# Brandiseer Video Prompting Guide This guide helps you write effective motion prompts for `POST /api/v2.1/video/generate`. It is optional — the endpoint works without reading this first — but following it will noticeably improve your results. ## What this endpoint actually accepts Before writing a prompt, know the shape of the input you're working with: - **Exactly one source image** — either a public `imageUrl` (JSON body) or a single multipart image upload. Never both, never more than one, no arrays, no base64/data URLs. If your source image is a local file, upload it as multipart form data; do not send a filesystem path. - **One text `prompt`** describing the motion, camera, and (optionally) audio/dialogue you want in the resulting video. - **No other reference inputs.** This endpoint does not accept separate audio clips, video clips, multiple subject images, storyboards, or keyframe sequences — only the single starting image plus your prompt. - **Image-to-video only.** There is no text-to-video mode; a starting image is always required. - **No editing or extension of existing videos.** This endpoint animates one still image into a new video; it does not edit, extend, or splice existing video files. Recommended workflow: generate your starting image with Brandiseer's image API (`POST /api/v2/generate` or `/api/v2/generate/async`), then pass its returned `imageUrl` straight into `/api/v2.1/video/generate`. ## Treat it as a visual storyteller, not a text completion Write structured prompts the way you'd brief a cinematographer: describe what the camera sees and how it moves, not just a vibe. A one-line mood description produces a much less controlled result than a shot-by-shot description. ## Basic prompt structure **One-sentence summary** — Subject + location + event + genre/style + camera movement. This anchors the whole generation before you add detail. **Detailed motion description** — Break the clip into a short timeline (see Timestamps below) and describe, for each segment: what the subject does, how the camera moves, and any sound. Use positive descriptions wherever possible; negative phrasing works well specifically for suppressing subtitles or audio ("no subtitles", "no background music"), described further below. **Consistency notes** — Call out anything that should stay constant across the whole clip: camera angle, framing, environment, lighting direction, atmosphere. Since there's only one source image, the model will already anchor subject appearance to it — you mainly need to describe *motion*, not re-describe the subject's look. ### Example ``` Realistic nature documentary style, natural lighting and shadows. A chubby panda cub rolls down a grassy forest slope on a warm afternoon. 0s-3s: The panda lies on the grassy slope, then begins rolling sideways down the hill with clumsy, adorable movements, gently bending the grass beneath it. A light breeze passes through; sunlight filters through the trees from the upper left, creating dappled light and shadow. 3s-6s: The panda rolls to a stop at the bottom of the slope, shifts from lying on its side to lying on its belly, and turns its round face toward the camera. Low camera position, slight handheld feel, subtly following the panda as it rolls. Natural depth of field: foreground grass slightly blurred, panda in focus, background forest softly out of focus. Natural environmental audio only — wind, rustling grass, the soft sound of the panda rolling. Warm, realistic, natural mood throughout. ``` ## Camera language Basic shot and camera terms can be written directly into the prompt: - **Shot size**: extreme wide shot, wide shot, medium shot, medium close-up, close-up. - **Camera movement**: push in, pull out, pan, track, follow, orbit, dive, pull back, tilt up, handheld shake. - **Camera angle**: low angle, overhead shot, first-person perspective. - **Named techniques**: one-shot/long take, dolly zoom, aerial perspective, FPV, bullet time, handheld shot, speed ramp. For a more obscure technical term, spell it out as [term + plain description]: e.g. "rack focus: the focus shifts smoothly — what was clear in the foreground becomes blurred while the subject in the background comes into focus." ## Action and expression description - **Actions**: favor general descriptions ("does several sets of high-knee raises", "the two figures engage in close combat") over listing every individual motion. Call out only a few specific, memorable beats, and avoid repeating the same action twice. - **Expressions**: describe them in plain sentences rather than idioms — "her eyes widen and her mouth falls slightly open" reads more reliably than "she's shocked." ## Timestamps Use timestamps to control pacing over the clip's duration. One-second intervals are a reasonable base unit. - Too little described in a time range and the model will improvise more freely than you intended. - Too much packed into a short range can cause rushed cuts or dropped detail — match how much you ask for to how much time you give it. - Timestamps aren't reliable for controlling very high-frequency repeated actions (e.g. "shakes its head three times per second"). Supported ways to express timing: - **Clear intervals**, kept continuous (no gaps): `"0-3s ... 3-7s ... 7-12s"`. - **Time-point cues**: `"At the 5-second mark, the camera quickly pans left."` - **Relative timing**: `"She stands there blankly. After 3 seconds, she smiles."` ## Audio and dialogue prompting You can describe audio directly in the prompt, including: - **Dialogue** — write the actual line and who says it: `Dialogue (the woman): "Wait for me."` Keep lines short enough to plausibly fit the shot's duration. - **Ambient/environmental sound** — wind, traffic, footsteps, water, room tone. - **Music** — genre, mood, or instrumentation ("gentle piano", "upbeat synth pop"), rather than naming a specific existing track. - **Silence or suppression** — you can explicitly ask for quiet or for specific audio elements to be left out. ### Negative audio/subtitle control - To suppress subtitles: `"no subtitles"` or `"do not add subtitles."` - To control which audio layers are present: `"no background music — generate only environmental and action sounds"`, or simply `"no audio."` ## What this guide does not cover Multi-image or multi-subject reference mapping, motion/style reference from a second video, 3D-model-to-render workflows, video instruction editing, video extension, multi-panel storyboards, keyframe sequences, and seamless transitions between two videos are all real techniques for other kinds of video-generation systems, but none of them apply to this endpoint — it takes exactly one starting image and one motion prompt. Prompts written for those richer multi-asset systems will mostly still work here; just drop any references to "Image 2", "Video 1", or similar asset numbering, since there is only ever one image.