A story video lives or dies on whether the viewer believes the world for a few seconds. That means the shot needs a clear subject, a deliberate camera move, and lighting that matches the mood of the scene, not just a beautiful image that happens to move.
HappyHorse 1.0 gives you three ways into that shot. Start with text when the scene exists only in your script. Start with an image when you already have a character or location reference. Use reference-to-video when the same face, costume, or set has to survive across several clips.
Last updated: 2026-10-20
The workspace below is preloaded with the mode, aspect ratio, and starter prompt for this page. You can change every setting before generating.
0 image references ready
0 video references ready
Generation Controls
These are the shared controls across all three workflow lanes.
Uses the faster route when available.
Include synchronized audio when the provider supports it.
Task Notifications
Useful for longer queue times. We keep the current task switch separate from your future default.
We send the prompt, task page link, and a direct download shortcut when the render finishes.
Save this as your default so new tasks start with notifications already enabled.
Nothing generated yet
Write the prompt like a shot list. Name the subject, what they are doing, where they are, what the camera is doing, and how the scene is lit. A line such as a lighthouse keeper climbing a spiral staircase at dusk gives the model a subject and a direction. Adding slow tracking shot, cold blue shadows, and a warm lantern gives it motion and mood. The result is easier to judge than a prompt that only lists adjectives.
Use the reference workflow when the same character appears in more than one shot. Supply a clear image of the face and costume, describe only the new action and camera move, and avoid re-describing features that are already visible in the reference. If a shot drifts, change one variable at a time: the action first, then the camera, then the lighting. That makes it obvious which instruction caused the change.
A lone lighthouse keeper climbing a spiral staircase at dusk, storm clouds gathering outside the tower windows, cinematic fantasy drama, slow tracking shot from below, realistic ocean spray, warm lantern light against cold blue shadows, 35mm film texture, no text, no watermark.
Swap the keeper and the lighthouse for your own subject and location. Keep the camera move and the warm-versus-cold lighting contrast; those two choices do most of the storytelling.
The page opens the workspace in text-to-video, image-to-video, or reference-to-video mode. That choice decides whether the model starts from words, a still image, or a set of references.
The loaded prompt describes a subject, a camera move, lighting, and a visual style. Replace the subject and setting with your own, but keep the structure so the model still has clear direction.
Generate one short clip, check the motion and subject consistency, then adjust one variable at a time. Once the look is stable, reuse the prompt with new subjects instead of starting from zero.
You do not have to read a guide, open another tab, and guess which controls matter. The workflow is already configured and the generator is directly below the explanation.
A 4 to 15 second shot is enough to judge motion, framing, and subject consistency. Iterate on a short clip first, then generate the final shot with the settings you trust.
Text, image, and reference workflows share the same task history and controls, so you can move from a still-led shot to a reference-led shot without learning a second product.
Short answers about the workflow, the inputs, and what to expect from a first generation.
Paste the scene description, not the entire screenplay. Pick one shot, describe the subject, action, location, camera move, and lighting, and keep it under the prompt limit. Repeat for the next shot.
Use reference-to-video with a clear image of the character. Keep the reference consistent, describe only the new action, and review each clip before generating the next one.
Start with a short clip that is long enough to read the action, usually 4 to 8 seconds. Once the motion and framing work, generate the final version at the length your edit needs.
Review the terms of service and the rights you have in your source material. You are responsible for the script, references, and how the output is used.
Generate anime-style AI video scenes with cel-shaded characters, painted backgrounds, and cinematic camera moves. Text and reference workflows.
Start creating
Generate cinematic B-roll for videos, ads, and documentaries. Coffee pours, city details, hands, textures, and slow camera moves in 16:9.
Start creating
Generate vertical YouTube Shorts clips with a strong opening second, centered subjects, and clean space for captions. Text-to-video in 9:16.
Start creatingLoad the story workflow, replace the lighthouse with your own scene, and generate a short clip you can judge before you build the next shot.
