FirstLast
MiniMax H3
Image to video from an opening frame, a closing frame, or both.
About MiniMax H3Image to video from your own frame
The frame decides the look
In image to video the picture you upload is the opening frame: its subject, lighting and colors carry into the clip. Your prompt only has to say what changes — the motion, the camera and the sound. Describing what is already visible wastes words.
- JPEG or PNG, up to 4 MBThat is what the image to video upload box accepts on this site.
- One clear subjectA cluttered frame gives image to video too many things to move.
- Match the shape you wantVeo 3.1 image to video asks for 16:9 or 9:16 and crops other shapes (fal's note).
How many frames each model takes
Image to video works with one opening frame on every model here. Only some also take a closing frame, which pins down where the shot ends.
First—
Veo 3.1
Image to video from an opening frame only; the end of the shot comes from your prompt.
About Veo 3.1First—
Veo 3.1 Fast
Image to video from an opening frame only; the end of the shot comes from your prompt.
About Veo 3.1 FastNeed both ends fixed? Use MiniMax H3. For a 1080P image to video clip with generated sound, try Veo 3.1 or the cheaper Veo 3.1 Fast.
Continue a clip with image to video
One clip is one shot, but shots can be chained. When a clip you generated here finishes, capture its final frame and send it straight back in as the first frame of the next image to video request. The next shot starts exactly where the last one stopped, and your library keeps the chain in order.
Write the change, not the picture
“A woman in a red coat on a bridge” repeats the frame. “She turns toward the river as the camera drifts left; wind, distant traffic” tells image to video what to do with it. The four-part shot structure from text to video still applies to image to video — skip the subject, keep action, camera and sound.
Image to video FAQ
Image to video takes a still you upload as the opening frame and animates it from your prompt: the motion, the camera and the sound. The look of the clip comes from the image.
JPEG or PNG files up to 4 MB. For Veo 3.1, fal recommends 720p or larger in 16:9 or 9:16; other shapes are cropped.
On MiniMax H3, yes — an opening frame, a closing frame, or both. Veo 3.1 and Veo 3.1 Fast take an opening frame only.
It keeps the look of the opening frame, but the frame is a starting point, not a still that stays put: image to video moves the subject and the camera as your prompt describes, and some models crop shapes they do not support.
Yes. Capture the final frame of a clip you generated here and use it as the first frame of the next image to video request. Your library keeps the shots in order.
Related pages
- H3 Video GeneratorCreate an H3 video with native audio and exact per-second credit pricing.
- H3 Max explainedA visual, evidence-led guide to the speed claim, quality rankings, real costs, limits and the Infinite Slop connection.
- Infinite Slop explainedWatch the official channel, understand the public queue and see why endless AI television is an airtime product — not one long video.
- AI Video MakerCreate an H3 video from a prompt, with native audio and the exact credit cost shown before submission.
- Text to videoWrite one shot — subject, action, camera, sound — and render it on the model that fits.
- Video to promptUpload a clip and get the structured prompt that recreates it — camera, cut timing and soundscape included.
- Zombie love storyTwo photos become a 15-second zombie couple drama — the standoff, the embrace and the memory, with sound.
- Gangnam selfie danceThree photos become the trio selfie dance: center, left, right, then a point at the lens, on the beat.
- All video toolsEvery tool on the site, grouped by what you start from: words, a frame, or a clip.
- Video promptsA library of complete prompts — copy one, or load it into the generator in a click.
- BlogEvery article in one place: model explainers, prompt guides and local-setup walkthroughs.
- PricingOne-time credit packs from $9.99 with no subscription or free allowance.