This AI Workflow Posts for Me While I Sleep

Comment "Workflow" to get the simplified workflow.

Want to create an AI talking avatar that sounds exactly like YOU — for free? In this step-by-step tutorial, I show you the full AI avatar creation workflow using three powerful free tools.
This is perfect for:

• ✅ AI news channels
• ✅ Social media content automation
• ✅ UGC (User Generated Content) campaigns
• ✅ Faceless YouTube channels (AI)
• ✅ Personal branding with AI

• 📦 MODELS & LINKS:

🎬 Main Model (LTX-2.3 Transformer)huggingface.co Lightricks/LTX-2huggingface.co

• 🔢 GGUF Variant (LTX-2.3 Dev):

ltx-2.3-22b-dev-Q8_0.gguf

unsloth/LTX-2.3-GGUFhuggingface.co

• 🔁 LoRA — Distilled Dynamic (22B):

Kijai/LTX2.3_comfyhuggingface.co 🔗huggingface.co

• 🔁 LoRA — IC Detailer:

Lightricks/LTX-2-19b-IC-LoRA-Detailerhuggingface.co

• 🗜️ VAE :

LTX23_audio+video_vae

Kijai/LTX2.3_comfyhuggingface.co

• 📐 Spatial Upscaler (2x):

ltx-2.3-spatial-upscaler-x2-1.0.safetensors

Lightricks/LTX-2.3huggingface.co

• 🧠 Text Projection:

ltx-2.3_text_projection_bf16.safetensors

Kijai/LTX2.3_comfyhuggingface.co

• 💬 Text Encoder — Gemma 3 12B (fp8):

gemma_3_12B_it_fp8_e4m3fn.safetensors

bartowski/gemma-3-12b-it-GGUFhuggingface.co

• 🎵 Audio Separator — MelBand Roformer:

Kijai/MelBandRoFormer_comfyhuggingface.co

• 🗣️ TTS — Qwen3 Voice Clone:

just install missing nodes and you will see where to download the models.

The Workflow

• Image Generation – Use NBP to generate a high-quality avatar image for free (use gemini)
• Voice Cloning – Clone your voice with Qwen 3 TTS for realistic, natural-sounding speech
• Video Creation – Bring it all to life using LTX 2.3 to generate a talking avatar video

• No subscriptions. No expensive software. Just a smart AI stack.

Timestamps

00:24Demo (What we're building) 02:56Workflow explaintation 06:35Clone your voice using Qwen 3 TTS 08:42Generate talking video with LTX 2.3 09:15Preview 09:45Other Samples i have created. 12:55Workflow executed video and Comparisions 15:01Console Timings 15:40Something i am working on 17:16Outro and crying

• ─────────────────────────

AI Jigyasa – Practical AI tutorials for creators, marketers & builders.
New videos every week. Subscribe so you don't miss a workflow.

• ─────────────────────────

• 💻 SYSTEM REQUIREMENTS:
• Minimum — 12GB VRAM (RTX 3060 12GB, 4060 Ti 16GB)

ComfyUI (latest version)

• 🔧 TROUBLESHOOTING:

Out of VRAM? Use FP8 quantized model or reduce resolution/length
Blurry results? Use spatial upscaler and increase steps
Inconsistent motion? Improve prompt detail and use negative prompts
Missing nodes? Update ComfyUI to latest version

• 📱 CONNECT WITH ME:
• 🔔 Subscribe for more AI tutorials
• 💬 Join our Discord: [ / discord ]
• 📸 Instagram: @ai_jigyasa
• 🐦 Twitter: @AshishKaprawan

This video is for educational purposes. LTX Video is developed by Lightricks and released under Apache 2.0 license. Always review generated content and respect copyright laws when using AI-generated media.

Like this tutorial? Hit that LIKE button and SUBSCRIBE for more cutting-edge AI tools and workflows!

• 🔻 Drop your generated videos in the comments - I want to see what you create! 🔻

#AIAvatar#TalkingAvatar#VoiceCloning#AITools2025#AIAutomation#FacelessYouTube#UGCContent#SocialMediaAI