Technology Related Blog

  Flux / Midjourney Here is the full picture: 1. Flux / Midjourney - Are they inside LTX? No, separate. LTX = only video. It takes an image and makes it move. Flux = image generator made by Black Forest Labs - successor to Stable Diffusion. It's open-source and free. 12 billion params. You run it locally on your 4060 Ti. Midjourney = image generator - paid, artistic style, $10/month basic. Runs on Discord. Workflow is 2 steps: Generate Indian characters image with Flux (free) or Midjourney Feed that image to LTX API to make video of them walking / talking. Flux is not included in LTX, you have to run it separately, but both can run in same ComfyUI. 1. Can you run Flux / Midjourney on RTX 2050 4GB? Midjourney: No install needed, runs on their website/Discord - works on any laptop. $10/month. Flux: No, your 2050 can't run it properly. Official FP16 needs 24GB VRAM Quantized Q8 needs ∼13GB Lowest Q4 needs 6-16GB Your RTX 2050 is 4GB . Even 8GB cards need special 4-bit qua...

  RunPod -- GPUS with pre installed configured MOdels -- and  cheap for running model per sec pulse --  https://salad.com/pricing -- very cheap -- Why salad so cheap: We utilize unused, latent compute resources from gamers and e-sports arenas worldwide to power our network — think Airbnb for GPUs. RunPod = real datacenters in US/EU - reliable, fast. Salad = thousands of gamer PCs worldwide sleeping at night - they rent their GPU when not gaming. That's why it's cheap. Salad good: Insanely cheap ($0.02/hr for RTX 5090) , 191 countries Salad bad: Distributed = unpredictable availability, Max 24GB VRAM per GPU, Not for low-latency workloads RunPod: Per-second billing Salad: You only pay for time hardware is available, they don't charge for cold boot. Hourly calculator, but billed per second. Both per hour, not per month, but you can keep running 720 hrs = monthly. Is Salad cheapest? For batch jobs / non-critical - Yes, Salad is cheapest in world at RTX 3090 $0.09/hr, RTX 40...

  Cheapest 24GB card to run both: RTX 4090 is the entry 24GB consumer card - 24GB GDDR6X - Rent from $0.69/hr on RunPod, Community Cloud from $0.34/hr RTX A5000 also 24GB - $0.27/hr - slower but cheaper, good for Wan -- best for server for long run RTX 3090 also 24GB - $0.124/hr on SaladCloud - cheapest 24GB if you find it  2 times faster than a5000  for online --  Why so cheap: We utilize unused, latent compute resources from gamers and e-sports arenas worldwide to power our network — think Airbnb for GPUs. RunPod = real datacenters in US/EU - reliable, fast. Salad = thousands of gamer PCs worldwide sleeping at night - they rent their GPU when not gaming. That's why it's cheap. Salad good: Insanely cheap ($0.02/hr for RTX 5090) , 191 countries Salad bad: Distributed = unpredictable availability, Max 24GB VRAM per GPU, Not for low-latency workloads

  4. Why and when to use Wan 2.1 vs LTX? And what's cheapest 24GB card? LTX 2.3: 22B params, Max 2560×1440 at 50 FPS, audio-native ✅, 10-14x faster than Wan. Best for product shots, portrait 9:16, social media, rapid iteration. -- need 16GB GTX card Wan 2.1: 14B params, Max 1280×720 at 24 FPS, motion coherence Very Good, identity preservation High. Best for nature, cinematic motion, realistic water/trees/faces. Rule of thumb: Need speed + vertical + cheap 1-min video? LTX Need best nature quality where leaves/water look real? Wan 2.1 - it has less artifacts on complex motion Cheapest 24GB card to run both: RTX 4090 is the entry 24GB consumer card - 24GB GDDR6X - Rent from $0.69/hr on RunPod, Community Cloud from $0.34/hr RTX A5000 also 24GB - $0.27/hr - slower but cheaper, good for Wan RTX 3090 also 24GB - $0.124/hr on SaladCloud - cheapest 24GB if you find it

  Here is short blog text you can copy: How to connect any Laptop to External GPU - 4 Methods Method 1: Thunderbolt / USB4 eGPU Enclosure - Easy but costly Box like Razer Core X V2. You plug one Thunderbolt cable, hot-plug while laptop is on. Cost: Rs. 35k just box + GPU cost. Needs laptop with Thunderbolt 3/4 or USB4. Pros: Easy, 1 cable charging + USB hub, no opening laptop. Flaws: Expensive, only 40Gbps (PCIe x4), 15% slower than desktop, cannot handle RTX 4090/5090 power. https://link.amazon/B02jSvOGI Method 2: OCuLink via M.2 NVMe - Cheapest & Fastest for AI Board like MINISFORUM DEG1 + M.2 to OCuLink adapter. Open laptop, put adapter in 2nd M.2 slot, cable out to DEG1 board + ATX PSU + GPU. Cost: Rs. 9.5k board + Rs. 2.1k adapter + Rs. 4.8k 650W PSU = ~Rs. 16k + GPU. Pros: Cheapest, 64Gbps faster than Thunderbolt, fits any GPU including RTX 4060 Ti 16GB, 4090. Flaws: Need to open laptop, warranty void, shutdown before unplug, cable fragile, no case. https://www.youtube.co...