Alex995647 commited on
Commit
8409e6e
·
verified ·
1 Parent(s): f352601

Add Turbo v4 Step-600 (pruned conversion)

Browse files
.gitattributes CHANGED
@@ -371,3 +371,9 @@ ref2v-turbo-4-step-v0-1-rank-21/example_images/05_video.mp4 filter=lfs diff=lfs
371
  gordon-pinewood/example_images/01_video.mp4 filter=lfs diff=lfs merge=lfs -text
372
  painterly-h3-by-cosmic/example_images/01_video.mp4 filter=lfs diff=lfs merge=lfs -text
373
  painterly-h3-by-cosmic/example_images/02_video.mp4 filter=lfs diff=lfs merge=lfs -text
 
 
 
 
 
 
 
371
  gordon-pinewood/example_images/01_video.mp4 filter=lfs diff=lfs merge=lfs -text
372
  painterly-h3-by-cosmic/example_images/01_video.mp4 filter=lfs diff=lfs merge=lfs -text
373
  painterly-h3-by-cosmic/example_images/02_video.mp4 filter=lfs diff=lfs merge=lfs -text
374
+ turbo-v4-step-600-pruned-conversion/example_images/01_image.webp filter=lfs diff=lfs merge=lfs -text
375
+ turbo-v4-step-600-pruned-conversion/example_images/02_image.png filter=lfs diff=lfs merge=lfs -text
376
+ turbo-v4-step-600-pruned-conversion/example_images/03_image.webp filter=lfs diff=lfs merge=lfs -text
377
+ turbo-v4-step-600-pruned-conversion/example_images/04_video.mp4 filter=lfs diff=lfs merge=lfs -text
378
+ turbo-v4-step-600-pruned-conversion/example_images/05_poster.jpg filter=lfs diff=lfs merge=lfs -text
379
+ turbo-v4-step-600-pruned-conversion/example_images/05_video.mp4 filter=lfs diff=lfs merge=lfs -text
turbo-v4-step-600-pruned-conversion/example_images/01_image.webp ADDED

Git LFS Details

  • SHA256: 669addd55c4fb48afc4ae21f38d0f3b2744679275473bade6ec8297f04f86104
  • Pointer size: 131 Bytes
  • Size of remote file: 250 kB
turbo-v4-step-600-pruned-conversion/example_images/02_image.png ADDED

Git LFS Details

  • SHA256: bea9f0cd32279b0dd7f84e07f39abd8762852e7f1287414a909a40dd16aa22d6
  • Pointer size: 132 Bytes
  • Size of remote file: 4.42 MB
turbo-v4-step-600-pruned-conversion/example_images/03_image.webp ADDED

Git LFS Details

  • SHA256: 84bdbcc616cc35ef3cf3e0080c4f366d24d5002f9fa897eb87fd163613ad9ef1
  • Pointer size: 131 Bytes
  • Size of remote file: 245 kB
turbo-v4-step-600-pruned-conversion/example_images/04_poster.jpg ADDED
turbo-v4-step-600-pruned-conversion/example_images/04_video.mp4 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:63a036f90f5aa07b65e64bf3a6cba8ba1ca02d3923c1bc86e2f127d37d309d63
3
+ size 12557337
turbo-v4-step-600-pruned-conversion/example_images/05_poster.jpg ADDED

Git LFS Details

  • SHA256: 4a5545ca484d7d3176c2fe8055e577cea33fc621d2c67ca9f16db9ec2ffd9d5f
  • Pointer size: 131 Bytes
  • Size of remote file: 293 kB
turbo-v4-step-600-pruned-conversion/example_images/05_video.mp4 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7bef43325ab365dd4ae99cae31cc1696a3f3fbb6dee19bb808b76465869f4e6b
3
+ size 11417915
turbo-v4-step-600-pruned-conversion/info.txt ADDED
@@ -0,0 +1,147 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ═══════════════════════════════════════════════════════════════════════════════
2
+ TURBO V4 STEP-600 (PRUNED CONVERSION)
3
+ ═══════════════════════════════════════════════════════════════════════════════
4
+
5
+ ───────────────────────────────────────────────────────────────────────────────
6
+ IDENTITY
7
+ ───────────────────────────────────────────────────────────────────────────────
8
+ Base model MiniMax H3
9
+ Generation mode Reference-to-Video (Ref2VA), Text-to-Video (T2VA)
10
+ Checkpoint fl2va (drives T2VA / I2VA / L2VA / FL2VA)
11
+ Category Speed / step-reduction
12
+ Author Halpak
13
+ Source https://huggingface.co/Halpak/MiniMax-H3-Turbo-Lora-ComfyUI
14
+ Provenance HuggingFace · Halpak/MiniMax-H3-Turbo-Lora-ComfyUI · minimax_h3_turbo_v4_step600_pruned_comfyui.safetensors
15
+
16
+ ───────────────────────────────────────────────────────────────────────────────
17
+ WEIGHTS
18
+ ───────────────────────────────────────────────────────────────────────────────
19
+ File minimax_h3_turbo_v4_step600_pruned_comfyui.safetensors
20
+ Version main
21
+ Size 591.6 MB
22
+ Format SafeTensor / unknown precision
23
+ SHA256 4f95afb2ab503eac876ae5571f4d78b0e0fd0198b626a89f4b7727018ed198a9
24
+ Published 2026-08-30
25
+
26
+ ───────────────────────────────────────────────────────────────────────────────
27
+ TRIGGER WORDS
28
+ ───────────────────────────────────────────────────────────────────────────────
29
+ No trigger word. The LoRA applies as soon as it is loaded —
30
+ strength is the only way to dial the effect up or down.
31
+
32
+ ───────────────────────────────────────────────────────────────────────────────
33
+ RECOMMENDED SETTINGS
34
+ ───────────────────────────────────────────────────────────────────────────────
35
+ Taken verbatim from the author's description:
36
+
37
+ Sampler steps 4
38
+
39
+ ───────────────────────────────────────────────────────────────────────────────
40
+ ADOPTION & POPULARITY
41
+ ───────────────────────────────────────────────────────────────────────────────
42
+ Downloads 385
43
+ Rating 100% positive (2/2)
44
+ Comments 0
45
+ Published 6 days ago
46
+ Download rate 67.5/day since release
47
+
48
+ Adoption rank #58 of 114 in this collection (Modest)
49
+ Momentum rank #40 of 114 by download rate (Steady · fresh release)
50
+
51
+ Adoption tier is the percentile of total downloads within this collection;
52
+ momentum is the percentile of downloads-per-day since release. Momentum is
53
+ ranked rather than measured against a fixed rate, because this base model
54
+ is itself new and every LoRA looks fast on an absolute scale. Both numbers
55
+ are CivitAI's, read on the scrape date at the foot of this file.
56
+
57
+ ───────────────────────────────────────────────────────────────────────────────
58
+ CHAINING / STACKING
59
+ ───────────────────────────────────────────────────────────────────────────────
60
+ Stacks with others not stated
61
+
62
+ The author gave no chaining guidance. The rules below are the
63
+ general MiniMax H3 ones, not anything specific to this LoRA.
64
+
65
+ General MiniMax H3 rules that apply here (see REFERENCE.txt):
66
+
67
+ • Load the LoRA with LoraLoaderModelOnly between UNETLoader and your
68
+ sampler; stacking means chaining several of those nodes in series.
69
+ • This is a step-distillation (turbo) LoRA. DISTILL LORAS DO NOT COMPOSE
70
+ — running this alongside another turbo/PDD/lightx2v LoRA breaks
71
+ generation. Pick exactly one.
72
+ • Style and character LoRAs stack on top of it normally, with no special
73
+ strength penalty beyond ordinary multi-LoRA tuning.
74
+ • Use Euler so the trained sigma schedule is consumed as intended, and
75
+ keep CFG at 1.0 with BasicGuider.
76
+ • Only the step counts the LoRA was trained for (usually 4, 6 or 8) work.
77
+ Other values produce noise — it is not a free dial.
78
+ • Chain this only onto an fl2va checkpoint (T2VA / I2VA / L2VA / FL2VA).
79
+ Loading it against ref2va can degrade output silently rather than
80
+ erroring.
81
+
82
+ ───────────────────────────────────────────────────────────────────────────────
83
+ PROS
84
+ ───────────────────────────────────────────────────────────────────────────────
85
+ + Author documents use across several modes (REF2V, T2V). [source:
86
+ description]
87
+ + Author published concrete recommended settings. [source: description]
88
+
89
+ ───────────────────────────────────────────────────────────────────────────────
90
+ CONS & CAVEATS
91
+ ───────────────────────────────────────────────────────────────────────────────
92
+ - No trigger word — the LoRA is always on once loaded, so strength is the
93
+ only control. [source: model metadata]
94
+ - Modest adoption (385 downloads) — less community feedback to rely on.
95
+ [source: CivitAI stats]
96
+
97
+ ───────────────────────────────────────────────────────────────────────────────
98
+ COMPATIBILITY WARNING
99
+ ───────────────────────────────────────────────────────────────────────────────
100
+ MiniMax H3 ships two transformers that share tensor key names: the fl2va
101
+ checkpoint (driving T2VA, I2VA, L2VA and FL2VA) and the separate ref2va
102
+ checkpoint (driving Ref2VA). Loading a LoRA against the wrong family can
103
+ silently degrade output instead of raising an error, so confirm the family
104
+ above matches the checkpoint you have loaded. Likewise, LoRAs built for
105
+ full checkpoints often throw adaln_proj shape errors on pruned ones — match
106
+ the build to your checkpoint.
107
+
108
+ ───────────────────────────────────────────────────────────────────────────────
109
+ AUTHOR'S DESCRIPTION
110
+ ───────────────────────────────────────────────────────────────────────────────
111
+ Non-EMA counterpart of the recommended v4 step-600 Turbo LoRA, in the same
112
+ pruned/curve-form ComfyUI layout. Distinct weights from the EMA build
113
+ (different SHA256); the README keeps both because their tensor values
114
+ differ.
115
+
116
+ ───────────────────────────────────────────────────────────────────────────────
117
+ EXAMPLE MEDIA
118
+ ───────────────────────────────────────────────────────────────────────────────
119
+ 5 preview(s) in ./example_images/ — community results for
120
+ this LoRA. Videos keep a matching _poster.jpg still frame.
121
+
122
+ No prompts are recorded below: CivitAI's public API flags these posts as
123
+ having generation metadata but does not return it, so the prompts cannot be
124
+ scraped. Open the model page to read them.
125
+
126
+ 01_image.webp (image, )
127
+ 02_image.png (image, )
128
+ 03_image.webp (image, )
129
+ 04_video.mp4 (video, )
130
+ still: 04_poster.jpg
131
+ 05_video.mp4 (video, )
132
+ still: 05_poster.jpg
133
+
134
+ ───────────────────────────────────────────────────────────────────────────────
135
+ HOW TO USE (COMFYUI)
136
+ ───────────────────────────────────────────────────────────────────────────────
137
+ 1. Copy the .safetensors file into ComfyUI/models/loras/
138
+ 2. Load the matching H3 checkpoint with UNETLoader.
139
+ 3. Insert LoraLoaderModelOnly between the loader and your sampler.
140
+ 4. Set strength as listed above, then add any trigger word to the prompt.
141
+ 5. Keep CFG at 1.0 with BasicGuider unless you know you want real CFG.
142
+
143
+ ───────────────────────────────────────────────────────────────────────────────
144
+ Scraped by lorakit on 2026-09-04 23:09 UTC.
145
+ Stats and description are the author's; pros/cons are derived from the
146
+ evidence tagged beside each line.
147
+ ───────────────────────────────────────────────────────────────────────────────
turbo-v4-step-600-pruned-conversion/metadata.json ADDED
@@ -0,0 +1,144 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "name": "Turbo v4 Step-600 (pruned conversion)",
3
+ "slug": "turbo-v4-step-600-pruned-conversion",
4
+ "base_model": "MiniMax H3",
5
+ "media_type": "video",
6
+ "model_id": 915909949,
7
+ "version_id": 915909950,
8
+ "version_name": "main",
9
+ "source": "huggingface",
10
+ "url": "https://huggingface.co/Halpak/MiniMax-H3-Turbo-Lora-ComfyUI",
11
+ "repo_id": "Halpak/MiniMax-H3-Turbo-Lora-ComfyUI",
12
+ "file_path": "minimax_h3_turbo_v4_step600_pruned_comfyui.safetensors",
13
+ "author": "Halpak",
14
+ "category": {
15
+ "key": "accelerator",
16
+ "label": "Speed / step-reduction"
17
+ },
18
+ "modalities": [
19
+ "ref2v",
20
+ "t2v"
21
+ ],
22
+ "primary_modality": "ref2v",
23
+ "checkpoint_family": "fl2va",
24
+ "trigger_words": [],
25
+ "recommended_settings": {
26
+ "Sampler steps": "4"
27
+ },
28
+ "chaining": {
29
+ "author": {
30
+ "stacks": "not stated",
31
+ "notes": "",
32
+ "avoid": [],
33
+ "requires": [],
34
+ "order": "",
35
+ "strength": "",
36
+ "stated": false
37
+ },
38
+ "general_rules": [
39
+ "Load the LoRA with LoraLoaderModelOnly between UNETLoader and your sampler; stacking means chaining several of those nodes in series.",
40
+ "This is a step-distillation (turbo) LoRA. DISTILL LORAS DO NOT COMPOSE — running this alongside another turbo/PDD/lightx2v LoRA breaks generation. Pick exactly one.",
41
+ "Style and character LoRAs stack on top of it normally, with no special strength penalty beyond ordinary multi-LoRA tuning.",
42
+ "Use Euler so the trained sigma schedule is consumed as intended, and keep CFG at 1.0 with BasicGuider.",
43
+ "Only the step counts the LoRA was trained for (usually 4, 6 or 8) work. Other values produce noise — it is not a free dial.",
44
+ "Chain this only onto an fl2va checkpoint (T2VA / I2VA / L2VA / FL2VA). Loading it against ref2va can degrade output silently rather than erroring."
45
+ ]
46
+ },
47
+ "pros": [
48
+ "Author documents use across several modes (REF2V, T2V). [source: description]",
49
+ "Author published concrete recommended settings. [source: description]"
50
+ ],
51
+ "cons": [
52
+ "No trigger word — the LoRA is always on once loaded, so strength is the only control. [source: model metadata]",
53
+ "Modest adoption (385 downloads) — less community feedback to rely on. [source: CivitAI stats]"
54
+ ],
55
+ "file": {
56
+ "name": "minimax_h3_turbo_v4_step600_pruned_comfyui.safetensors",
57
+ "size_bytes": 620285592,
58
+ "size_mb": 591.55,
59
+ "sha256": "4f95afb2ab503eac876ae5571f4d78b0e0fd0198b626a89f4b7727018ed198a9"
60
+ },
61
+ "stats": {
62
+ "downloads": 385,
63
+ "thumbs_up": 2,
64
+ "thumbs_down": 0
65
+ },
66
+ "popularity": {
67
+ "downloads": 385,
68
+ "thumbs_up": 2,
69
+ "thumbs_down": 0,
70
+ "comments": 0,
71
+ "approval": "100% positive (2/2)",
72
+ "age_days": 5.7,
73
+ "downloads_per_day": 67.49,
74
+ "rank_by_downloads": 58,
75
+ "rank_by_momentum": 40,
76
+ "of_total": 114,
77
+ "adoption_tier": "Modest",
78
+ "momentum": "Steady · fresh release"
79
+ },
80
+ "selection_score": 409.0,
81
+ "selection_reasons": [
82
+ "385 downloads",
83
+ "2 thumbs-up",
84
+ "top-ranked for its category"
85
+ ],
86
+ "example_media": [
87
+ {
88
+ "file": "01_image.webp",
89
+ "type": "image",
90
+ "resolution": "",
91
+ "prompt": "",
92
+ "source_path": "assets/1.webp"
93
+ },
94
+ {
95
+ "file": "02_image.png",
96
+ "type": "image",
97
+ "resolution": "",
98
+ "prompt": "",
99
+ "source_path": "assets/1a.png"
100
+ },
101
+ {
102
+ "file": "03_image.webp",
103
+ "type": "image",
104
+ "resolution": "",
105
+ "prompt": "",
106
+ "source_path": "assets/2.webp"
107
+ },
108
+ {
109
+ "file": "04_video.mp4",
110
+ "type": "video",
111
+ "resolution": "",
112
+ "prompt": "",
113
+ "source_path": "assets/MiniMax_H3_00060-audio.mp4",
114
+ "poster": "04_poster.jpg"
115
+ },
116
+ {
117
+ "file": "05_video.mp4",
118
+ "type": "video",
119
+ "resolution": "",
120
+ "prompt": "",
121
+ "source_path": "assets/MiniMax_H3_audio.mp4",
122
+ "poster": "05_poster.jpg"
123
+ }
124
+ ],
125
+ "tags": [
126
+ "minimax-h3",
127
+ "lora",
128
+ "adapter",
129
+ "comfyui",
130
+ "text-to-video",
131
+ "reference-to-video",
132
+ "audio-video",
133
+ "synchronized-audio",
134
+ "few-step",
135
+ "turbo",
136
+ "accelerated-inference",
137
+ "lightx2v",
138
+ "dynamic-rank",
139
+ "svd",
140
+ "safetensors"
141
+ ],
142
+ "scraped_at": "2026-09-04T23:09:13+00:00",
143
+ "path": "minimax-h3/turbo-v4-step-600-pruned-conversion"
144
+ }
turbo-v4-step-600-pruned-conversion/minimax_h3_turbo_v4_step600_pruned_comfyui.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4f95afb2ab503eac876ae5571f4d78b0e0fd0198b626a89f4b7727018ed198a9
3
+ size 620285592