Free Addons & Tools

We need to prepare an Indian VPN, this is an essential prerequisite. Click here to install free Indian node VPN. Of course, to increase availability, it's recommended to use paid VPN Click here to get Surfshark.

They both have mobile apps, direct installation is more convenient!

After switching to Indian IP, then enter ChatGPT official website, you can register a new account, or use an old account (Foo IT Zone uses an old account).

After entering, you can see free trial ChatGPT for 12 months at the top, you can see "Free Gift" or "Activate Plus" prompt, click to see the following page:

Choose GO meal, original price is 399 rupees but now it's 0 rupees,说明 you can enjoy this 1-year discount package. If you don't see the above page, it means your IP address is not pure enough, or your ChatGPT is a new account,建议 changing account to try. It's recommended to use UnionPay credit card,建议 using your real card (doesn't need to be Indian), also no need to switch regions, activate directly!

To avoid being cancelled like before, using it for a few days and then GO membership gets cancelled, so a stable Indian node VPN is essential. Whether you're on PC or mobile, when using ChatGPT GO paid account, always keep the Indian node VPN connected. Free is certainly nice, but unstable, so you can choose paid VPN, need to have Indian nodes, such as Surfshark or ProtonVPN both are available! They both have mobile apps, direct installation is more convenient!

Or you can be like me, have a permanent free Oracle Cloud Server, then deploy OpenVPN or WireGuard inside, you can permanently get a free Indian proxy node.

wget https://git.io/vpn -O openvpn-install.sh && bash openvpn-install.sh

In OCI console, add to Network Security Group (NSG) or Security List:

• Protocol: UDP
• Port: 51820
• Source: 0.0.0.0/0

Or directly open all ports to save trouble each time!

Open all ports:

iptables -P INPUT ACCEPT
iptables -P FORWARD ACCEPT
iptables -P OUTPUT ACCEPT
iptables -F

Ubuntu mirror sets iptables rules by default, disable it:

apt-get purge netfilter-persistent
reboot

Or force delete:

rm -rf /etc/iptables && reboot

Or install WireGuard, security encryption is better, suitable for special users:

Ubuntu / Debian:

apt update
apt install -y wireguard qrencode

wg genkey | tee server_private.key | wg pubkey > server_public.key
wg genkey | tee client_private.key | wg pubkey > client_public.key

View public keys:

cat server_public.key
cat client_public.key

View private keys:

cat server_private.key
cat client_private.key

Create configuration file:

sudo nano /etc/wireguard/wg0.conf

Note: Modify the network card inside: eth0 (everyone's is different). eth0 might not be your network card name, you can use ip a to check (OCI is usually ens3 or enp0s3).

If it's ens3, please change all eth0 above to ens3.

sudo nano /etc/sysctl.conf

Uncomment or add:

net.ipv4.ip_forward=1

Then execute:

sudo sysctl -p

sudo systemctl enable wg-quick@wg0
sudo systemctl start wg-quick@wg0

Client configuration example:

Register for a free NVIDIA NIM account:

After logging in, generate your own API Keys in the settings center. Select "Never Expires" for expiration time. Currently can be called directly for free, no limits discovered yet.

Call API through Cherry Studio to easily achieve: Intelligent Dialogue · Autonomous Agent · Unlimited Creation, unified access to mainstream large models!

1. Install and start Cherry Studio, then click "Settings" (gear icon) at the top
2. Select "Model Services" on the left side
3. Find "NVIDIA" models in the dropdown on the right
4. Fill in the API Key you obtained from NVIDIA on the right

Click the "Manage" button at the bottom, manually add these two models:

Just copy the model names above and search in the management to add the models.

True "All-in-One" Model

LTX-2 is not just another AI model - it's the first truly "all-in-one" AI video generation model. Unlike others that can only generate video without sound, or have mismatched audio/video, or require ridiculous hardware specs, LTX-2 delivers synchronized audio + video + high quality + local deployment + low requirements.

Multiple Performance Modes

Extreme Mode: For maximum quality output
VRAM-Saving Mode: Optimized for 8GB graphics cards
High-Speed Mode: For quick drafts and prototyping

🚀 One-Click ComfyUI Deployment (Recommended for Beginners)

Super convenient one-click deployment with ComfyUI latest version!

Download ComfyUI

📦 LTX-2 Model Checkpoint (Download One)

ltx-2-19b-dev-fp8.safetensors - Download
ltx-2-19b-dev.safetensors - Download
ltx-2-19b-distilled.safetensors - Download
ltx-2-19b-distilled-fp8.safetensors - Download

🔍 Essential Upscalers

ltx-2-spatial-upscaler-x2-1.0.safetensors - Required for current two-stage pipeline
ltx-2-temporal-upscaler-x2-1.0.safetensors - Supported for future pipeline implementation
ltx-2-19b-distilled-lora-384.safetensors - Simplified LoRA (required for current pipeline except DistilledPipeline and ICLoraPipeline)

🎬 TI2VidTwoStagesPipeline

Production-grade text/image-to-video, supports 2x upscaling (recommended)

⚡ TI2VidOneStagePipeline

Single-stage generation for rapid prototyping

🔥 DistilledPipeline

Fastest inference with only 8 predefined sigma values (8 steps first stage, 4 steps second stage)

Performance Optimization

Use DistilledPipeline: Fastest inference with 8 predefined sigma values
Enable FP8: Reduce memory usage with --enable-fp8 (CLI) or fp8transformer=True (Python)
Attention Optimization: Use xFormers (uv sync --extra xformers) or Flash Attention 3 for Hopper GPUs
Gradient Checkpointing: Reduce inference steps from 40 to 20-30 while maintaining quality
Skip Memory Cleanup: Disable automatic memory cleanup between stages if you have sufficient VRAM

1️⃣ Chinese Couple Conversation (Lip Sync + Emotion Test)

A 20-year-old Asian couple sitting in a café, girl smiling and speaking Mandarin: "Do you still remember when we first met?" Boy nods gently, replying in Mandarin: "Of course I remember, you were wearing a white dress that day, I fell for you at first sight." Natural lighting, realistic photography style, slight camera movement, perfect lip sync with audio.

2️⃣ Comedy Couple Short Drama

Asian young couple arguing at home, girl speaking Mandarin angrily: "You forgot to do the dishes again!" Boy looks innocent, replying humorously: "I didn't forget, I was waiting for inspiration!" Light comedy style, exaggerated but natural expressions, lip sync, fast rhythm.

3️⃣ Gaming Live Stream Style

First-person shooter game footage, player fighting in city ruins while commenting in Mandarin: "This gun's recoil is too strong, but the damage is really high, I need to circle around from the right." Smooth gameplay, gun sounds sync with audio, slight game HUD.

Revolutionary 3D Visual Reconstruction System

SAM 3D is a 3D visual reconstruction system that Meta has upgraded based on its classic Segment Anything Model (SAM). It's not simply about "identifying objects from images," but rather directly reconstructing usable 3D models, poses, and spatial structures from a single image or video.

Two Main Components

SAM 3D Body: Focuses on 3D pose, motion, skeleton, and mesh reconstruction of the human body
SAM 3D Objects: Used to recreate various objects in the real world, such as furniture, tools, and electronic products

🛍️ AR Shopping: Bringing Products "Into Your Home"

Merchants upload product photos → SAM 3D generates 3D models → You open AR on your phone → Place it directly in your living room. This upgrades e-commerce from "ordering based on images" to "ordering after a real preview."

🏥 Medical and Rehabilitation: AI Understands Your Every Move

SAM 3D Body can reconstruct human skeleton from video, identify joint angles, and analyze movement standards. AI acts like a "virtual therapist," monitoring movement correctness in real time for improved rehabilitation accuracy.

SAM 3D Body: Transformer Architecture

Meta uses a 3D pose regression system based on a Transformer encoder-decoder architecture. Input: Ordinary image → Output: 3D human body mesh + pose parameters. It doesn't predict keypoints, but directly predicts the complete 3D human body model.

mytv-android Currently the best Android TV live TV software

My TV Live TV software developed using native Android

2. Live Streaming Software APK + Live Streaming Source + TV Assistant Package Download:

Click to Download

Includes Live Streaming Software, Live Streaming Sources, and TV Assistant Package

1. No-Installation Deployment (One-Click Installation Package)

If you don't have time to read tutorials, don't want to manually download and install, or your network environment doesn't allow it, you can choose to directly open the model integration package below for no-manual deployment.

Z-Image Model Integration Package Download

⭐ Realistic Portrait Style (Natural Light, High-Quality Look)

A super realistic photo of an East Asian beauty. Her skin is naturally smooth, dark and lustrous, with a sweet smile. The warm and soft ambient lighting creates a cinematic portrait effect. The photo uses shallow depth of field, rich detail in the eyes, 8K ultra-high-definition resolution, photo-realistic feel, professional photography, extremely clear facial details, perfect composition, softly blurred background, and a fashionable, high-fashion style.

🌸Sweet Japanese Style

Adorable Japanese girl, dressed in casual school uniform style, soft pastel tones, sweet smile, delicate makeup, brown eyes, fluffy hair, bright sunlight, extremely cute aesthetic, magazine cover style, delicate skin texture, clear facial features, perfect lighting, HDR

💄Korean Cool and High-End Style

Korean fashion model, elegant and simple beauty, smooth straight hair, moist lips, perfectly symmetrical face, neutral studio lighting, Vogue-style photography techniques, delicate makeup, sharp eyes, high-end portrait lens effects, ultra-high definition image quality, fashionable and modern styling

Protocol Support

Implements SSH v2, Telnet, Raw TCP, Serial, and Shell protocols with comprehensive authentication support.

• SSH automatic authentication during session verification
• SSH ControlMaster, ProxyCommand or ProxyJump support
• SSH proxy and automatic login with password, public key, keyboard interaction, and gssapi-with-mic
• X11 forwarding and port forwarding (direct/local, reverse/remote, dynamic)

File Management

Integrates SFTP and SCP clients with comprehensive file operations.

• Downloading, uploading, deleting, renaming, and creating files/directories
• Integrated local file manager with full file operations
• XModem, YModem, and ZModem support

Shell Support

Supports multiple shell environments across different operating systems.

• Windows: Cmd, PowerShell, and Cmd/PowerShell as administrators
• Linux: bash, zsh, PowerShell core, etc.
• macOS: bash, zsh, PowerShell core, etc.

Step 1: Download ComfyUI

Download the latest version of ComfyUI: Click to go

If you have already installed the ComfyUI client, it is recommended to upgrade it to the latest version.

Step 2: Download the JSON Workflow

Download the JSON workflow Click to get. Drag and drop it into the workflow; the required models will be downloaded automatically. If you don't have an external network environment, you'll need to use a VPN or proxy and enable TUN global mode!