Minimax H3 Image is fun, trains well in my early testing. Feels malleable despite some flaws with smaller details. These are some sample outputs from my first run of training a LoRA in ai-toolkit. It's super fast at inference and there seems to be plenty of seed variance.
@RDKLInc Picked up a cheap 1st gen Kindle Scribe this week for this purpose, used Claude Code and SSH to get jailbreak going and KOReader installed, been having a blastβ¦Found a plugin that can connect to and browse public Calibre servers and Z Library. Totally different device now. ππΌππΌ
All she wanted was her keys. Her husbandβs outside mowing the lawn, her son is turning into a furry creature, and Kimberley has a boy hidden in her wardrobe.
Meet Kimberleyβs family. Suddenly, her missions with The Super Spies donβt look quite so stressful.
Jimmy Stewart and zombies, mashed up the old fashioned way. Our short film Itβs a Wonderful Night of the Living Dead took a year to make, ended up playing in a few Oscar-qualifying film festivals, and is now available to watch from your toilet.
Trying to figure out longer form video generation using context latent and sharper video with #MiniMax#H3 using https://t.co/ijmGOtLvJy and https://t.co/BlsOMPgGMe - Still has noticeable seams at the join though.
Seedance is so foolish it's hard to justify using anything else.
I have a tool built that renders this type of v2v basically 2 mouse clicks
Also, don't sleep on draft mode... goated feature (for v2v at least)
Source: Samy Elbanna on FB π
Video training is admittedly one of my weakest skills, it is time intensive and you have to align a lot of things to get decent results. Trained on MiniMax H3 and I love how this came out.
Took about 15 hours training time on 33 clips overnight. The vid got compressed so Iβll add the full version to my site.
@dt_sqr@fal img to svg is only 10 cents, vs an Adobe Illustrator sub for it's tools or https://t.co/8ai3CNpz7i which is about $10 a month for something you might not use in a month. It's nice to have the option of just calling an API and getting work done.
Was looking around for Halloween window projection videos, I was bored by most of what I found so I made some with Krea and H3. Had @claudeai make a preview tool based on a photo of my windows that fit 16:9 perfectly to see how they might look. @krea_ai@Hailuo_AI
We just open-sourced an entire 3D team for Blender.
Introducing the Scenario plugin.
Use it inside Blender, or let Claude Code run it through MCP.
π/scenario-plugin
β
β£ π Concept Artist
β β£ π Concept Art From a Prompt
β β π References From Your Viewport
β
β£ π 3D Artist
β β£ π Meshes From Text or a Photo
β β π Retexture or Remesh the Selected Mesh
β
β£ π Animator
β β£ π Rig the Selected Character
β β π Motion From Text
β
β£ π Surfacing Artist
β β£ π PBR Materials Onto Any Slot
β β π Maps From a Photo
β
β£ π Layout Artist
β β£ π Greybox From a Description
β β π 360 Worlds and Skies
β
β£ π Camera Operator
β β£ π Camera Path From a Sentence
β β π Orbits, Dollies, Cranes and Flyovers
β
β£ π Render Artist
β β£ π Viewport to Finished Still
β β π Playblast to Finished Shot
β
β π Sound Designer
β£ π Music, Voice and SFX
β π Dropped on Your Sequencer
Comment "BLENDER" and I'll send you the GitHub repo.