Status: Soon for generation. You can prepare briefs now. Midways' initial V1 paths cover text-to-image and one short text-to-video clip; public access is pending.
Give the model a shot list in one sentence
For an image, use subject + setting + composition + light + essential constraint. Example: A red hiking boot on wet stone, low three-quarter view, overcast light, sharp product texture, no lettering. This says what must be visible without overloading the prompt with conflicting styles.
For video, use subject + single action + setting + camera behavior + mood. Example: A paper boat floats down a narrow stream at sunset; camera follows smoothly at water level; warm reflections; one continuous shot. The initial V1 clip lasts five seconds, so favor one action. Automatic multi-scene films are Soon.
Revise by changing one variable
If a product fills too little of the frame, change only framing: close-up, product fills two thirds of the frame. If motion feels chaotic, change only the camera direction: locked camera, subject moves left to right. Keep the working parts of the prompt. The next generation is a new request with a new quote.
Use planning as a draft
Help me plan can suggest an image composition or a video storyboard when planning access is configured. Edit Action and setting, Camera, Lighting and What must stay in the plan? before choosing Use this plan or Use this scene for one clip. Planning does not start image or video generation. For video, a 2–4 scene storyboard is a writing aid; generate and assemble each clip separately.
Avoid unsupported assumptions
Do not write instructions that depend on an uploaded visual reference, an automatically continuous multi-shot movie, or a guaranteed exact text rendering in the image. Visual references and combined multi-shot generation are Soon. After generation, use the manual editor for title text, timing and sound.
Should I name the model in my prompt?
No. Select a model in Model & output when offered; use the prompt for visual and motion instructions.
How long should my video prompt be?
The field permits up to 2,000 characters, but a short prompt focused on one five-second event is easier to evaluate and revise.