Instructions to use infosave/MiniMax-H3-Turbo-cmf with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- cortiq
How to use infosave/MiniMax-H3-Turbo-cmf with cortiq:
# one Rust binary, no additional dependencies cargo install cortiq-cli # or a prebuilt binary from github.com/infosave2007/cmf/releases hf download infosave/MiniMax-H3-Turbo-cmf --include "*.cmf" --local-dir . ls *.cmf # some repos ship more than one quantization
cortiq animate FILE.cmf --prompt "a corgi in a chef hat flipping a pancake" --out clip.avi
- Notebooks
- Google Colab
- Kaggle
Commit History
4 days ago
-
Add discoverability tags
Verified signatureAdd verified format and task metadata without changing model files.899b706
-
Upload README.md with huggingface_hub
Verified signature7c4e218 -
Upload samples/refine_ab_refined_1024x576.mp4 with huggingface_hub
Verified signature051be0f -
Upload samples/refine_ab_draft_512x288.mp4 with huggingface_hub
Verified signaturec96a8da
-
Upload mmh3-raven-streaming-q4tp.cmf with huggingface_hub
Verified signature5be0529 -
0.5.94: the q8_2f build runs on the card — 357.3 s to 171.5 s
Verified signature24d7fb2 -
the q8_2f build's host-path cost, measured
Verified signature7419377 -
Upload mmh3-turbo-clipproj4b-fl2va-v2-q8_2f.cmf.sha256 with huggingface_hub
Verified signature70d3fcb -
the two-field eight-bit build: q8_2f, 26.90 GB, cortiq verify clean
Verified signature6abc462 -
0.5.93: the latent upscaler, the q8_2f build, reference clips
Verified signature1669c83 -
Upload images/upscaler_lanczos_1024.png with huggingface_hub
Verified signature2e07ddf -
Upload images/upscaler_net_1024.png with huggingface_hub
Verified signature6ab8d7e -
the ported latent upscaler: 512x288 render decoded at 1024x576
Verified signature147dc2a -
the vision-row ClipProj twin fitted for the v2 build, so the container can be rebuilt
Verified signaturef4485d3 -
the adapter's cost re-measured on a machine with headroom: 1.06x, which is the noise floor
Verified signaturebe7e6c4 -
adapter A/B still: lora
Verified signature9a66017 -
adapter A/B still: base
Verified signature3d02815 -
adapter A/B sample: routed
Verified signature0257265 -
adapter A/B sample: lora
Verified signature435887c -
adapter A/B sample: base
Verified signature37516c3 -
0.5.92: runtime adapters, the branch router, and what this port does not do yet
Verified signature72a3ac3
-
card: field notes from discussions #1/#2/#4 — 24 GB frame budgets, keyframe VAE encode on CPU, kill disarmed during prompt encode (0.5.80), P-core QoS
Verified signature41b5099
-
clipproj4b-fl2va v2: split projection — vision rows through a vision-fitted map, faces hold
Verified signature9e08d5a -
clipproj4b-fl2va v2: split projection — vision rows through a vision-fitted map, faces hold
Verified signature0540810 -
images: 4x face-crop A/B panels, clipproj v1 vs v2
Verified signature0b67068 -
images: 4x face-crop A/B panels, clipproj v1 vs v2
Verified signaturea422bf6 -
images: 4x face-crop A/B panels, clipproj v1 vs v2
Verified signatureebc0c6f -
images: 4x face-crop A/B panels, clipproj v1 vs v2
Verified signatureb4b9e11 -
images: clipproj v1 vs v2 face A/B frames
Verified signature1e40c25 -
images: clipproj v1 vs v2 face A/B frames
Verified signature30e0074
-
card: the compact fl2va file — start/end frames on 16-24 GB machines
Verified signature77201ae -
clipproj + fl2va in one 14.5 GB file: the 4B tower and the VAE encoder join the compact build; first-frame conditioning verified frame-for-frame
Verified signaturea8020b6
-
README: lead with the three-way size comparison, decide by VRAM up front
Verified signaturee8a0878 -
README: one 'Making it smaller' section, a Which-file table keyed on VRAM, three-way sample comparison
Verified signature0914add -
ClipProj section: scope the timings to the card they were taken on
Verified signature79fdefb -
ClipProj 4B encoder variant: 23.47 GB -> 13.16 GB, 2x faster, still four bits everywhere
Verified signature7097d8d -
ClipProj 4B prompt encoder: 25.20 GB -> 14.13 GB, everything still q4tp
Verified signature2088e7b -
samples: ClipProj 4B variant, same prompt and seed as the q4tp/q2tp A-B
Verified signature4d25f5d -
samples: ClipProj 4B variant, same prompt and seed as the q4tp/q2tp A-B
Verified signature8d8ee49 -
samples: ClipProj 4B variant, same prompt and seed as the q4tp/q2tp A-B
Verified signaturee9805ba
-
add the host-only row: 646.1 s against 60.2 on the card, same machine and seed
Verified signature94c6382 -
0.5.68: measured on the current build — 60.2 s at four steps, per-stage table, and how to use two cards
Verified signature60fd8b2 -
0.5.68: the render is on the card end to end — 91.6 s at 8 steps, and a seed finally reproduces
Verified signature76c9afc
-
card: name the library that reads this file
Verified signature701638b -
Upload README.md with huggingface_hub
Verified signaturee14c9c9 -
Upload mmh3-turbo-fl2va-q2tp.cmf with huggingface_hub
Verified signature921dce2 -
Upload README.md with huggingface_hub
Verified signature4fa8278 -
Upload samples/ab_q4tp.mp4 with huggingface_hub
Verified signaturefca9023 -
Upload samples/ab_q2tp.mp4 with huggingface_hub
Verified signaturef187d9b -
Upload samples/ab_q2tp.gif with huggingface_hub
Verified signature3891fa7