Ashraff Hathibelagal

Tried video editing using Codex (GPT-6-Luna). The original video was generated using Grok Imagine. Codex used hyperframes to add in cuts, text, and transitions. The result is much better than the original video...

Female versions of Draco (long straight platinum hair, green eyes, green-and-silver tie, slight smirk) have existed in fan art for over a decade. But there's a gender-swap fan art wave in progress... I tried generating my own little short clip using Minimax H3 and Krea 2 Turbo

Asked Grok Imagine in agent mode to create a stunning product launch video for the new Tesla Roadster. This was the first time I tried the agent mode, and it was a simple one-line prompt. Very impressive result...

Just used Claude Sonnet 5.5 to create a music video, without Seedance or any such video generation models. It made all this using Hyperframes... All I had to say to it was "use hyperframes to create a stunning lyrics video for this song" 🤯

Testing lip sync with minimax h3. Using the guide node in comfyui definitely helps. I think this result is quite good...

Experimenting with special effects using Minimax H3. In this video, continuous blue-white lightning traces the guy's outline... his tribal tattoos pulse with inner light, electric sparks and motion trails follow his limbs, and faint energy ripples spread across the floor with each step.

Trying to create action scenes using Minimax H3. Here's an anthropomorphic snow leopard circling and striking a large, heavy bag inside a modern gym with dark speckled flooring, metal framing, and large glass walls revealing an adjacent room I think the scene looks realistic...

More prompting practice. Tried creating a video today with a realistic nighttime phone-video style with high-contrast street lighting, deep shadows, and a slightly cool color temperature. Used Minimax H3 with Grok and Krea 2 Turbo.

Experimenting with character intros using Grok Imagine and Krea 2 Turbo. Here's a high-end title sequence that treats four characters as opposing visual systems. Includes premium motion design. The theme is dark editorial, and photorealistic. The visual language combines dramatic studio cinematography, kinetic typography, graphic transitions, film grain, and atmospheric particles.

Just a test clip. I think I'm going to make a movie using this character.

Another test video I created using Krea 2, Minimax H3, and Grok.

He has nothing to offer. Only this. Single-shot performance, about 12 seconds. Locked eye-level full-body frame, no cuts, no camera move. Subject stays centered for the whole take... made using minimax h3 r2v.

Testing character swap using Minimax-H3 R2V. You don't even need a character sheet. I just created portraits of these two characters using Krea 2. You really don't need Seedance 2.5, imo. Minimax H3 running locally is good enough most of the time...

Had to try creating those "dancing cat" videos that are trending on Instagram. Used Minimax H3 R2V to make this 🐱

Created another infographic using ChatGPT. ChatGPT is very good at generating educational infographics, especially when you want the content, layout, and visuals created from one prompt. You can go from "explain this grammar point" to a finished infographic very quickly...

Hit my Grok Imagine usage limit, so I tried Minimax R2V running locally instead. Minimax R2V is far better. The actors show a lot more realistic emotions. I used the Turbo 4-step LoRA, so it took just a few minutes to generate two 13s-long clips. Used Grok to write the prompts

Created two characters and the first frame using Muse Image. Those were the three references I passed to Grok Imagine. The prompt for the video was generated using Claude Sonnet.

Cinematic strength-training video showing extremely muscular anthropomorphic animals. Made with Grok Imagine. Because I was using references, the text prompt is almost entirely about action, choreography, and cinematography.