NemoClaw Knowledge Wiki

Making AI videos locally with Pinokio - Kevin Stratvert channel

3 min read

Making AI videos locally with Pinokio - Kevin Stratvert channel

https://www.youtube.com/watch?v=G2Ec3h5CfA8 Here is a summary of the guide on generating AI videos locally on your PC, based on the video transcript.

How to Generate Free AI Videos Locally on PC

This guide covers how to run open-source AI video models (like LTX-2 and Wan) directly on your computer using a tool called Pinokio. This method ensures privacy, requires no subscription, and has no usage limits.

1. System Requirements

Running AI video generation locally is resource-intensive.

  • GPU: A dedicated NVIDIA GPU is highly recommended.
  • VRAM: You need at least 6-8 GB of VRAM. More VRAM allows for faster generation and longer clips.
  • How to check VRAM (Windows):
    1. Press Ctrl + Shift + Esc to open Task Manager.
    2. Click the Performance tab.
    3. Click GPU.
    4. Look for “Dedicated GPU memory.”

2. Install Pinokio

Pinokio is a “one-click installer” for AI tools that manages complex dependencies (Python, CUDA, etc.) automatically.

  1. Go to the Pinokio website.
  2. Download the installer for your OS (Windows, Mac, or Linux).
  3. Run the installer. Note: The first install may take a few minutes as it downloads necessary base files.

3. Install Wan2GP

Wan2GP is the specific script used to run the video models inside Pinokio.

  1. Open Pinokio and click on Discover.
  2. Search for Wan2GP.
  3. Click Install.
  4. Pinokio will list required dependencies; click Install to confirm.
  5. Once installed, click Start to launch the Web UI.

4. Configuring the Model (LTX-2)

In the Wan2GP Web UI:

  • Select Model: Choose LTX-2 from the dropdown menu.
    • Note: LTX-2 is highlighted because it can generate audio and narration along with the video.
  • Model Type: Select Distilled. (Should be able to run the non-distilled on my Nvidia rtx8000
    • The distilled version is roughly half the size (~20GB) of the full model, making it more stable and practical for consumer GPUs with minimal quality loss.
  • Performance Settings:
    1. Go to the Configuration tab > Performance.
    2. Select a VRAM Profile that matches your hardware (e.g., Profile 2 for ~12GB VRAM).

5. Generating Video

Option A: Text-to-Video

  1. Select Text Prompt Only.
  2. Prompt: Enter a descriptive prompt.
    • Tip: You can generate dialogue by typing syntax like: She says, "We need the brand to feel more authentic."
  3. Settings:
    • Resolution: Start with 720p for faster results.
    • Aspect Ratio: Choose 16:9 (YouTube) or 9:16 (TikTok/Shorts).
    • Duration: Set the frame count (e.g., 240 frames @ 24fps = 10 seconds).
  4. Click Generate.
    • Note: The first generation takes longer as the model loads into memory. Subsequent runs will be faster (~30 seconds per clip on good hardware).

Option B: Image-to-Video

  1. Select Start Video with Image.
  2. Drag and drop your source image into the media box.
  3. Enter a prompt describing how the image should move or animate (e.g., “The person continues walking towards the castle”).
  4. Click Generate.

6. Advanced Features & Outputs

  • Advanced Mode: Toggle this on to generate multiple variations (batches) of a single prompt automatically.
  • Viewing Files:
    • Recent videos appear at the bottom of the UI.
    • To see all files: Click the storage size icon in the top bar > Open app folder > Open outputs folder.

Summary of Benefits

  • Cost: 100% Free.
  • Privacy: Everything stays on your local machine.
  • Control: No restrictions on content or usage frequency.
  • Capabilities: Generates video, sound, and speech.

Graph View