AI Video Generator
The AI Video Generator makes short video clips from a text prompt or from your own images. You choose from 26 leading video models, including Veo 3.1, Seedance 2.5 and Kling 4.0, and many of them add sound to the clip. It is made for creators and marketers who need b-roll, product shots, social clips or scenes for a longer edit.
This guide also covers the Video Translator, which subtitles or dubs a video in 16 languages, and AI Avatar Video, which turns one portrait photo into a talking presenter.
The AI Video Generator needs purchased credits, with one exception: a registered account with a verified email that never made a purchase gets one free 5 second 720p clip with Runway. A clip costs about 3 to 600 credits, depending on the model, the length and the resolution. Runway makes a 5 second 720p clip for 3 credits.
Quick answers
How many models does the AI Video Generator have?
26 video models, including Veo 3.1 (the default), Seedance 2.5, Kling 4.0, Wan 3.0, Vidu Q4 and Runway. Four of them take your own video clips: Kling Motion Control and Wan 2.2 Animate copy the motion of a clip, Kling 4.0 Reference builds a new scene from reference clips and images, and Topaz Video Upscaler sharpens or enlarges a clip.
What is the cheapest AI video clip?
A 5 second 720p Runway clip costs 3 credits, and a 6 second 480p Grok Imagine Video clip costs 4 credits. Both are good for testing a prompt before you run it on a stronger model.
Which models make sound?
Veo 3.1, Kling 4.0, Kling 4.0 Flash, the Seedance models, Wan 3.0, Wan 3.0 Prime and Vidu Q4 always add sound. Kling 3.0, Kling 3.0 Omni and PixVerse V6 have a sound switch. Turning the sound off costs less on Kling 3.0 and PixVerse V6.
What you can make
- A scene from a text description, such as a product on a kitchen table at sunrise.
- An animated version of a photo: your image becomes the first frame of the clip.
- A clip that moves from one image to another, with a first and a last frame.
- Clips with sound effects and ambient sound, with models that make native sound.
- Your character performing the moves of a video you upload, with the motion transfer models.
- A new scene built from your own clips and images, with Kling 4.0 Reference, or a sharper, larger copy of a clip, with Topaz Video Upscaler.
- Vertical clips for TikTok, Reels and Shorts, or wide clips for YouTube.
- Shots for a longer video that you cut together in the AI B-Roll Editor.
How it works
- Open the AI Video Generator. Veo 3.1 is selected by default.
- Pick a model. The controls change to the options that model supports.
- Write your prompt. Describe the subject, the action, the camera and the light.
- Optional: add a reference image as the first frame, and a second one as the last frame.
- Choose the ratio, the length, the resolution and the sound, where the model offers them.
- Press Generate. The button shows the price before you start.
- Your clip appears when it is ready, and it is saved in My Works.
The models
| Model | Best for | Length | Maximum resolution | Sound | Credits per clip |
|---|---|---|---|---|---|
| Veo 3.1 (default) | All-round default, with Lite, Fast and Quality tiers | 8 s | 4K | Always | 8 to 93 |
| Seedance 2.5 | Clips with sound up to 15 s at 1080p | 5, 10 or 15 s | 1080p | Always | 35 to 593 |
| Kling 4.0 | Kuaishou's newest Kling, text or first and last frame | 5, 10 or 15 s | 1080p | Always | 36 to 160 |
| Kling 4.0 Flash | The fast, low-cost Kling 4.0, text or a first frame | 5, 10 or 15 s | 720p | Always | 22 to 64 |
| Kling 3.0 | Sound on or off, the widest choice of lengths | 3 to 15 s | 1080p | Switch | 11 to 102 |
| Kling 3.0 Omni | Sound on or off, 5 to 15 s | 5, 10 or 15 s | 1080p | Switch | 18 to 87 |
| Kling 3.0 Turbo | Faster Kling clips | 5, 10 or 15 s | 1080p | No | 23 to 85 |
| Gemini Omni Video | Up to 7 reference images | 4 to 10 s | 4K | No | 16 to 53 |
| Gemini Omni 1.1 Flash | First and last frame, 4K | 4 to 10 s | 4K | No | 16 to 53 |
| Seedance 2.0 | Clips with sound at a lower price than 2.5 | 5, 10 or 15 s | 1080p | Always | 24 to 383 |
| Seedance 2.0 Fast | Fast clips with sound | 5, 10 or 15 s | 720p | Always | 15 to 93 |
| Wan 3.0 | Low-cost clips with sound | 5, 10 or 15 s | 1080p | Always | 10 to 120 |
| Wan 3.0 Prime | Clips with sound at 1080p | 5, 10 or 15 s | 1080p | Always | 16 to 189 |
| Wan 2.7 | Silent clips | 5, 10 or 15 s | 1080p | No | 20 to 90 |
| Runway | The cheapest test clip | 5 or 10 s | 1080p | No | 3 to 8 |
| MiniMax H3 | Clips up to 2K resolution | 6, 10 or 15 s | 2K | No | 12 to 51 |
| Hailuo 2.3 | Animate one photo (needs 1 image) | 6 or 10 s | 1080p | No | 8 to 23 |
| PixVerse V6 | Low price per second, sound on or off | 5, 10 or 15 s | 1080p | Switch | 7 to 69 |
| HappyHorse 1.1 | Silent clips at 1080p | 5, 10 or 15 s | 1080p | No | 29 to 109 |
| Vidu Q4 | Animate one photo or combine 2 to 15 reference images, up to 16 s, with sound (needs 1 image) | 5, 10 or 16 s | 4K | Always | 9 to 250 |
| Grok Imagine Video 1.5 | Cheap drafts | 6, 10 or 15 s | 720p | No | 4 to 17 |
| Grok Imagine Video (Fast) | Cheap, fast drafts: 6 s at 480p for 4 credits | 6, 10 or 15 s | 1080p | No | 4 to 30 |
| Kling Motion Control | Copy the motion of your clip onto a character image (Kling 3.0 or 2.6) | Length of your clip, 3 to 30 s | 1080p | No | 3.15 to 7.55 per second of your clip |
| Wan 2.2 Animate | Animate a character with your clip (Move) or put it into your clip (Replace) | Length of your clip, up to 30 s | 720p | No | 1.5 to 3.125 per second of your clip |
| Kling 4.0 Reference | A new clip from up to 10 reference images and 5 reference videos, named in the prompt as @image_1 or @video_1 | 5, 10 or 15 s | 1080p | – | 36 to 160 without a video, 43 to 287 with videos |
| Topaz Video Upscaler | Enhance a clip at 1x or upscale it 2x or 4x, no prompt | Length of your clip, 1 to 60 s | 4x the input | No | 2 to 3.5 per second of your clip |
A credit range runs from the shortest, lowest-resolution clip to the longest, highest-resolution clip of that model. Some examples:
- Kling 3.0, 5 seconds, 720p, with sound: 25 credits.
- Runway, 5 seconds, 720p: 3 credits.
- Grok Imagine Video, 6 seconds, 480p: 4 credits.
Options and limits
| Item | Limit / options |
|---|---|
| Models | 26 |
| Clip length | 3 to 16 seconds, depending on the model |
| Ratios | 16:9, 9:16, 1:1, 4:3, 3:4, 21:9, 3:2 or 2:3, depending on the model |
| Reference images | Used as the first frame, and on many models also the last frame; Vidu Q4 takes up to 15 references; up to 10 MB each |
| Models that need an image | Hailuo 2.3, Vidu Q4, Kling Motion Control and Wan 2.2 Animate |
| Models that need a video | Kling Motion Control, Wan 2.2 Animate and Topaz Video Upscaler; Kling 4.0 Reference takes videos optionally |
| Motion clips | Kling Motion Control and Wan 2.2 Animate: one MP4 or MOV clip. Kling: 3 to 30 s (10 s when it follows the image), up to 100 MB. Wan: up to 30 s, 100 MB for Move and 10 MB for Replace. Kling 4.0 Reference: up to 5 clips of 2 to 30 s, 30 s in all, up to 100 MB each. Topaz: one clip of 1 to 60 s, up to 50 MB |
| Native sound | Always on: Veo 3.1, Kling 4.0, Kling 4.0 Flash, Seedance, Wan 3.0, Wan 3.0 Prime, Vidu Q4. Switch: Kling 3.0, Kling 3.0 Omni, PixVerse V6 |
| Draft mode | Seedance 2.5 only: a 480p preview at the 480p price, upgraded to 1080p within 7 days |
| Speed limit | Up to 5 submissions a minute |
| Credits | Purchased credits or invite credits; one free 5 second 720p Runway clip per person before the first purchase |
Pricing
Every clip costs credits. The price depends on the model, the length, the resolution and, on some models, the sound switch. The Generate button shows the exact price before you start.
The AI Video Generator needs purchased credits. Invite credits also work. Free check-in credits do not pay for AI video clips. Any purchase unlocks all 26 models.
The two motion transfer models price by your uploaded clip instead of a length option: the per-second rate times the clip length, rounded up to whole seconds and whole credits. The browser reads the length for the price on the button, and the server reads the stored file again before charging and bills the longer of the two. A 10 second clip on Kling 3.0 Motion Control at 720p costs 57 credits. Try it with Kling Motion Control. Topaz Video Upscaler prices the same way, per second of the clip: 10 seconds at 2x cost 20 credits, at 4x 35 credits (open the Topaz Video Upscaler).
Kling 4.0 Reference prices by the output length you pick. Without a reference video it costs its per-second rate times that length. With one or more videos the rate is lower, and it applies to twice the output length, or to the summed video length plus the output length when the videos are longer; every started second of a video counts. A 10 second clip at 720p costs 72 credits without a video, 85 with a 5 second video and 107 with a 15 second video. Try it with Kling 4.0 Reference.
The one exception is the free first video: one 5 second 720p clip with Runway, from text or one start image, for a registered account with a verified email that never made a purchase. Each person gets it once, so another account, email, browser or network of the same person does not get a second one.
Seedance 2.5 draft mode
Turn on Draft mode under the Seedance 2.5 options to make a cheap 480p preview first. A draft costs the same as a 480p clip: 35 credits for 5 seconds, 70 for 10 seconds and 105 for 15 seconds. When you like the draft, press Upgrade to 1080p under the result. The upgrade renders the same clip again at 1080p and costs the 1080p price of its length: 198 credits for 5 seconds, 395 for 10 seconds and 593 for 15 seconds. The result card shows how many days are left: a draft can be upgraded for 7 days, and the button then disappears. A second press on the same draft opens the upgrade you already started and does not charge again.
A failed clip is refunded automatically. See the credits and pricing guide for plans and packs. Plan prices are in USD, and checkout shows other currencies.
Tips for better results
- Test your prompt on a cheap model first, such as Runway or Grok Imagine Video. When the idea works, run it again on Veo 3.1, Seedance 2.5 or Kling 4.0.
- Start at 720p and the shortest length. Move up to 1080p or 4K only for the final clip.
- On Seedance 2.5, test with Draft mode on, then upgrade the draft you like to 1080p instead of paying for 1080p tries.
- Use a reference image when the look must stay the same. Make one in the AI Image Generator first.
- Write one action per prompt. Describe the subject, the movement, the camera and the light.
- Pick a model with native sound if you need sound effects. Otherwise turn the sound off where the model has a switch, which costs less on Kling 3.0 and PixVerse V6.
- Match the ratio to where you post: 9:16 for TikTok, Reels and Shorts, 16:9 for YouTube.
FAQ
Which model should I start with?
Veo 3.1 is the default and a safe choice. For cheap tests, use Runway or Grok Imagine Video. For a photo you want to animate, try Kling 3.0, Hailuo 2.3 or Vidu Q4.
What happens if a clip fails?
The credits go back to your balance automatically.
Can a free account make AI videos?
One. A registered account with a verified email that never made a purchase can make one free 5 second 720p clip with Runway, from text or one start image. After that, AI video needs purchased credits or invite credits. A free account can still make videos with AI video templates, which free check-in credits can pay for.
Who owns the clips?
You can use the videos you create, subject to the terms of service and the usage policy of the model you chose. You keep all rights to the images you upload.
Can I make videos longer than 16 seconds?
One clip is 3 to 16 seconds. For a longer video, make several clips and cut them together in the AI B-Roll Editor with a script, a voice and captions.
Can I download the clip?
Yes. Every clip is saved in My Works, where you can download, share, view or delete it. See the works, sharing and billing guide.
Video Translator
The Video Translator translates a video of up to 10 minutes into one of 16 languages. You get translated SRT and VTT subtitles, the same subtitles burned into the picture, or a dub in which every speaker talks in the new language with a voice cloned from the original. The dub keeps only the voices: music and background sound are removed.
The Video Translator needs purchased credits or invite credits. Subtitles cost about 1 credit for a 1 minute video and about 5 credits for 10 minutes. A dub costs about 243 credits per minute, and a dub shorter than one minute costs about the same as one minute. If any step fails, the full price comes back.
Subtitles or a dub
Subtitles keep the original voices and add text in the new language. They are the cheapest way to reach a new audience. A dub replaces the speech itself, which suits viewers who do not like to read subtitles, children and small screens.
In a dub, each speaker's voice is cloned from the video and speaks the translation. When a voice cannot be cloned, a similar stock voice speaks instead, and the page tells you. The voices are taken out of the sound track first, so the original speech never plays under the translation; music and background sound are removed with it.
Languages and subtitle formats
The 16 languages are English, Chinese, Spanish, Portuguese, French, German, Japanese, Korean, Hindi, Arabic, Indonesian, Russian, Italian, Turkish, Vietnamese and Polish. The spoken language can be any of these, or the tool detects it.
Subtitles come as SRT and WebVTT. Both work with YouTube, Vimeo, most video editors and the HTML video player. You can also burn the subtitles into the MP4.
What you can make with the translator
- A YouTube video with an SRT caption track in another language.
- A short video with burned-in translated subtitles for apps that do not support caption files.
- A dubbed version of a tutorial, an ad or a talk, in the speaker's own voice.
- Translated versions of one video for many markets.
How to translate a video
- Open the translator. Go to the Video Translator.
- Upload your video. Pick an MP4 or MOV of any size, up to 10 minutes.
- Pick the languages. Choose the spoken language or let the tool detect it, then pick one of 16 target languages.
- Choose the output. Subtitles, subtitles burned in, or a dub. The page shows the price before you start.
- Download the result. Watch it on the page and download the MP4, the SRT or the VTT file. Every translation is saved in My Works. You can close the page while it runs.
Video Translator options and limits
| Item | Limit / options |
|---|---|
| Source | MP4 or MOV with a sound track; a video without sound is refused before any charge |
| Length | 1 second to 10 minutes |
| File size | No limit |
| Target languages | 16 |
| Spoken language | One of the 16, or automatic detection; it must differ from the target language |
| Outputs | Subtitles (SRT, VTT), burned-in subtitles (MP4), dub (MP4) |
| Subtitle cues | At most 6 seconds and 2 lines each, timed to the original speech |
| Dub | Cloned voice per speaker; voices only, music and background sound removed |
| Time | Subtitles usually take 1 to 3 minutes; a dub takes a few minutes per minute of video |
Video Translator pricing
| Output | 1 minute | 5 minutes | 10 minutes |
|---|---|---|---|
| Subtitles | About 1 credit | About 3 credits | About 5 credits |
| Subtitles burned in | About 1 credit | About 4 credits | About 7 credits |
| Dub | About 243 credits | About 1211 credits | About 2421 credits |
The page shows the exact price before you start. Free check-in credits do not pay for translations, and a failed translation is refunded in full.
Tips for translations
- Use clear speech. Loud music under the voice, many people talking at once or a strong echo cause transcription errors, and every error carries into the translation.
- Set the spoken language yourself when the video mixes languages or starts with music.
- Keep the caption file when the platform supports it. A separate track can be switched off and replaced later without a new export.
- For a dub, use one or two clear speakers. That gives the most natural cloned voices.
- Trim a longer video first. Cut a video over 10 minutes into parts, for example with the YouTube Clip Maker.
AI Avatar Video
AI Avatar Video turns one portrait photo into a talking presenter. The face speaks your typed script or your own recording with lip sync. The video is as long as the speech, up to 5 minutes, in 9:16, 16:9 or 1:1, with optional burned-in captions.
AI Avatar Video needs purchased credits or invite credits. Standard 720p costs about 4 credits a second and Pro 1080p about 8 credits a second, so a 30 second Standard video costs about 120 credits. A failed video is refunded.
What you need
One portrait photo (JPEG or PNG, up to 10 MB) and the speech: a typed script read by an AI voice, or your own MP3, WAV or M4A recording of up to 5 minutes.
To speak with your own voice, upload a recording of it, or pick a voice you cloned in the AI B-Roll Editor when you type a script. Voice cloning needs a paid account.
The lip sync follows the sound, so it works for any language that the voice speaks. The AI voices speak English, Chinese, Spanish, Japanese and many more.
Use only your own face, or the face of a person who gave you permission. Do not make a public figure or a private person say something they did not say.
What you can make with an avatar
- TikTok, Reels and Shorts with a steady presenter, without filming every day.
- A product explainer or an offer that you update by changing the script.
- Course lessons and onboarding videos that are easy to fix later.
- The same presenter in other languages, with captions in each language.
How to make an avatar video
- Open the tool. Go to AI Avatar Video.
- Upload a portrait. Pick a sharp, front-facing photo of one person. Choose a vertical, wide or square frame; the photo is cropped to it.
- Add the speech. Type a script of up to 3500 characters and pick a voice, or upload a recording of up to 5 minutes.
- Hear the voice first. With a typed script, you hear the voiceover before you pay for the video.
- Pick the quality and captions. Standard 720p or Pro 1080p, with or without burned-in captions. The page shows the price.
- Download and share. Download the MP4, SRT and VTT files. Every video is saved in My Works.
AI Avatar Video options and limits
| Item | Limit / options |
|---|---|
| Photo | JPEG or PNG, up to 10 MB after cropping, one person |
| Script | Up to 3500 characters |
| Recording | MP3, WAV or M4A, up to 100 MB, 1 second to 5 minutes |
| Video length | As long as the speech, up to 5 minutes |
| Frame | 9:16, 16:9 or 1:1 |
| Quality | Standard 720p or Pro 1080p |
| Voices | AI voices with a free sample to play in the voice list, or a voice you cloned (paid accounts) |
| Captions | Optional, burned in; SRT and VTT files included |
AI Avatar Video pricing
| Video | Standard 720p | Pro 1080p |
|---|---|---|
| 30 seconds | About 120 credits | About 240 credits |
| 60 seconds | About 240 credits | About 480 credits |
| 3 minutes | About 720 credits | About 1440 credits |
| Extra | Typical cost |
|---|---|
| AI voiceover of a typed script | About 11 credits per 500 characters |
| Captions | About 1 to 4 credits |
The page shows the exact price before you start. Free check-in credits do not pay for avatar videos, and a failed video is refunded. If only the captions cannot be burned in, you get the clean video, and the caption files still download.
Tips for avatar videos
- Use a sharp, well-lit photo. The face looks at the camera and is not covered by hands, hair or sunglasses.
- Leave space around the head. The photo is cropped to the frame, so keep room above the head and around the shoulders.
- Use clean audio. Clear speech without loud music gives the most natural lip sync.
- Tell viewers when a presenter is AI. Clear labels keep trust and follow the rules of most platforms.
- Translate it later. Run the finished video through the Video Translator to reach another language.