Skip to content

Free, open-source AI video generator for Mac. Run LTX-2.5 and MiniMax H3 (Hailuo) locally on Apple Silicon: text-to-video and image-to-video with sound, lip-sync, character LoRA training, songs from lyrics (YuE2) and image editing (Qwen-Image-Edit). One-click install via Pinokio. Offline: no cloud, no API key, no NVIDIA GPU.

2.7 KB
Raw Blame

Settings

The gear in the header opens Settings. Changes are saved to panel_settings.json when you press Apply and affect every new render; files already in the gallery are not re-encoded.

Output format {#output-format}

How finished clips are encoded.

Choice One line
Standard visually lossless to almost everyone, about 7 MB per 5 s clip
Video production (lossless) for grading and editing elsewhere, about 50 MB per 5 s clip
Web / social the smallest files, about 3 MB per 5 s clip
Custom set the pixel format and CRF yourself: 0 is lossless, 18 visually lossless, 23 the web default, 28 and up visibly lossy

Memory / speed {#memory-speed}

  • Live preview — On shows the shot as it forms in the Now card. It never changes the result (the file is byte-for-byte the same either way); turning it off also removes the Stop early button.
  • Hailuo H3 model (when H3 is installed) — Automatic picks the best this Mac can hold; Full (bf16) peaks around 42 GiB; Compact (Q8) peaks around 21 GiB, about half the memory for a little more time. On a Mac too small for Full, that option says so and is disabled.

Model files {#model-files}

Verify model files (checksum) checks every installed weight file against its published checksum and offers a one-click re-download of any that are damaged. It takes 1–2 minutes.

Storage {#storage}

Shown when this disk holds weights this version does not render with by default. Each pack lists its size and a Remove button, which asks first. Removing a pack never touches your renders in mlx_outputs.

API tokens {#tokens}

Saved on this Mac only, and sent nowhere except as sign-in to the site they belong to.

  • CivitAI API key — needed to install LoRAs from CivitAI. save & test checks it.
  • Hugging Face token — needed for gated models; read access is enough.

[[pinokio]]

Spicy mode (adult content) {#spicy}

Off by default: the CivitAI browser hides NSFW results and the Show NSFW switch. Enable Spicy mode asks for a second click, then lets you choose per search. [[/pinokio]]

Appearance {#appearance}

Dark, Light or System (follows the Mac, live). Saved in this browser only.

Completion alerts {#alerts}

A chime when a render finishes, and a browser notification when the tab is in the background once you allow it. Send a test tries it.

Anonymous usage analytics {#analytics}

Sends anonymous counts — version, hardware class, engine, tier, resolution, the first line of an error. Never your prompts, file names, paths, images or video; a copy of what is sent is kept in state/usage-log.jsonl. Turn off stops it.