Every video model on Pictx offers both paths: write a prompt and generate from scratch, or upload a still and animate it. They are not two ways to get the same result — they are different tools for different jobs.
Text-to-video: maximum freedom
Start from nothing but words. The model invents the scene, the subject, the camera — everything.
Use it when:
- You have no source imagery — the idea only exists in your head.
- You want surprising, exploratory results and are iterating on concepts.
- The shot needs elements that are hard to source: fantasy scenes, impossible camera moves, generated environments.
The tradeoff: less control. Small prompt changes can produce very different frames, and keeping a character consistent across clips takes prompt discipline.
Image-to-video: maximum control
Start from a real frame — a photo, a product shot, a generated still you already approved. The model adds motion while preserving composition.
Use it when:
- The visual is already decided: a product photo, a client-approved keyframe, a character sheet.
- Consistency matters more than novelty — the subject must look exactly like the source.
- You are building a sequence: generate stills in one style, then animate each.
The tradeoff: the model can only work with what is in the frame. Camera moves are more limited, and motion beyond the original composition is constrained.
The pro workflow: combine both
The strongest pipeline on Pictx is a hybrid:
- Text-to-image first. Generate the keyframe with an image model — you get full control over composition, subject and style, and can iterate cheaply on stills.
- Image-to-video second. Send the approved frame to Seedance 2.5 or Kling i2v to add motion.
This costs less and looks better than repeatedly re-rolling text-to-video until a frame you like happens to animate well.
Credit reality
Drafting with stills is dramatically cheaper than drafting with video. Settle the look in images (a few credits per try), then spend video credits once on the approved frame.
Start with a still in the image studio, then animate it in image-to-video.