The AI video generator
Type a scene. Choose who films it.
The studio runs every major text-to-video model side by side - Seedance 2.5 with synced audio and clips up to 30 seconds, Veo 3.1 in 4K, Kling v3 Pro, LTX Video, and the unrestricted Wan family through Wan 3.0 plus Happy Horse. Aspect ratios from 21:9 cinema to 9:16 vertical, priced per clip before you render.
Free messages to start · no card · no identity attached
Every clip above was generated in the Studio - open one as a template and remix it with your own prompt.
The lineup
Every major text-to-video model. One picker.
Every model has a temperament, and the picker badges them so you can read it at a glance. Don't want to choose? The Quality selector (Draft, Balanced, Best) and a filter level from standard to zero filters pick the model for you.
Seedance 2.5
Best OverallThe narrative flagship: up to 30 seconds in one generation with synced dialogue, and a 10,000-character brief it actually reads. Give it a production note, not a one-liner.
Veo 3.1
4KWhen the shot has to pass for camera footage. Reads technical language - lenses, camera moves, lighting - better than anything else in the picker.
Kling v3 Pro
Hi-FiThe body-mechanics specialist: when people have to move convincingly - sports, dance, fight choreography - Kling keeps weight and joints honest.
Wan 3.0
New · UnrestrictedOne continuous take up to 30 seconds with audio, and no provider-side prompt moderation. It reads plain visual description and doesn't fill gaps - say exactly what happens.
Happy Horse 1.1
Most UnrestrictedThe least filtered model in the Studio, with widescreen support - for the legal briefs everything else bounces.
Wan 2.2 14B
Draft laneFind the shot before you pay for it: run cheap 480p drafts, change one thing per pass, then carry the winning prompt up to a flagship.
One studio, the whole pipeline
From a prompt to a finished, edited video.
Most tools stop at the clip. The Studio's home screen reads as a pipeline - Create, Quick start, Post-production - and every stage is one click deep.
Direct it, or type it
Describe the clip to the Director and it picks the model and writes the prompt with you - or type it yourself in Text to Video and let the Quality and Filters selectors choose the model.
Text to VideoStart from an image or a recipe
Animate a still in Image to Video - Seedance 2.5 pins up to 30 reference images - drop in a ready-made face from Character Studio, or fill in a Recipe: UGC ads, product spots, logo stings.
Image to VideoStoryboard multi-scene stories
Give Storyboard one premise and it plans 3-6 scenes with keyframes you approve before a second of video is paid for - then animates the set with matched cuts.
StoryboardCut, upscale, export
The Movie Editor puts your clips on a timeline - trim, split, reorder, music, crossfades - and exports one MP4 up to 1080p in your browser. Upscale sharpens it 2x or 4x for delivery.
Movie EditorFree credits on sign-up · price shown before every render
The model lottery
Every text-to-video model has a personality. Pick wrong, re-render all day.
One model nails physics but muffs faces. Another does gorgeous light and ignores your camera direction. Locked into a single tool, you discover its blind spots one failed render at a time - on your budget.
Single-model tools hide the tradeoff
You can't know a model is wrong for your brief until you've paid to find out - and switching tools means new accounts, new subscriptions, new credit systems.
And they all share one door policy
Mainstream tools moderate the same way, so an awkward brief fails everywhere equally. Without an unrestricted lane, some scenes are simply unrenderable.
Prompts people bring to the studio
Here's how OpenGradient is different
One prompt box, the whole market
Audition models, keep the winner
Run the same prompt on two or three models for cents and compare - motion, faces, adherence. The picker shows strengths, badges and prices.
Audio, styles, upscaling
Seedance, Veo, Kling v3 and the newer Wan generations render synced dialogue and effects; style presets from cinematic to claymation shape the look; 2x/4x upscaling finishes it for delivery.
An unrestricted lane
Wan and Happy Horse run without provider-side prompt moderation - clearly badged, 18+ capabilities gated - so edgy-but-legal briefs still render.
Free messages to start · no card · no identity attached
The anonymity layer
Every model, and none of them know you.
Not a policy promise - a protocol. No single party, including us, ever holds both your identity and your question.
Read the full security architectureEncrypted on your device
Messages are encrypted locally before they leave your browser.
Anonymized by Oblivious HTTP
Network relays split who you are from what you're asking. No one sees both.
Unlocked only in a secure enclave
The model reads your prompt inside attested hardware that never learns who sent it.
Questions, answered
Frequently asked
Which text-to-video model is best?
For overall quality with audio, Seedance 2.5. For maximum resolution, Veo 3.1 in 4K. For hi-fi motion, Kling v3 Pro. For unrestricted briefs, Wan 3.0 or Happy Horse 1.1. Don't want to choose? Two selectors pick for you: Quality (Draft, Balanced, Best) and Filters (Standard, Uncensored, Zero filters) - Best + Standard is Seedance 2.5, Zero filters is always Happy Horse.
How long can clips be?
Up to 30 seconds in one continuous take on Seedance 2.5 and Wan 3.0, with 8-15 seconds typical elsewhere - and for anything longer, generate scenes and join them in the Movie Editor.
How do I write text-to-video prompts that work?
Like a note to a film crew, not a list of tags: who is in frame and what they do, where the camera is and whether it moves, what the light is doing - and if the model renders sound, what to hear. Match the register to the model: technical for Veo, plain and explicit for Wan, a long production brief for Seedance 2.5, which gives you 10,000 characters and rewards every one. The full prompt guide
What does it cost?
Per clip, by model, duration and resolution - the price sits on the Generate button before you click. Seedance 2.0 at 5 seconds, 720p with audio, is 988 credits (about $0.99); a 5-second Wan 2.2 14B draft is 88 credits. 1,000 credits = $1, free credits on sign-up, no card, and credits never expire.
Give the scene to three directors
Free to start. Type one prompt, render it on two models, and learn more about text-to-video in five minutes than a month of one-tool loyalty would teach.
Unrestricted means no provider-side prompt moderation on the open models - not no rules. Illegal content is banned, explicit capabilities are gated 18+, and sexual content depicting real people without their consent is prohibited.