Decide where the shot begins and where it lands. Upload two keyframes, describe the journey between them, and Yovid renders every frame in the middle on Kling 3.0, Veo 3.1, Seedance 2.5 and more.
Give a video model one photo and it decides on its own how the clip ends. Give it a closing keyframe as well and the ending is no longer a guess: the model only has to invent the camera path, the motion and the change of light that join your two pictures. Reach for it whenever a clip has to finish on an exact composition, such as a reveal, a product turn, a location change or a before-and-after.
Animate one still and let the model choose where the shot goes. Best for bringing a single photo, poster or product image to life.
Image to video AIFeed in photos of a character, an outfit or a product as references, and the model stages a brand-new shot around them.
Reference to video AIUpload the clip and continue it from its final moment with the same people, set and sound.
AI video extenderImage one sets the composition the clip starts on, image two the composition it must reach. In this Seedance 2.5 run the camera opens on leaf shadows over white curtains and finishes on the model beside the stream, with a sweep of gauze across the lens doing the hand-off.
Seedance 2.5The two keyframes can be different places entirely. Google’s Veo 3.1 example begins outside a barn at sunrise and ends on a rider crossing the prairie; the model works out the push through the barn and the spill of light from the far doors on its own.
Veo 3.1Eleven models in the Yovid composer accept a keyframe pair. Kling 3.0 and Veo 3.1 are the go-to choices for tight transitions, Seedance 2.5 and Wan 3.0 for long ones, and Seedance 2.0 Mini or Seedance 1.5 Pro for a first try on free credits.
Every number below is read from the live generator settings, so the table moves when a model changes.
| Model | Clip length | Top resolution | End frame alone | Audio switch | On free credits | Credits per second |
|---|---|---|---|---|---|---|
 Seedance 2.5 | 4–30 s | 1080p | No | Yes | No | 6.1 at 720p |
| Seedance 2.0 | 4–12 s | 1080p | No | Yes | No | 4 at 720p |
| Seedance 2.0 Fast | 4–15 s | 720p | No | Yes | No | 3.75 at 720p |
| Seedance 2.0 Mini | 4–15 s | 720p | No | Yes | Yes | 2.5 at 720p |
| Seedance 1.5 Pro | 4–12 s | 1080p | No | Yes | Yes | 1.2 at 720p |
 Kling 3.0 | 4–15 s | 4K | No | Yes | No | 2.5 at 720p |
| Kling O3 | 4–15 s | 1080p | No | Yes | No | 2.5 at 720p |
 Veo 3.1 Fast | 4, 6 or 8 s | 4K | No | No | No | 3 at 720p |
| Veo 3.1 Pro | 4, 6 or 8 s | 1080p | No | No | No | 8 at 720p |
 Wan 3.0 | 2–30 s | 1080p | No | Yes | No | 2.4 at 720p |
 MiniMax H3 | 5–15 s | 2K | Yes | No | No | 4.55 at 2K |
Credits per second are catalogue rates with both keyframes attached, before plan discounts; the composer prints the exact cost for your settings before you generate. The end frame is free: each rate equals that model’s image to video rate. Free credits stop at 720p, Seedance 1080p needs Pro, Premium or a credit pack, and 4K needs Premium or a credit pack. A “No” under Audio switch means sound is not yours to toggle: Veo 3.1 and MiniMax H3 add their own.
Use cases
Move from an exterior to an interior, day to night or one set to the next in a single camera move instead of a hard cut.
Start on the packshot, end on the hero angle, and let the model rotate, open or light the product on the way.
Renovations, makeovers, seasons and edits read best when both states are pinned and the change plays out between them.
Two storyboard panels become a moving beat, so clients and crews can judge pacing before anything is shot.
Step 1
Same aspect ratio, light from the same side, the same subject or place. Photos, renders and AI images all work as JPG, PNG or WebP.
Step 2
In the Image to Video workspace, add the opening image, then the closing one. On Kling, Veo and Seedance 1.5 Pro, switch the frame menu to First + Last Frames first.
Step 3
Describe the camera path and what changes along it, pick a length and resolution, check the credit cost and press Generate.
Shared lighting, lens and style give the model a path it can actually film. Two unrelated pictures tend to melt into each other instead.
“The camera pushes through the doorway into the hall” is a path; “go from image 1 to image 2” is not. Name the move, the speed and the ending.
A small turn or a light change fits in 4 to 6 seconds. Crossing a room, a street or a season needs 8 seconds or more.
List the logo, the face, the building or the product that may not change, and ask for a single take with no cuts when you want one.
Use one image as both keyframes and describe a motion that comes back around, and the last frame meets the first for a seamless loop.
MiniMax H3 can work backwards from a closing image with no opening frame. Every other model needs the first frame before it accepts a last one.
Related tools
Keyframes are one input among several. Switch to whichever fits the material you already have, in the same account and credit balance.
Prompt guide6 image to video prompt examplesStarting frames, the prompts behind them and what each one produced.Read guide
Model comparisonSeedance 2.5 vs Kling 3.0Length, dialogue, references and cost, compared model to model.Read guide
Model comparisonAI video models comparedFrames, references, length and resolution for 14 models.Read guide
