Generate motion between two images — locally, with controlled prompting

Turn images into video. On your machine.

First + last frame • Positive / negative prompts • Local-only AI

PictaMotus is an AI image-to-video generator that creates intermediate frames between a start image and an end image. You can steer motion and style with a positive prompt and block unwanted artefacts with a negative prompt. Output is saved as MP4 locally, without sending frames or prompts to external servers.

COMMERCIAL USE: FREE

DATA FLOW: OFFLINE

CONTROL: POS/NEG PROMPTS

PLAYBACK: UP TO 30 FPS

MAX AREA: UP TO 1609

Access requirement: PictaMotus is free for commercial use, but it requires a valid license and a password. Contact Virtuo to confirm you are a company or a professional (B2B eligibility) and to receive access credentials.

Download → How it works

How it works

Choose a first frame (start image) and a last frame (end image).
Write prompts: a positive prompt for motion/style, and a negative prompt for what to avoid.
Set timing: pick Frames and FPS (up to 30 FPS) for the playback speed.
Generate: the model synthesises intermediate frames and writes an MP4 to your output folder.

Core idea: you define the endpoints; the AI invents the in-between — locally.

Controls & prompts

Frames: more frames = longer video (constraint: (Frames − 1) ÷ 4 must be an integer).
Steps: quality/time trade-off (more steps = slower, usually better).
Guidance: how strongly the model follows prompts (typical range: 4.5–6.0).
Max Area: internal resolution cap (up to 1609) — affects detail and VRAM usage.
Seed: reproducibility when you want the same result again.

Prompts: positive = what you want; negative = what you refuse to see.

Performance modes

CUDA (recommended): runs on NVIDIA GPU. Practical baseline is > 16 GB VRAM.
CPU fallback: supported, but slow — a few seconds of video can take hours.
VRAM optimisations: slicing / tiling and CPU offload can reduce peak memory.
Offline by default: once installed and activated, generation does not require internet.

Reality check: if you want predictable throughput, you need CUDA and a serious GPU.

Hardware recommendations

Minimum practical: NVIDIA GPU with 64 GB VRAM (CUDA required).
Ideal (workstation-grade):
NVIDIA RTX PRO 6000 Blackwell Workstation Edition, 96 GB GDDR7 with ECC, Blackwell, Dual-slot, PCIe Gen 5 x16, Full Height, Full Length — HIGHLY RECOMMENDED
NVIDIA H100 PCIe, 80 GB HBM3, Hopper, Dual-slot, PCIe Gen 5 x16
NVIDIA A100 PCIe (80 GB version), 80 GB HBM2e, Ampere
NVIDIA A16 PCIe, 64 GB GDDR6 (4x 16 GB), Ampere, Dual-slot, PCIe Gen 4 x16

CUDA is non-negotiable if you expect this to behave like a tool, not a weekend hobby project.

First launch & access flow

First launch: Terms & Conditions must be read to the end before acceptance is enabled.
Password required: access is provided after B2B verification via Virtuo.
Licenses exist: Trial, Monthly, Annual, and Lifetime.
All variants are $0: difference is access window and operational controls, not price.

Intent: controlled distribution and predictable support — not payment.

Responsible use

Use frames you own or have explicit permission to use.
Don’t impersonate or produce misleading content that could reasonably deceive.
Keep work local when confidentiality matters.
Validate outputs before publishing: artefacts and unexpected motion can happen.

GUI Preview

Click image to enlarge