This is a scannable reference for prompting Higgsfield AI, the browser-based video and image studio built around cinematic Camera Controls. Keep it open in a tab: grab the formula, pick a preset, copy the vocab, and paste. For the deep-dive picks, see the best Higgsfield prompts roundup and the camera controls guide.
The prompt formula
Higgsfield's signature workflow is simple: pick a camera-move preset in the app, then write a prompt describing the subject and scene. Naming the move in the prompt and selecting the matching preset gives the most repeatable results. Follow the order below and end every prompt with a settings reminder.
Camera move (preset) + Subject + Action + Setting + Lighting + Lens/style + MoodDescribe the subject concretely (who or what, wardrobe, expression, action) and keep to one main camera move per clip — stacking moves gets messy and simpler motion renders cleaner. Ready-made structures live in the prompt templates.
Camera-move presets
Higgsfield ships 70+ camera-move presets. Select one in the app and name it in your prompt. Use bold, distinctive moves (Crash Zoom, Bullet Time) for a single big beat — overusing them kills the punch.
| Group | Presets | What it does / Best for |
|---|---|---|
| Zoom | Crash Zoom In/Out, Rapid Zoom In/Out, YoYo Zoom | Fast punch toward or away from the subject for impact; save Crash Zoom for one reveal or reaction beat. |
| Dolly | Dolly In/Out/Left/Right, Super Dolly In/Out, Double Dolly | Smooth camera travel on an imaginary track; steady approach, reveal, or lateral pass. Super Dolly is a longer, more dramatic push. |
| Vertigo | Dolly Zoom In/Out | Zoom counter to a dolly move so the background warps while the subject stays fixed — unease, dread, realization. |
| Orbit / Arc | 360 Orbit, Arc Left/Right, Lazy Susan | Circle or sweep around the subject to show it in the round; product spins, hero shots, dance. |
| Aerial | FPV Drone, Crane Up/Down, Jib Up/Down, Overhead | Fly, lift, or look straight down for scale and geography; establishing shots and reveals. |
| Signature FX | Bullet Time, Snorricam, Robo Arm, Whip Pan, Through Object In/Out, Object POV | Bold stylized moves — frozen-time orbits, body-locked cameras, seamless transitions. High impact; use sparingly. |
| Time | Hyperlapse, Timelapse Human/Landscape/Glam | Compress time as the camera moves or holds; crowds, skies, transformations, glam reveals. |
Models & tools
Higgsfield bundles its own models and external engines in one workspace, so you pick the tool that fits the shot without leaving the app.
| Model / tool | What it's for |
|---|---|
| Soul (Soul 2.0) | Flagship photo model tuned for fashion, aesthetics, and realism. |
| Soul ID | Train a personal identity model from 20+ reference photos so the same face carries across every generation without re-uploading. |
| DoP (Director of Photography) | Image-to-video engine that applies real camera moves to a still photo. |
| Speak | Lip-synced talking avatars generated from a photo or character plus your audio. |
| Image-to-Video | Animate an uploaded still with a chosen camera move in the Create workspace. |
| Draw-to-Video | Sketch motion or guides on a frame and turn them into video in the Create workspace. |
| Higgsfield Assist | AI prompt copilot that refines prompts and recommends camera settings before you spend credits. |
| Cinema Studio 3.5 | Narrative workspace with an AI Director that breaks one concept into individual shots and assigns per-shot camera moves — build a multi-shot sequence instead of a single 5-second clip. |
| Marketing Studio | Ad workspace: paste a product URL and it pulls the images, name, and features to generate publish-ready UGC ads (talking heads, unboxings, virtual try-ons) and CGI product spots — with 40+ AI avatars or a custom one you describe. |
| Seedance 2.5 | Newest bundled engine for long-form, audio-first clips: a continuous take up to 30 seconds, native synced audio (ambience, foley, score), 4K upscaling, and up to 50 reference images/clips for character and style consistency. Best for multi-shot narrative and transformations. Prompt it with its own formula (below). |
| Wan 2.6 | Restyle existing footage — feed it a reference clip and it re-skins the look while keeping the original motion. |
| Sora 2 / Kling 3.0 / Veo 3.1 / Hailuo 02 | The rest of the 15+ bundled external engines you can switch to for different looks, motion styles, and clip lengths — all in one subscription. |
Seedance 2.5 reads a different prompt formula than Higgsfield's own presets. State the shot count, total duration, and aspect ratio at the top, then number each shot and give it one camera move. It reads camera language directly — dolly in, truck left, push in, pull back wide, arc shot, crane up, orbital move, handheld follow — and generates sound in sync, so describe the audio too. Add no 3D, no cartoon, no VFX to force realism. See example 8 below.
Lighting & lens modifiers
Drop these terms into the Lighting and Lens/style slots of the formula to steer mood and framing. One or two per prompt is plenty.
| Modifier | Effect |
|---|---|
| Golden hour | Warm, low, directional sunlight; long shadows and a soft, flattering glow. |
| Blue hour | Cool twilight after sunset; moody, cinematic, city-lights look. |
| Low-key | Mostly dark frame with selective highlights; dramatic, high-contrast. |
| High-key | Bright, even, shadow-light exposure; clean, airy, commercial. |
| Rim light | Light from behind the subject that outlines its edges and separates it from the background. |
| Practical neon | Visible in-scene colored lights (signs, tubes) casting saturated color on the subject. |
| Softbox | Diffused, wraparound studio light; smooth skin and gentle falloff. |
| Volumetric | Visible light beams and atmosphere (haze, dust, god rays) for depth and mood. |
| 24mm | Wide angle; expansive framing, exaggerated depth, good for interiors and drone shots. |
| 35mm | Natural, near-eye field of view; documentary and street feel. |
| 50mm | Standard, honest perspective; portraits and product without distortion. |
| 85mm | Short telephoto; flattering portrait compression and background separation. |
| 100mm macro | Extreme close-up detail; textures, product surfaces, tiny subjects. |
| Anamorphic | Wide cinematic frame, oval bokeh, and horizontal lens flares. |
| Shallow depth of field | Sharp subject over a soft, blurred background; isolates and premiumizes. |
Settings (app, not prompt)
Duration, aspect ratio, and model choice are app settings, not inline parameters you type in the prompt. Set them in the interface, then add a reminder at the end of the prompt so you don't forget.
| Setting | Options / note |
|---|---|
| Aspect ratio | 16:9 (landscape), 9:16 (vertical social), 1:1 (square). Chosen in the app, not typed inline. |
| Duration | Higgsfield's own camera moves run about 5 seconds; the bundled Seedance 2.5 engine generates up to a 30-second continuous take with native audio. Set duration in the app. |
| Model | Pick Soul, DoP, Speak, or a bundled engine (Sora 2, Kling 3.0, Veo 3.1, Seedance 2.5, Wan 2.6, Hailuo 02) in the app. |
| Preset | The camera-move preset is selected in Camera Controls, then repeated in the prompt text. |
| Reminder convention | End every prompt with a middot-separated note, e.g. (Preset: Crash Zoom In · 9:16 · 5s) — a reminder, not prompt text. |
Copy-paste example prompts
Copy-paste prompts that follow the formula. Each ends with a settings reminder — set those in the app, then paste the prompt and pick the matching preset. Example 8 uses the different Seedance 2.5 formula for a longer, audio-first sequence.
1. Bullet Time rooftop freeze
Bullet time orbit around a breakdancer frozen mid-freeze on a rooftop at night, time nearly stopped as the camera arcs 180 degrees, dust and water droplets hanging motionless in the air, dramatic blue backlight and city glow, anamorphic lens, slow-motion realism, ultra-detailed, cinematic. (Preset: Bullet Time · 16:9 · 5s)Why it works: one bold signature move, a concrete frozen subject, and volumetric detail (dust, droplets) give the frozen-time effect something to hang on.
2. Crash Zoom reaction
Crash zoom in on a chef's face as she tastes the sauce and her eyes widen with delight, steam rising in a warm restaurant kitchen, practical neon and soft tungsten light, 85mm shallow depth of field, energetic and joyful mood. (Preset: Crash Zoom In · 9:16 · 5s)Best for: punchy vertical social beats — save the crash zoom for this single reaction so it lands.
3. FPV Drone reveal
FPV drone flight racing low over crashing coastal waves at golden hour, then climbing to reveal a lone lighthouse on the cliff, warm directional sunlight and volumetric sea spray, 24mm wide angle, epic and cinematic mood. (Preset: FPV Drone · 16:9 · 5s)Why it works: the drone move plus a wide 24mm lens sells scale, and the climb-to-reveal gives the 5-second clip a clear arc.
4. Image-to-Video with DoP
Slow dolly in on the uploaded portrait of a woman in a red coat on a rainy street, subtle rain and breath fog, preserve the original face, lighting, and geometry, moody blue hour with practical neon reflections, shallow depth of field, quiet cinematic mood. (Preset: Dolly In · DoP · 9:16 · 5s)Best for: animating a still without redrawing the subject — the "preserve the original face, lighting, and geometry" line keeps identity stable.
5. 360 Orbit product hero
360 orbit around a matte-black wireless headphone floating above a reflective pedestal, slow full rotation, high-key studio lighting with a soft rim light on the edges, 100mm macro detail on the mesh and stitching, clean premium mood. (Preset: 360 Orbit · 1:1 · 5s)Why it works: a single orbit shows the product in the round, and macro plus rim light premiumizes the surface for ads and marketplaces.
6. Cinema Studio 3.5 multi-shot sequence
A barista opening a small café at dawn. Shot 1: crane down from the shopfront sign to the front door (Crane Down, wide 24mm, cool blue-hour light). Shot 2: dolly in on hands tamping espresso, steam rising (Dolly In, 85mm shallow depth of field, warm practical light). Shot 3: crash zoom to her satisfied smile as the first cup pours (Crash Zoom In, golden window light). Consistent character and warm cinematic mood throughout. (Cinema Studio 3.5 · AI Director · 16:9)Best for: multi-shot stories. Let the AI Director split the concept into shots, then keep one main camera move per shot — the same one-move-per-clip rule, applied across a sequence.
7. Marketing Studio UGC product ad
UGC-style talking-head ad for the linked product. A friendly female creator in her late twenties films herself selfie-style in a bright bedroom, holding the product to camera and reacting with genuine excitement. Natural handheld phone look, soft window light, casual and authentic tone. She unboxes it, shows one clear benefit, then gives a quick thumbs-up call to action. (Marketing Studio · paste product URL · Avatar: library or custom · 9:16)Best for: fast social ads. In Marketing Studio you paste the product URL and it pulls the images and features — this prompt steers the creative: the avatar, setting, tone, and the beat structure (react → show benefit → CTA). Swap in CGI product spot or virtual try-on for a different format.
8. Seedance 2.5 multi-shot with native audio
8 shots, 20 seconds, 16:9, native audio. Montage, multi-shot cinematic. A lone storm-chaser steps out of a pickup onto a cracked desert road as a supercell builds. Shot 1: pull back wide to reveal the storm wall (crane up, 24mm). Shot 2: handheld follow as she raises a camera. Shot 3: push in on her focused eyes (85mm). Shot 4: orbital move around the truck as the wind rips. Shot 5: low arc shot of hail starting to fall. Shot 6: dolly in on the first lightning strike hitting the plain, volumetric. Shot 7: truck left across debris skittering past. Shot 8: pull back wide as she drives off into the wall of rain. Audio: rising wind and thunder ambience, foley of gravel and hail, tense orchestral score. No 3D, no cartoon, no VFX — ultra-real. (Engine: Seedance 2.5 · 16:9 · 20s)Why it works: Seedance 2.5 uses its own formula — shot count, total duration, and aspect ratio up front, then one camera move per numbered shot. It reads those camera terms directly and generates the ambience, foley, and score in sync, so the audio line pays off. Because it holds a single take up to 30 seconds, the whole sequence renders in one generation instead of stitching 5-second clips, and no 3D, no cartoon, no VFX pushes it toward realism.
Frequently Asked Questions
Do I have to name the camera move in the prompt if I already picked the preset?
Name it in both places. Selecting the preset in the app sets the motion engine, and repeating the move in the prompt text (for example, "crash zoom in") reinforces it. Doing both gives the most repeatable, on-target results.
Can I stack two camera moves in one clip?
Avoid it. Higgsfield renders cleanest with one main camera move per clip. Stacking moves gets messy and often produces warping or drift. If you need two moves, split them into two clips and cut between them.
Are aspect ratio, duration and model typed into the prompt?
No. Aspect ratio (16:9, 9:16, 1:1), duration (about 5 seconds) and the model choice are app settings you select in the interface, not inline parameters. We add a reminder like (Preset: Crash Zoom In · 9:16 · 5s) at the end of each prompt so you don't forget to set them.
What is the difference between Soul and Soul ID?
Soul (Soul 2.0) is the flagship photo model tuned for fashion, aesthetics and realism. Soul ID lets you train a personal identity model from 20 or more reference photos so the same face carries across every generation without re-uploading a reference each time.
How do I keep a face and lighting consistent in image-to-video?
Use DoP or Image-to-Video and add the line "preserve the original face, lighting, and geometry" to your prompt. That tells the engine to apply the camera move without redrawing the subject, which keeps identity and scene stable.
Which external models can I use inside Higgsfield?
Higgsfield bundles 15+ engines in one subscription — Sora 2, Kling 3.0, Veo 3.1, Seedance, Wan and Hailuo — in the same workspace, alongside its own Soul, DoP and Speak tools, so you can pick the engine that fits the shot without leaving the app.
What is Higgsfield Marketing Studio?
Marketing Studio is Higgsfield's ad workspace. Paste a product URL and it pulls the product images, name and features, then generates publish-ready ads — UGC formats like talking heads, unboxings and virtual try-ons, plus CGI product spots — using 40+ ready AI avatars or a custom presenter you describe. It's built for ads rather than the cinematic camera-move clips the rest of the app makes.
What can Seedance 2.5 do that Higgsfield's own models can't?
Seedance 2.5 is the bundled engine for long-form, audio-first clips. It generates a continuous take up to 30 seconds (versus the ~5-second Higgsfield camera moves), adds native synchronized audio — ambience, foley and score — upscales to 4K, and accepts up to 50 reference images or clips for character and style consistency. Prompt it differently too: state the shot count, total duration and aspect ratio at the top, then number each shot with one camera move.