17–23 NOVThis Wan's on us.Free Wan 3.0 + $5 in bonus credits. Paid plans only.How it worksRun any Wan 3.0 or Wan 3.0 Prime workflow from 17 to 23 Nov. On 24 Nov we refund every Wan credit you spent, plus $5 in bonus credits. Flotime is billed as normal. No code needed.Full details →Run Wan 3.0 →
Floyo
Floyo
Workflows
API
Pricing
17–23 NOVThis Wan's on us.Free Wan 3.0 + $5 in bonus credits. Paid plans only.How it worksRun any Wan 3.0 or Wan 3.0 Prime workflow from 17 to 23 Nov. On 24 Nov we refund every Wan credit you spent, plus $5 in bonus credits. Flotime is billed as normal. No code needed.Full details →Run Wan 3.0 →
Floyo
Floyo
Workflows
API
Pricing
17–23 NOVThis Wan's on us.Free Wan 3.0 + $5 in bonus credits. Paid plans only.How it worksRun any Wan 3.0 or Wan 3.0 Prime workflow from 17 to 23 Nov. On 24 Nov we refund every Wan credit you spent, plus $5 in bonus credits. Flotime is billed as normal. No code needed.Full details →Run Wan 3.0 →

MiniMax H3 Camera Control · Image to Video

Turn a still image into a video that follows a camera path you choose, using the MiniMax H3 (Hailuo 3.0) model. Upload a reference image, supply a camera reference video, write a prompt, and hit run.

camera control
image to video
minimax h3

228

Gen time: -- secs

Nodes & Models

LoadImage
UNETLoader
KSamplerSelect
LoadVideo
CLIPLoader
RandomNoise
PrimitiveStringMultiline
VAELoader
VAEDecodeAudio
PrimitiveFloat
VAEDecode
ResolutionSelector
ModelPatchLoader
BasicScheduler
PrimitiveBoolean
LoraLoaderModelOnly
SamplerCustomAdvanced
PrimitiveInt
ComfySwitchNode
BasicGuider
CreateVideo
ComfyMathExpression
MiniMaxH3ReferenceToVideo
Video Slice
GetVideoComponents
ResizeImageMaskNode
DepthAnythingV2Preprocessor
MiniMaxH3FunControlNetApply
SaveVideo
PreviewImage

ABOUT THE WORKFLOW

Animate a photo with custom camera movement
Upload a reference image and a short camera reference video. The workflow extracts depth maps from your camera video and uses them to guide MiniMax H3’s video generation, so the output follows the exact camera path from your reference. You get back an MP4 with synced audio.

Model

  • MiniMax H3 (Hailuo 3.0) by MiniMax. A 33-billion parameter open-source video model strong at subject preservation, natural motion, and built-in audio generation. Outputs up to 1080p at 24fps.


HOW IT WORKS

Step 1. Upload a reference image
This is the visual starting point for your video. The model will preserve the subject and style of this image throughout the output.
Works great with: portraits · product shots · landscapes

Step 2. Upload a camera reference video
Record or find a short clip with the camera movement you want. The workflow extracts depth maps from this video and uses them to steer the generated output. The clip will be sliced to match your chosen output duration.
Works great with: drone footage · handheld tracking shots · slow pans

Step 3. Write a prompt
Describe the scene and the motion you want. Reference the subject in your image using <Picture 1>. Example: "<Picture 1> a woman walking through a neon-lit alley, cinematic camera movement, rain on the pavement".

Step 4. Set duration and resolution
Choose your output length in seconds (default: 5) and aspect ratio (default: 9:16 portrait at 0.4 MP). Longer durations and higher resolutions increase generation time. Maximum duration depends on resolution.

Step 5. Adjust camera control strength (optional)
ControlNet Strength (default: 0.4) sets how closely the output follows the camera reference. Higher values lock tighter to the camera path. Lower values give the model more creative freedom.

Step 6. Enable Lightning LoRA (optional)
Turn on Lightning LoRA to cut generation from 20 steps down to 4. Faster results with a slight trade-off in fine detail. Useful for drafts and quick iterations.

Step 7. Run and download
Hit run. The workflow generates your video with synced audio and saves it as MP4. You can preview extracted depth maps alongside the output.
Ready for: Premiere Pro · DaVinci Resolve · After Effects

First time? Leave every setting as-is. The defaults are the right starting point for almost everyone.


RECOMMENDED SETTINGS

Quick-start guide. Find the goal that matches yours and copy the settings.

  • Standard quality, portrait output — Duration: 5s, Resolution: 9:16 / 0.4 MP, Steps: 20, ControlNet Strength: 0.4. The right starting point for almost everyone.

  • Want faster drafts to test camera angles? — Enable Lightning LoRA. Steps drop to 4. Good enough to judge framing and movement before committing to a full render.

  • Need tighter camera tracking? — Raise ControlNet Strength to 0.6 or 0.7. The output will follow your reference video more precisely, but the model has less room to fill in natural motion.

  • Want more cinematic, organic movement? — Lower ControlNet Strength to 0.2 or 0.3. The camera path becomes a suggestion rather than a constraint.

  • Output looks jittery or inconsistent? — Make sure your camera reference video has smooth, continuous motion. Handheld shake and fast cuts confuse the depth extraction.

  • Subject doesn’t match the reference image? — Add <Picture 1> at the start of your prompt to anchor subject preservation. Keep the rest of the prompt focused on scene and motion, not on re-describing the subject.

Prompt: Describe the scene, mood, and type of camera motion. Be specific about lighting and environment. "<Picture 1> a man standing on a rooftop at golden hour, slow dolly forward, city skyline in background" works better than "person outside, moving camera".


LEARN

📹 Videos

✨ Quick links


USE CASES

🎬 Filmmakers & Directors
Previsualize camera moves before a shoot. Upload a still from a location scout and a reference clip with the dolly or crane move you want, and get a rough cut in seconds.

📱 Social Media Creators
Turn a single photo into a scroll-stopping video with cinematic camera movement. Choose 9:16, write a short prompt, and post the result straight to Reels or TikTok.

🏠 Real Estate & Architecture
Animate a rendered interior or exterior shot with a walkthrough camera path. Show clients what moving through a space feels like without booking a videographer.

🎮 Game & Concept Artists
Test camera angles on concept art or environment paintings. Upload the artwork, supply a camera path, and evaluate framing before building the scene in-engine.


WHAT WORKS BEST / WHAT TO AVOID

✅ Works great

  • Clean, well-lit reference images with a clear subject

  • Smooth camera reference videos with steady, continuous movement (dolly, pan, orbit)

  • Prompts that describe scene and motion rather than re-describing the image

  • 5-second clips at default resolution for fast iteration

⚠️ May produce softer results

  • Shaky handheld footage as camera reference (depth extraction struggles)

  • Camera reference videos with fast cuts or scene changes

  • Very long durations at high resolution (heavier on resources, more room for drift)

  • Prompts that contradict the reference image (asking for snow when the image shows a beach)


FAQ

What is MiniMax H3 and how does it compare to other video models?
MiniMax H3, also called Hailuo 3.0, is a 33-billion parameter open-source video generation model released by MiniMax in July 2026. It handles subject preservation and natural motion well, and generates synced audio alongside the video. It outputs up to 1080p at 24fps. It is less suited for long clips and text rendering compared to some closed-source alternatives.

How does camera control work in this workflow?
The workflow extracts depth maps from the camera reference video you upload. These depth maps feed into a ControlNet that guides the generation process, so the output video follows the same camera path as your reference. You control how tightly it follows with the ControlNet Strength slider.

What kind of camera reference video should I use?
Any short video clip with the camera motion you want to replicate. Drone footage, slow pans, dolly shots, and orbit moves all work well. The clip gets sliced to match your chosen output duration, so it does not need to be an exact length. Avoid clips with fast cuts or heavy shake.

Can I speed up generation without losing too much quality?
Yes. Enable the Lightning LoRA toggle. It reduces steps from 20 to 4, cutting generation time significantly. Fine detail may soften slightly, but it is fast enough for drafting and testing camera angles before committing to a full-quality render.

Does MiniMax H3 generate audio with the video?
Yes. The model produces synced audio alongside the video output. The workflow loads both a video VAE and an audio VAE, and the final MP4 includes the generated soundtrack.

What resolution and duration combinations are available?
The default is 9:16 portrait at 0.4 MP with a 5-second duration. Maximum duration depends on the resolution you choose. Lower resolutions allow longer clips. Higher resolutions cap out at shorter durations.

How to run MiniMax H3 camera control online?
You can run MiniMax H3 camera control online through Floyo. No installation, no setup, no API key to wire up. Open the workflow in your browser, upload your inputs, and hit run. Free to try.


WHY FLOYO?

Floyo is the only platform with team collaboration for ComfyUI in the browser. You run workflows with no install. You share run history, assets, and models across your team. You pay only when you generate. Floyo supports open-source and closed-source models.

A designer runs an edit and likes the result. A teammate opens that exact run from shared history and keeps going. No file handoffs. No version confusion.

For studios and enterprise teams, Floyo adds private workspaces, pooled resources, and a team usage dashboard. Other ComfyUI cloud tools run for one person at a time. Floyo runs for the whole team, with transparent per-generation costs.


Ready to try it?
Upload a reference image, grab a camera clip, and generate your first camera-controlled video.

→ Launch Workflow, Free

Questions? Watch the free course or check the FAQ above.

Read more

N