowance belongs to your Google account. The only no-account route is running an open model yourself. Wan 2.2's 5B model needs a graphics card with 24 GB of memory or more (an RTX 4090 qualifies) and outputs 720p at 24 fps.
Best AI video generator app
The best AI video generator app for your phone is the one whose model you already want, since the apps differ more in model than in design:
- Kling. Kling 3.0 in apps for iPhone and Android phones, with desktop downloads for Windows and Mac too.
- Google Flow. Flow's phone app downloads in the same 150-plus countries as the web version, the US and UK among them, and it's for adults only.
- Gemini. Video with Gemini Omni on a Google AI plan, from $4.99 a month.
ChatGPT is no longer an option for video: OpenAI closed the Sora app on April 26, 2026. The Sora alternatives guide covers every replacement. Toybox is a web app: open it in your phone's browser, with no download.
Best AI video generator for YouTube
The best AI video generator for YouTube is one that makes vertical 9:16 clips for Shorts and 16:9 clips for regular videos. YouTube counts a square or vertical video of up to three minutes as a Short, so a Short can hold several AI clips joined together. Veo 3.1, Kling 3.0, Gemini Omni and Toybox's Video Creator all offer both shapes.
YouTube's disclosure rule matters more than the tool. You have to disclose AI content that shows a real person saying or doing something they never did, changes what happened at a real event or place, or creates a realistic-looking scene that never took place. When you upload in YouTube Studio, you answer Yes under "AI use" in the Attributes section, and YouTube then labels the video. Unrealistic content, such as a fantasy creature, and minor edits like color correction don't need it. The guide to AI video for YouTube Shorts covers formats and prompts.
Which one should you pick
Start from the one thing your project can't do without:
- The highest-rated text to video. Use Gemini Omni Flash in the Gemini app or Google Flow.
- No budget. Use Google Flow's 50 daily credits with Veo 3.1 Lite.
- Scenes over 10 seconds, or characters speaking Spanish, Japanese or Korean. Use Kling 3.0, compared in detail in Veo 3 vs Kling.
- A scene assembled from your photos plus an audio clip to match. Use Seedance 2.0 on Dreamina.
- Many models under one bill. Use Runway, or see the Higgsfield alternatives comparison for other multi-model platforms.
- Short clips next to your flyers and product images. Use Toybox Video Creator, and describe the sound you want in the same prompt.
Try a Veo 3.1 clip in Video Creator
Video Creator in Toybox AI is powered by Google Veo 3.1 and makes one clip per run from a description and an optional start photo. Choose Vertical (9:16) under "Format" for Shorts and Reels or Wide (16:9) for YouTube, pick 4, 6 or 8 sec under "Length", and tap "Create video"; clips take a few minutes. A clip currently costs 350 credits at any length on a Pro plan (see Toybox pricing), and if a clip fails, the credits go back automatically. Every clip has sound, and since there's no separate audio setting, describe the sounds in the prompt. Two limits to check against this list: Video Creator makes 720p clips, and it doesn't take audio or video references. A vertical opener for a Short, described for one 8-second shot:
Vertical close-up of a small potted basil plant on a sunny kitchen windowsill. A hand slowly waters the soil, and droplets catch the morning light as the leaves move slightly. The camera eases in from shoulder height and stops on the top leaves. Soft trickle of water and birdsong through the open window. Bright, natural colors, shallow depth of field, a single unbroken take with no captions.