Frame-to-frame transition video
You set the start and end frames; the model walks the road between them: bud to bloom, empty room to furnished, morning to night.
Open Video StudioFrame-to-frame generation, also called first and last frame or keyframe video, gives the model two fixed points instead of one. In ordinary image-to-video the model starts from your picture and decides where the scene goes; here you also choose where it ends. The model only invents the motion in between, which makes transformations, reveals and before/after clips far more predictable.
In Ryna AI Video Studio this is the Frame→Frame flow, and it works with the engines that take both a first and a last frame: Seedance 2.5 and the Seedance 2.0 family, Kling 3.0, Veo 3.1 and Hailuo 3. Wan 3.0, Runway Gen-4.5 and Grok Imagine take a first frame only; use them for ordinary image-to-video.
Steps
- 1
Pick the Frame→Frame flow in the studio.
- 2
Upload first and last frames (make them in the Images tool).
- 3
Describe the transition in one sentence, and say whether the camera moves.
- 4
Generate 5–10 seconds; if the change looks rushed, lengthen the clip first.
What works well as a frame-to-frame video
Before and after: a room before and after renovation, a face with and without makeup, a plain product and the same product in its packaging. Transformation and morph: a bud into a flower, a pencil sketch into a finished painting, one object becoming another. Time-lapse: the same street from morning to night, a skyline as the lights come on. Product reveal: a closed box in the first frame, the product standing in the last.
Logo intros and invitations fit too: put a plain background in the first frame and the finished design in the last, so the clip ends on a clean frame.
How to prepare the two images
The model tries to explain the difference between your frames with motion; the smaller and more logical that difference, the cleaner the result. Keep the same framing and camera angle in both images, the same aspect ratio as the video you will generate, and the subject in roughly the same place. If the subject jumps from the left edge to the right, the model has to carry it across the frame and often blurs it on the way.
Keep the lighting continuous unless the change of light is the point: a morning-to-night clip should change light, a before/after room should not. The easiest way to get matching frames is to make the first one, then derive the second from it in the Images tool by editing only what should change.
Prompting the transition
The images already define the start and the end, so the prompt should describe the journey, not the scenes. Say what changes, in what order and how fast: "the petals unfold one by one, slowly, static camera". Always mention the camera; if it should not move, write "static camera" or "locked-off camera", otherwise many models add a slight drift.
Match the duration to the size of the change. A simple morph reads well in 5 seconds; a time-lapse with several stages needs 8–10.
Common artifacts and fixes
A cross-fade instead of motion, where the first frame simply dissolves into the last, usually means the frames are too different: bring them closer, or describe a physical process (growing, filling, folding) instead of "turns into". Melting or warping mid-clip: shorten the clip, simplify the change, or split one big transformation into two clips.
A pop at the end, where the last second jumps to the final frame, often means the last frame's light or framing contradicts the prompt. If faces change, frame the same person the same way in both images and avoid big pose changes.
Example prompts
“A bud opening into a flower, macro, soft light”
“An empty room filling with furniture, time skip”
“A city from morning to night, time-lapse, lights coming on”
“A closed cardboard box opens and the product rises out of it into its position in the last frame, static camera”
“A pencil sketch fills with color and becomes the finished painting, slow, top-down”
“The same living room before and after renovation: walls repainted, furniture arriving piece by piece, static camera”
Tips
- Same camera angle in both frames makes it believable.
- Derive the second frame from the first in the Images tool.
- Use the same aspect ratio for both images and the video.
- Write "static camera" unless you want camera movement.
- Rehearse at 480p on an economy engine, then re-run the keeper at a higher resolution.
- Seedance is the smoothest on transitions.
Recommended engines
Related pages
Frequently asked questions
- What is first and last frame video generation?
- You give the model a start image and an end image and it generates the motion between them. It is also called keyframe or frame-to-frame video.
- Which engines support a first and last frame?
- In Ryna AI: Seedance 2.5, Seedance 2.0 including Mini and Fast, Kling 3.0 Pro and Standard, Veo 3.1 including Fast and Lite, and Hailuo 3 including Max. Wan 3.0, Runway Gen-4.5 and Grok Imagine take a first frame only.
- Can I try it on the welcome credits?
- Yes, with the welcome-credit engines that take a last frame: Seedance 2.0 Mini and Fast, Veo 3.1 Lite and Hailuo 3 Max.
- Why does my video just fade from one image to the other?
- The frames are too different for the model to connect with motion. Match the angle and subject position, and describe a physical change in the prompt.
- Do the two images need the same size?
- Prepare both in the aspect ratio of the video you will generate, with the same framing. Differently cropped frames are the most common cause of a jumpy transition.
- How many credits does it cost?
- It depends on engine, duration and resolution; the studio shows the exact amount before you generate. A 5-second clip is ≈10–15 credits on economy engines and ≈30–70 on cinema-class ones. New accounts get 20 welcome credits.
Open Video Studio
20 welcome credits for new accounts. No subscription.
Open Video Studio