Web developer building apps with Generative AI tools. Sharing AI demos, experiments, and creations. New Zealander living in Peru.

Cusco
I tested all of the Best Avatar Models on @FAL to find which creates the most realistic product in hand videos. Setup: • Generated a product shot (lady + body lotion) using Wan 2.2 Instagirl LoRA, Ideogram Character and Qwen Image. • Created a custom voice with MiniMax Speech-02 HD • Generated a silent video with Wan 2.2 image-to-video Which wins? Let's see 🧵👇
25
48
530
153,324
I tested the top text-to-video AI models to find the most realistic "yapping doctor" video. Dreamina, Flux, Grok, MiniMax, ​Omni, Wan. Full prompt below. Which one wins?
1
301
6. ByteDance Dreamina Seedance 2.5 $0.2205 per second at 480p
Made with AI
1
61
Text-to-video prompt used: Subject & Attire: Middle-aged Caucasian male neurologist, salt-and-pepper hair, forehead lines, thin black rectangular glasses. Sits slightly off-centre in a black leather executive chair, wearing blue scrubs under a white lab coat embroidered with "Dr Fraser, MD, PhD / Neurology". Silver wedding band and watch visible on wrist. Setting & Framing: Native 9:16 vertical, medium close-up, front-facing camera. Background has tightly closed black horizontal blinds blocking nearly all outside light, save for a faint daylight slit at the windowsill. Viewer's left shows a desk with papers and an illuminated lava lamp. Realistic office lighting and natural acoustics. Action & Dialogue Timeline: 0:00–0:02: Waves right hand (viewer's left, palm partially cropped): "Good morning everyone. I'm Dr. Fraser." 0:02–0:06: Counts by tapping left fingers with right index finger: "I'm a neurologist who focuses on brain health, functional medicine, peptides, nutrition..." 0:06–0:08: Opens hands outward and upward: "...and optimizing how the body and brain perform." [0:08 Jump Cut]: Quick edit cut abruptly skips a breath, cleanly snapping torso and hands into resting position. 0:08–0:15: Holds hands interlocked at chest level: "I get this asked in the clinic almost daily: 'Doc, what do you think of peptides?'" Based on the concept by @Mho_23 x.com/Mho_23/status/20989292…
just a reminder that you can make AI videos this realistic for less than $0.50... most people still have no idea this is possible
101
Wan 3.0 at 5 cents per second is crazy good. I generated several 5-second UGC-style clips, took just over 3 minutes for each. Price is for 480p: generate cheap at scale, upscale only the winners. Verdict: Let's gooooo!
155
Spent the afternoon at Ancasmarca, an archaeological site at 4,000m (13,000ft). Such a special atmosphere, interesting ruins, and not another soul there.
1
1
84
Ideogram 4 can also generate seamless tiles! 😍
1
68
Graphic design was a real weakness for AI images but with Ideogram 4 it becomes a strength. Suddenly prompting looks just like using Figma. I vibecoded this tool to sketch boxes and play with the layout. Iterate on layouts in less than 10 seconds.
1
80
Controlling layout in AI images used to be an art Ideogram 4 makes it a science. Just draw your bounding boxes and annotate them. Crazy cool. This is next level.
1
84
🚨Cosmos 3 Super - Text to Image - A new open weights image model ranked #1 - Works best with automatic prompt expansion - It uses a multi-turn agentic harness to automatically evaluate and improve images across several rounds Tested on @fal with prompt expansion and agentic generation enabled.
2
93
phantom_brush image model on @arena loving the detail 🔥
1
37
What happened to the Krea 2 hype? It's a very capable image model. Great for style and creativity. Worth playing with.
51
MAI Image 2 is a top 10 text-to-image model by Microsoft AI. Strong on photorealism amd portraits. Max output 1024x1024 pixels. Available in their playground.
Made with AI
46
Checking out the new fal assets. Its really cool, it gives you a description for each image. I uploaded a folder of AI images from 2023. Then used Recraft 4.1 to generate new versions.
1
6
517
I built a fal connector for WordPress 7's new AI Connectors system. AI media generation right inside of WordPress. 😎
76
Great aesthetics from the new Recraft v4.1 Utility Text to Image model. 😻
🚨 Recraft V4.1 is here on fal! ✨ Built for high aesthetics and emotional imagery, not generic stock looks. 🎨 Stronger personality in 3D, illustrations and soft gradients. 🖼️ Image, vector, and utility flavors. Standard and Pro tiers for any scale.
72