AI video models, compared

Klakix gives you 10 text to video models and 17 image to video models in one account: Google (Gemini Omni Flash, Veo 3.1 Lite), MiniMax (the Hailuo H3 Max family), ByteDance (Seedance 2.5, 2.0, and 2.0 Mini), and Alibaba (Wan 3.0). Try an idea on a fast, cheap model, then render the final version on a more realistic one. Every option shows its credit cost before you generate.

Clip lengths, resolutions, and credit costs on this page are read straight from our model catalog, so they change when prices do. You pay in US dollars.

Model	Tool	Tier	Good for	Credits
Seedance 2.5ByteDance	Text to Video	Best	A single shot up to 30 seconds long, or up to 25 seconds in Full HD. The most ambitious model in the lineup, and the most expensive.	147
Seedance 2.0ByteDance	Text to Video	Balanced	The more affordable Seedance: video with sound up to 15 seconds, up to 4K, at a balanced price.	95
Seedance 2.0 MiniByteDance	Text to Video	Fast	The cheapest Seedance: quick, short clips with sound for social media.	48
MiniMax H3 Max TurboMiniMax	Text to Video	Fast	Fast and cheap, good enough for social posts. Generates sound too.	20
MiniMax H3 MaxMiniMax	Text to Video	Balanced	A clear step up in motion and detail. Ranks near the top of independent tests for video with sound.	40
Google Gemini Omni FlashGoogle	Text to Video	Best	The most realistic results, with synced sound. Made for ads and promo videos.	50
Google Veo 3.1 LiteGoogle	Text to Video	Balanced	The more affordable version of Google's video model. Generates dialogue and ambient sound along with the picture.	40
Wan 3.0Alibaba	Text to Video	Balanced	A single shot up to 30 seconds long, in Full HD. Good for longer scenes.	100
Kling 4.0Kuaishou	Text to Video	Balanced	Kling's newest model: a single shot up to 30 seconds, always with stereo sound. 720p or Full HD, including cinematic 21:9.	79
Kling 3.0Kuaishou	Text to Video	Balanced	Kling's newest video model: up to 15 seconds, with or without sound. Standard is 720p, Pro is 1080p. Built-in speech is English or Chinese.	63
Seedance 2.5ByteDance	Image to Video	Best	A single shot up to 30 seconds long, or up to 25 seconds in Full HD. The most ambitious model in the lineup, and the most expensive.	147
Seedance 2.0ByteDance	Image to Video	Balanced	The more affordable Seedance: video with sound up to 15 seconds, up to 4K, at a balanced price.	95
Seedance 2.0 MiniByteDance	Image to Video	Fast	The cheapest Seedance: quick, short clips with sound for social media.	48
Seedance 2.5 ReferenceByteDance	Image to Video	Best	Add up to 30 photos, 10 videos and 10 audio clips as references: product, place, motion and music come together in one video.	116
Seedance 2.0 ReferenceByteDance	Image to Video	Balanced	Up to 9 photos, 3 videos and 3 audio clips as references, combined into one video. Up to 4K.	84
Seedance 2.0 Mini ReferenceByteDance	Image to Video	Fast	The cheapest way to use multiple references: up to 9 photos, 3 videos and 3 audio clips, for quick social clips.	42
Seedance 2.5 Video EditByteDance	Image to Video	Best	Upload your video and describe the change: remove an object, swap the background or the product. Everything else stays the same.	154
Seedance 2.5 Video ExtendByteDance	Image to Video	Best	Picks up from your video's last frame and continues the scene for up to 30 seconds. You write what happens next.	154
MiniMax H3 Max TurboMiniMax	Image to Video	Fast	Brings a photo to life quickly and cheaply.	20
MiniMax H3 MaxMiniMax	Image to Video	Balanced	Best at keeping faces and details intact. Recommended for old photos.	40
Google Gemini Omni FlashGoogle	Image to Video	Best	The highest realism, with synced sound. For your most important projects.	50
Google Veo 3.1 LiteGoogle	Image to Video	Balanced	The more affordable version of Google's video model. Generates dialogue and ambient sound along with the picture.	40
Wan 3.0Alibaba	Image to Video	Balanced	A single shot up to 30 seconds long, in Full HD. Good for longer scenes.	100
Kling 4.0Kuaishou	Image to Video	Balanced	Kling's newest model: a single shot up to 30 seconds, always with stereo sound. 720p or Full HD, including cinematic 21:9.	79
Kling 4.0 ReferenceKuaishou	Image to Video	Balanced	Up to 10 photos and 5 videos as references: keeps the character, product, place and motion consistent in one video. A single shot up to 30 seconds, with sound.	79
Kling 4.0 Video EditKuaishou	Image to Video	Balanced	Upload your video and describe the change: background, weather, lighting, colors, outfit or camera move. Everything else stays the same.	94
Kling 3.0Kuaishou	Image to Video	Balanced	Kling's newest video model: up to 15 seconds, with or without sound. Standard is 720p, Pro is 1080p. Built-in speech is English or Chinese.	63
Kling effectsKuaishou	Effects	Fast	Ready-made templates like heart hands, birthday, money rain and pet dance.	28
PixVerse effectsPixVerse	Effects	Balanced	Popular transformation and dance effects that work from a single photo.	20
MiniMax H3 Max Lip SyncMiniMax	Talking Photo	Fast	More natural facial movement, with a high-resolution option.	74
HeyGen Avatar IVHeyGen	Talking Photo	Balanced	The person in the photo says your text with natural expressions. It's fast.	57
OmniHuman 1.5ByteDance	Talking Photo	Best	The most realistic face and body movement. Made for promos and ads.	146
HeyGen Avatar IVHeyGen	UGC Ads	Balanced	Fast and natural. Keeps the presenter's face consistent, with lively expressions. Good enough for most ads.	62
OmniHuman 1.5ByteDance	UGC Ads	Best	The most realistic body language and gestures. For ads that are going live.	154
Muse ImageMeta	Images	Fast	Very cheap and fast. Good for testing ideas and making lots of images.	2
Nano Banana 2Google	Images	Balanced	High quality with accurate lettering. The best pick for putting text on an image.	8
GPT Image 2.5OpenAI	Images	Best	OpenAI's newest image model, at the top of independent comparisons. Excellent at fine detail, realism and text inside the image.	25
Voice v4Klakix	Voiceover	Balanced	The newest, most expressive model. Steer the delivery with tags like [excited], [whispers] and [laughs].	2
Voice v4 TurboKlakix	Voiceover	Fast	Half-price, very fast version of v4. Understands tags; great for long scripts and drafts.	1
Voice ClassicKlakix	Voiceover	Best	The most consistent read with every fine control: style, similarity, clarity and speed. For corporate narration and audiobooks.	2
Voice v3Klakix	Voiceover	Balanced	The previous expressive generation. Understands audio tags; try it for a different tone than v4.	2
Voice FlashKlakix	Voiceover	Fast	The fastest and cheapest. Good for drafts and very long scripts.	1
Multi-voice DialogueKlakix	Voiceover	Balanced	A scene with two or more speakers: a voice per line, with laughs and pauses in between. Podcasts, sketches, ad dialogue.	1
Sound effectKlakix	Voiceover	Balanced	Generates the sound you describe: a creaking door, applause, rain, a whoosh. Royalty-free audio for your videos.	2
Music ComposerKlakix	Voiceover	Best	Composes the music you describe: ad background, intro, or a full song with vocals. Cleared for commercial use.	18
Kling 3.0 Motion ControlKuaishou	Motion Transfer	Balanced	Transfers the dance or motion in a clip onto the character in your photo. 720p, up to 30 seconds.	63
Kling 3.0 Motion Control ProKuaishou	Motion Transfer	Best	The same motion transfer in 1080p, with steadier faces and hands. For videos you'll post.	84
Topaz Video AITopaz Labs	Upscaler	Balanced	Upscales your video to Full HD or 4K and cleans up noise and compression artifacts. Smooth 60 fps if you like.	20
MotifKlakix	Motif Video	Best	Type a topic, pick one of 46 templates and get a 30–90 sec video with voiceover, captions and motion graphics, drawn in code. Text and numbers come out exactly right.	159
Viral ClipsKlakix	Viral Clips	Balanced	Upload a long video and the AI picks the strongest moments, then cuts vertical clips with face tracking and word-by-word captions, ready for Reels, TikTok and Shorts.	7

Credit costs use each model's default settings (suggested length and resolution). Change the length or resolution and the cost changes too.

Model families

Find the model you're looking for

Versions change fast, so each page covers a model family rather than a single version number.

Kuaishou

Kling AI

Kling AI is the video model from Kuaishou, the Chinese short-video company. Kuaishou launched Kling 4.0 on October 15, 2026, and it's available on Klakix: turn text or a photo into a single shot up to 30 seconds long with sound, combine several photos and videos as references, or edit a clip you already have. A 5-second 720p Kling 4.0 clip costs 79 credits, with no subscription required. Kling 3.0 and the Kling effects are here too.

Google

Google Veo

Veo is Google's model for turning text and photos into video with sound. Veo 3 launched in May 2025, Veo 3.1 in October 2025, and the cheaper Veo 3.1 Lite in March 2026. On Klakix, use Veo 3.1 Lite with no subscription: 4 to 8 seconds in 720p or 1080p. An 8-second 720p clip with sound costs 40 credits.

ByteDance

Seedance

Seedance is ByteDance's video model family; ByteDance also owns TikTok. The newest version, Seedance 2.5, makes clips with sound up to 30 seconds in a single take. On Klakix you can use 2.5, 2.0, and 2.0 Mini, plus Reference, Edit, and Extend models, with no subscription. A 5-second 720p clip on 2.5 costs 147 credits.

MiniMax

Hailuo AI

Hailuo AI is the video brand of MiniMax, an AI company based in Shanghai, and its current model since July 2026 is MiniMax H3. On Klakix, H3 Max and H3 Max Turbo turn a prompt or a photo into a 5, 10, or 15 second clip with sound. A 5-second clip at 480p starts at 13 credits.

Alibaba

Wan AI

Wan AI is Alibaba's video model family, also called Tongyi Wanxiang. Wan 2.1 and 2.2 are open source, 2.5 added built-in sound, and Wan 3.0 became generally available on Aug 24, 2026. On Klakix, Wan 3.0 turns a prompt or photo into video with sound up to 30 seconds; 5 seconds at 480p costs 25 credits.

AIsphere

PixVerse

PixVerse is an AI video platform from AIsphere, a company founded in Beijing in April 2023. Its flagship model, PixVerse V6, came out in March 2026. On Klakix you can use PixVerse's ready-made effects from a single photo: a 5-second 720p clip costs 20 credits, with no subscription.

ShengShu Technology

Vidu

Vidu is the AI video model from ShengShu Technology, a Beijing company founded in March 2023. Its newest versions, Vidu Q2 and Q3, make video from text, photos, and reference images. On Klakix, Vidu runs only inside 7 ready-made photo templates, from 20 credits per clip; you can't write your own Vidu prompt here.

HeyGen

HeyGen is a US-based AI company that turns a photo into a talking avatar video. Its own site has a watermarked free plan and paid plans from $29 a month. On Klakix you can use HeyGen Avatar IV to make a portrait say your script: up to 70 characters costs 32 credits, with no subscription.

Google

Nano Banana

Nano Banana is the name of Google's Gemini image models: the original in August 2025, Nano Banana Pro in November 2025, and Nano Banana 2 on February 26, 2026. Nano Banana 2 is Gemini's default image model, known for clean, readable text in images. On Klakix, you can use it with no subscription, from 8 credits per image.

OpenAI

GPT Image

GPT Image 2.5 is the image model OpenAI released on September 8, 2026, the same generation as Images 2.5 in ChatGPT. The API offers two versions: Flare, the faster default, and Sunburst, which spends longer on fine detail. Klakix uses Sunburst, which ranked first on the Artificial Analysis text-to-image leaderboard as of September 23, 2026. No subscription needed.

Topaz Labs

Topaz Video AI

Topaz Video AI is Topaz Labs' video upscaler, now sold as the desktop app Topaz Video. On Klakix you run its Proteus model in the browser instead: upload a clip up to 30 seconds and pick Full HD or 4K. A 10-second 720p clip costs 13 credits in Full HD, with no GPU or subscription.

OpenAI

Sora

Yes, Sora is gone. OpenAI shut down the Sora app and website on April 26, 2026, and set September 24, 2026 as the date sora-2 and sora-2-pro were removed from its API. OpenAI named no replacement. Klakix never offered Sora, but it has text-to-video models with sound, like Veo 3.1 Lite, Gemini Omni Flash, and Wan 3.0.

Real outputs

One sample from each model

All made on this site, each labeled with the model that made it. Tap a video to enlarge it and turn on the sound.

A product photo and a location photo in one ad: a white sneaker steps into a puddle on a rainy Istanbul… Still from a video made with Seedance 2.5 Reference. — Seedance 2.5 Reference · 5 secA product photo and a location photo in one ad: a white sneaker steps into a puddle on a rainy Istanbul street, with a slow-motion splash, neon reflections and the camera circling the shoe.

An epic sunrise over Cappadocia: hundreds of colorful hot air balloons drift above the fairy chimneys, golden… Still from a video made with Seedance 2.5. — Seedance 2.5 · 5 secAn epic sunrise over Cappadocia: hundreds of colorful hot air balloons drift above the fairy chimneys, golden sunbeams cut through the morning haze, and a crane shot rises to reveal the valley. Rich detail, large-format film look.

A cat running the register at a corner shop: in its tiny apron, it taps the cash register with a paw. Still from a video made with Seedance 2.0. — Seedance 2.0 · 5 secA cat running the register at a corner shop: in its tiny apron, it taps the cash register with a paw

A ferry crossing the Bosphorus with seagulls overhead, golden hour light. Still from a video made with MiniMax H3 Max Turbo. — MiniMax H3 Max Turbo · 5 secA ferry crossing the Bosphorus with seagulls overhead, golden hour light

A tin robot walks the waterfront carrying a tray of sesame bread rings, with seagulls in pursuit. Still from a video made with Seedance 2.0 Mini. — Seedance 2.0 Mini · 5 secA tin robot walks the waterfront carrying a tray of sesame bread rings, with seagulls in pursuit

A cinematic aerial shot over Istanbul's Bosphorus at blue hour: the bridge lights flicker on and a ferry… Still from a video made with Google Veo 3.1 Lite. — Google Veo 3.1 Lite · 8 secA cinematic aerial shot over Istanbul's Bosphorus at blue hour: the bridge lights flicker on and a ferry leaves a glowing wake as the camera glides forward. Anamorphic lens, teal and amber grade, crisp modern cinema look. A distant ferry horn and seagulls.

Hot air balloons filling the sky over Cappadocia at sunrise, the camera slowly rising. Still from a video made with MiniMax H3 Max. — MiniMax H3 Max · 5 secHot air balloons filling the sky over Cappadocia at sunrise, the camera slowly rising

Turkish coffee foaming in a copper pot, then poured into a cup, macro shot. Still from a video made with Google Gemini Omni Flash. — Google Gemini Omni Flash · 5 secTurkish coffee foaming in a copper pot, then poured into a cup, macro shot

Sunset on the Galata Bridge: anglers casting their lines, the New Mosque and ferries behind, the camera… Still from a video made with Wan 3.0. — Wan 3.0 · 10 secSunset on the Galata Bridge: anglers casting their lines, the New Mosque and ferries behind, the camera moving slowly along the bridge. Seagulls and waves.

The baker pulls warm loaves out of the oven with a wooden peel, steam rises, then he smiles at the camera. Still from a video made with Kling 4.0. — Kling 4.0 · 5 secThe baker pulls warm loaves out of the oven with a wooden peel, steam rises, then he smiles at the camera. With bakery sounds.

A character photo and a product photo in one video: the woman walks through an autumn park and drinks coffee… Still from a video made with Kling 4.0 Reference. — Kling 4.0 Reference · 5 secA character photo and a product photo in one video: the woman walks through an autumn park and drinks coffee from the thermos, and both stay true to the photos.

We turned the Cappadocia video into winter: the fairy chimneys and the valley are covered in snow, and the… Still from a video made with Kling 4.0 Video Edit. — Kling 4.0 Video Edit · 5 secWe turned the Cappadocia video into winter: the fairy chimneys and the valley are covered in snow, and the balloons float along the same paths while the camera move stays the same.

The chef ladles lentil soup into a bowl as steam rises, then smiles at the camera. Still from a video made with Kling 3.0. — Kling 3.0 · 5 secThe chef ladles lentil soup into a bowl as steam rises, then smiles at the camera. Kitchen sounds.

Choosing

Which model fits your project?

Nothing to download: every model runs in your browser on the same account. Three questions usually settle it.

1. How long does the clip need to be?

Most models do their best work at 5 to 10 seconds. If you need a longer single shot, Wan 3.0 and Seedance 2.5 go up to 30 seconds; Seedance 2.0 and the MiniMax H3 family stop at 15. Longer clips cost more and give the model more room to lose the scene.

2. Is there a real face in the frame?

If your photo shows a real person, the Seedance models won't accept it; pick MiniMax H3 or the Google models instead. Seedance is strong with landscapes, products, places, and objects.

3. Testing an idea, or making the final cut?

While you explore, the fast tier is enough to see several variations cheaply. For ads and anything you'll publish, realism matters more, so the best tier pays off. Running the same prompt on both shows exactly what the difference buys you.

Specs: length, resolution, sound

Models in the text to video and image to video tools. A model that appears in both tools gets one row.

Model	Length	Resolution	Sound	Input and notes
Seedance 2.5ByteDance	4, 5, 6, 8, 10, 12, 15, 20, 25, 30 sec	480p / 720p / 1080p	Yes, can be turned off	Text or a photoUp to 25 sec in Full HD · Accepts an end frame · Doesn't accept photos of real faces
Seedance 2.0ByteDance	4, 5, 6, 8, 10, 12, 15 sec	480p / 720p / 1080p / 4K	Yes, can be turned off	Text or a photoAccepts an end frame · Doesn't accept photos of real faces
Seedance 2.0 MiniByteDance	4, 5, 6, 8, 10, 12, 15 sec	480p / 720p	Yes, can be turned off	Text or a photoAccepts an end frame · Doesn't accept photos of real faces
MiniMax H3 Max TurboMiniMax	5, 10, 15 sec	480p / 768p / 1080p	No	Text or a photoAccepts an end frame
MiniMax H3 MaxMiniMax	5, 10, 15 sec	480p / 768p / 1080p	No	Text or a photoAccepts an end frame
Google Gemini Omni FlashGoogle	5, 8, 10 sec	360p / 720p / 1080p / 4K	No	Text or a photoAccepts an end frame
Google Veo 3.1 LiteGoogle	4, 6, 8 sec	720p / 1080p	Yes, can be turned off	Text or a photoAccepts an end frame
Wan 3.0Alibaba	5, 10, 15, 20, 30 sec	480p / 720p / 1080p	Yes, can be turned off	Text or a photoAccepts an end frame
Kling 4.0Kuaishou	3, 5, 8, 10, 15, 20, 25, 30 sec	720p / 1080p	No	Text or a photoAccepts an end frame
Kling 3.0Kuaishou	3, 5, 8, 10, 15 sec	720p / 1080p	Yes, can be turned off	Text or a photoAccepts an end frame
Seedance 2.5 ReferenceByteDance	4, 5, 6, 8, 10, 12, 15, 20, 25, 30 sec	480p / 720p / 1080p	Yes, can be turned off	Photo, video, and audio references plus textUp to 25 sec in Full HD · Doesn't accept photos of real faces
Seedance 2.0 ReferenceByteDance	4, 5, 6, 8, 10, 12, 15 sec	480p / 720p / 1080p / 4K	Yes, can be turned off	Photo, video, and audio references plus textDoesn't accept photos of real faces
Seedance 2.0 Mini ReferenceByteDance	4, 5, 6, 8, 10, 12, 15 sec	480p / 720p	Yes, can be turned off	Photo, video, and audio references plus textDoesn't accept photos of real faces
Seedance 2.5 Video EditByteDance	sec	480p / 720p / 1080p	Yes, can be turned off	A video to edit plus text; optional photoDoesn't accept photos of real faces
Seedance 2.5 Video ExtendByteDance	4, 5, 6, 8, 10, 12, 15, 20, 25, 30 sec	480p / 720p / 1080p	Yes, can be turned off	A video to extend plus text; optional photoUp to 20 sec in Full HD · Doesn't accept photos of real faces
Kling 4.0 ReferenceKuaishou	3, 5, 8, 10, 15, 20, 25, 30 sec	720p / 1080p	No	Photo, video, and audio references plus text
Kling 4.0 Video EditKuaishou	sec	720p / 1080p	No	A video to edit plus text; optional photo

FAQ

About the models

What's the best AI video model?

There's no single answer; the job decides. For long single shots, Wan 3.0 and Seedance 2.5 stand out. For photos with real faces, the MiniMax and Google models work where Seedance doesn't. For ready-made effects, Kling's templates are the easy pick. Running the same prompt on two models is the quickest way to compare.

Which model leads in independent tests?

On Artificial Analysis's blind-vote leaderboard (September 23, 2026), Gemini Omni Flash ranked 1st for text to video with audio, Wan 3.0 2nd, and MiniMax H3 Max 3rd. For image to video with audio, MiniMax H3 Max ranked 1st. Seedance 2.0 was 5th on both audio lists, and Seedance 2.5 wasn't listed yet. All of these models are available here.

Which models make video with sound?

Every model in the text to video and image to video tools generates ambient sound along with the picture. Veo 3.1 Lite, Wan 3.0, and the Seedance family (2.5, 2.0, and 2.0 Mini) also let you turn sound off. For a person speaking your script, use Talking Photo or UGC Ads, which voice the script and lip-sync it.

Which model makes the longest clips?

Wan 3.0 and Seedance 2.5 give the longest single clip: up to 30 seconds. Seedance 2.5 goes up to 25 seconds in Full HD. The MiniMax H3 Max family and Seedance 2.0 go up to 15 seconds, Gemini Omni Flash up to 10, and Veo 3.1 Lite up to 8.

Which models don't accept photos of people?

The Seedance family (2.5, 2.0, 2.0 Mini, plus its reference, edit, and extend models) rejects photos that contain real human faces. For portraits and family photos, choose MiniMax H3 Max, Veo 3.1 Lite, or Gemini Omni Flash. Seedance works well with landscapes, products, places, and food.

Hailuo or Seedance?

Hailuo (MiniMax H3) for photos with faces and cheap short clips; Seedance for long single shots of products, places, and landscapes without people. Seedance is the priciest family here: at today's prices, 5 seconds costs 40 credits on H3 Max, 95 on Seedance 2.0, and 147 on Seedance 2.5.

Can I make 4K video?

Google Gemini Omni Flash and Seedance 2.0 both offer 4K. The other video models top out at 1080p, and Seedance 2.0 Mini at 720p. Higher resolution costs more credits. You can also take any clip to 4K with the AI video upscaler.

Learn more: AI video upscaler

How much does it cost to try a model?

The cheapest try is a 5-second 480p clip on MiniMax H3 Max Turbo: 13 credits ($0.47–$0.87). Video and image models unlock with your first credit purchase; the smallest option is the Mini pack, 75 credits for $5. No subscription is required.

Do I need to install anything?

No. The models run on the provider's servers and you use them from your browser, on a phone or a computer. Finished videos download as MP4 files.

Sources

The models' official pages and an independent leaderboard. Klakix isn't affiliated with any of these companies; we offer access to their models through API providers.

Hailuo AI (MiniMax)Official site of the H3 family
Seedance 2.0 (ByteDance Seed)Official model page
Seedance 2.5 (ByteDance Seed)Official model page
Veo (Google DeepMind)Official page of the Veo family
Video generation in GeminiGoogle's Gemini Omni overview
Wan (Alibaba)Official site of the Wan models
Kling AI (Kuaishou)Kling's official site
Artificial Analysis: text to videoIndependent blind-vote leaderboard
Artificial Analysis: image to videoIndependent blind-vote leaderboard

Try the models yourself

Signing up is free. Watch the samples first, then buy credits and compare models side by side.

Get started