Add same-day LTX and MiniMax workflows

2026-08-09 18:08:00 +00:00
parent 05e121776f
commit 7e10542b4b
12 changed files with 1223 additions and 0 deletions
@@ -0,0 +1,45 @@
+# MiniMax H3 Turbo v4 8-Step T2VA
+
+## Summary
+
+`MiniMax H3` text-to-video-plus-audio workflow tuned for the Turbo v4 LoRA and an 8-step recipe. This is the practical fast-path the same-day `BennyDaBall` release is aiming for.
+
+## Model Family
+
+- `minimax`
+
+## Status
+
+- `imported-only`
+
+## Source
+
+- Workflow pack page: <https://huggingface.co/BennyDaBall/MiniMax-H3-Workflows>
+- Workflow file: <https://huggingface.co/BennyDaBall/MiniMax-H3-Workflows/resolve/main/MiniMax_H3_Turbo_v4_8step_T2VA_BennyDaBall.json>
+- Original publisher: `BennyDaBall`
+- Date imported: `2026-08-09`
+
+## Developer notes
+
+- The source README says this workflow uses only core ComfyUI nodes.
+- It is set up around the `minimax_h3_turbo_v4_step600_ema_pruned_comfyui` LoRA.
+- The workflow bundles a prompt pattern for timed audio and dialogue control.
+
+## Our notes
+
+- This is the most immediately usable MiniMax release of the day for simple prompt-to-video testing.
+- If MiniMax feels like a real option on the `3070`, this is the first workflow I would try before the heavier reference-control packs.
+- Still likely tight on `8 GB` once you keep the official MiniMax model stack in memory, but the 8-step recipe is the main reason it is worth a look.
+
+## Required custom nodes
+
+- none beyond current official ComfyUI MiniMax support
+
+## Required models
+
+- `Comfy-Org/MiniMax-H3`
+- `drbaph/MiniMax-H3-Turbo-Lora-ComfyUI`
+
+## Notes
+
+- Best same-day MiniMax workflow to test first.