Zhenzhen's AI Radar · MiniMax H3 · 2026-11-08 Morning Report
This issue features 10 new versions: Strongly Recommended 1 · Recommended 3 · Worth Watching 6 · Listed Only 0
Strongly Recommended (1)
[S] Batkin - MiniMax H3 · H3 V1
LoRA · New Model · Author AriaMaiven · Published 11-08 06:20 · Downloads 42
Ask plain H3 for a kitten with bat wings and you get a cat in a Halloween costume: stiff red dragon wings with bony spikes, glued to its ribs.
Image
Reasons for Recommendation
- Within 4 hours of release, it has already reached 42 downloads and 6 likes, with a download rate higher than 93%% of H3 models over the last 7 days
- Includes 5 sample videos, allowing you to preview the results before deciding to download
- The author provides specific parameters (1 instances), lowering the barrier to reproduction
- Free download, license allows image generation and in-site rental
How to Use
- Load this LoRA (591 MB) in the H3 workflow and include the trigger word "b4tk1n" in the prompt
- Author's parameters: "Weight 1.0."
Risks and Limitations
- 3 sensitive-level examples have been hidden
Trigger Word: b4tk1n
Recommended (3)
[A] P-H3-Image Generate + Edit | 8 Pictures + Auto Prompt · v1.1
Workflow · Version Update · Author AusBoss · Published 11-08 08:49 · Downloads 26
Edit a picture in plain words or make a new one from scratch. P-H3 makes its best pictures at 8 MP. Smaller sizes can give blurry faces and artifacts.
Image
Reasons for Recommendation
- The 2th H3 version of "P-H3-Image Generate + Edit | 8 Pictures + Auto Prompt". This update: Added a Size choice: 4 MP or 8 MP, for edits and new pictures. 8 MP gives faces and fine detail more room, with more memory and time, and P-H3 makes its best new pictures at 8 MP.
- Within 1 hours of release, it has already reached 26 downloads and 4 likes, with a download rate higher than 95%% of H3 models over the last 7 days
- Includes 1 sample videos and 1 sample images, allowing you to preview the results before deciding to download
- Documentation is relatively complete (approx. 3450 words)
- Free download, license allows in-site rental
How to Use
- ComfyUI workflow: P-H3-Image Generate + Edit (AusBoss).json (105 KB); after importing, follow the instructions to add custom nodes and models. This is not a new weight file.
Risks and Limitations
- Dependent nodes and models must be verified by the user; this tool has not been run.
[A] minimax h3 Continuum linear WF with refMODs and pic refs · v1.0a
Workflow · Version Update · Author Jas_A_Hook · Published 11-08 05:05 · Downloads 37
The model-loading boxes supply MiniMax H3 and the supporting parts needed to generate video.
Image
Reasons for Recommendation
- The 2th H3 version of "minimax h3 Continuum linear WF with refMODs and pic refs". This update: +added instructions in post The model-loading boxes supply MiniMax H3 and the supporting parts needed to generate video.
- Within 5 hours of release, it has already reached 37 downloads and 4 likes, with a download rate higher than 88%% of H3 models over the last 7 days
- Includes 1 sample images, allowing you to preview the results before deciding to download
- Documentation is relatively complete (approx. 6582 words)
- Free download, license allows image generation, in-site rental, rental, sale, and sale of fusions
How to Use
- ComfyUI workflows: minimaxH3ContinuumLinear_v10a_3356568.json (108 KB), minimaxH3ContinuumLinear_v10a_3356575.json (108 KB); after importing, follow the instructions to add custom nodes and models. This is not a new weight file.
Risks and Limitations
- 1 sensitive-level examples have been hidden
- Dependent nodes and models must be verified by the user; this tool has not been run.
[A] RefMod baker · v1.1
Workflow · Version Update · Author Jas_A_Hook · Published 11-08 06:21 · Downloads 22
update 1.1 added text2image gen on same WF with swtich ++ Help keep a character on-model” is accurate;
Image
Reasons for Recommendation
- The 2th H3 version of "RefMod baker". This update: How to use this workflow On the group switch, turn on Text2Image. The baking group turns off automatically.
- Within 4 hours of release, it has already reached 22 downloads and 2 likes, with a download rate higher than 85%% of H3 models over the last 7 days
- Includes 1 sample images, allowing you to preview the results before deciding to download
- Documentation is relatively complete (approx. 2482 words)
- Free download, license allows image generation, in-site rental, rental, sale, and sale of fusions
How to Use
- ComfyUI workflow: KREA 2 - Text2image + RefMod Bake on1WF.json (38 KB); after importing, follow the instructions to add custom nodes and models. This is not a new weight file.
Risks and Limitations
- Dependent nodes and models must be verified by the user; this tool has not been run.
Worth Watching (6)
[B] H3 Multi First-to-Last-Frame: Automated, Looping · v1.0
Workflow · New Model · Author qdr1en · Published 11-07 22:01 · Downloads 94
Add prompt instructions for specific runs. The number of frames to generate.
Image
Reasons for Recommendation
- Within 12 hours of release, it has already reached 94 downloads and 8 likes, with a download rate higher than 90%% of H3 models over the last 7 days
- The author provides specific parameters (2 instances), lowering the barrier to reproduction
- Free download, license allows image generation and in-site rental
How to Use
- ComfyUI workflow: h3_fflf_auto.json (97 KB); after importing, follow the instructions to add custom nodes and models. This is not a new weight file.
- Author's parameters: "minimax_h3_fl2v_lightx2v_turbo_4step_v0.1_comfy lora, at 0.75 strength and 8 steps."
- Author's parameters: "It is not essential since the videos are generated at 24 fps, but interpolation helps at …"
Risks and Limitations
- All examples are sensitive-level (7 hidden), so the effect cannot be previewed
- Dependent nodes and models must be verified by the user; this tool has not been run.
- The card image is an AI-generated diagram, not the actual model output; the author has no public non-sensitive examples
[B] [Krea 2 DiT] Clementine (zclclementine) — Fictional AI Influencer & Consistent Character · v1.0 - MiniMax H3 RefMod
LoRA · Version Update · Author MaisonLatentia · Published 11-07 23:46 · Downloads 31
[Krea 2 DiT] Clementine (zclclementine) — Fictional AI Influencer & Consistent Character
Image
Reasons for Recommendation
- The 2th H3 version of "[Krea 2 DiT] Clementine (zclclementine) — Fictional AI Influencer & Consistent Character". This update: ### MiniMax H3 Visual Reference Adapter (RefMod) for Clementine (
zclclementine) Architecture**: Standalone MiniMax H3 Multimodal Reference Adapter (4 Master Visual Anchors) - Within 10 hours of release: 31 downloads, 2 likes
- The author provides specific parameters (3 instances), lowering the barrier to reproduction
- Free download, license allows image generation, in-site rental, rental, sale, and sale of fusions
How to Use
- Load this LoRA in the H3 workflow (2 MB · FP16), include the trigger words "zclclementine" and "zclclementinewoman" in the prompt
- Author parameters: "Recommended Strength**:
0.80 – 0.85(Default:0.85)" - Author parameters: "| LoRA Weight |
0.90(zclclementine_krea2_best.safetensors) | Active in chain…"
Risks and Limitations
- All examples are sensitive-level (1 hidden), so the effect cannot be previewed
- The card image is an AI-generated schematic, not the actual model output; the author has no publicly available non-sensitive examples
Trigger Words: zclclementine, zclclementinewoman
【B】PlagueKind - Minimax H3 Sparse Attention - Refmod - LTX Workflow - Eros | Sulphur Compatible | FaceID · H3 V12
Workflow · Version Update · Author Plaguekind · Published 11-08 09:53 · Downloads 8
help me get a 5090 which is only 0.5% off of int8HQ, makes the same video for 5gb less.
Image
Reasons for Recommendation
- The 4th H3 version of "PlagueKind - Minimax H3 Sparse Attention - Refmod - LTX Workflow - Eros | Sulphur Compatible | FaceID"; this update: which is only 0.5% off of int8HQ, makes the same video for 5gb less. Added Lokken Text encoder, same deal as my model, amazing quality.
- Published 1 hours ago: 8 downloads, 0 likes
- Includes 1 example images, allowing you to preview the effect before deciding to download
- The author provides specific parameters (3 instances), resulting in a low barrier to reproduction
- Free download, license allows image generation, on-site rental, rental, sale, and sale of merges
How to Use
- ComfyUI workflow: PlagueKind-MinimaxH3-V12.json (123 KB); after importing, follow the instructions to add custom nodes and models, as this is not a new weight file itself
- Author parameters: "Recommended 5-12 steps, with default settings."
- Author parameters: "Added my turbo lora, functional strength up to 5.0"
Risks and Limitations
- 2 sensitive-level examples are hidden
- Dependent nodes and models must be verified by yourself; this tool has not been run
【B】Marnie Thatch · v1.0 H3
LoRA · Port · Author MikoGoblin · Published 11-07 23:19 · Downloads 28
She can mend fences, handle animals, maintain equipment, stack a trailer properly, identify what needs doing before somebody asks, and work through the kind of…
Image
Reasons for Recommendation
- Porting an old model to H3: previous versions existed for Krea 2 / ZImageTurbo
- Published 11 hours ago: 28 downloads, 1 likes
- Includes 1 example videos, allowing you to preview the effect before deciding to download
- Description is relatively complete (approx. 2899 characters)
- Free download, license allows on-site rental
How to Use
- Load this LoRA in the H3 workflow (148 MB · BF16), include the trigger word "marnielora" in the prompt
Risks and Limitations
- 1 sensitive-level examples are hidden
Trigger Words: marnielora
【B】Anne-Tayanita Cree - native american woman - video · MiniMax H3
LoRA · New Model · Author Guinwen · Published 11-07 21:56 · Downloads 59
Anne-Tayanita, a Native American woman of the Cree people.
Image
Reasons for Recommendation
- Published 12 hours ago, already has 59 downloads and 7 likes; download speed is higher than 82%% of H3 models from the last 7 days
- Includes 1 example videos, allowing you to preview the effect before deciding to download
- Free download, license allows on-site rental and rental
How to Use
- Load this LoRA in the H3 workflow (233 MB · BF16), include the trigger words "Anne-Tayanita Cree" and "young woman" in the prompt
Risks and Limitations
- 3 sensitive-level examples are hidden
Trigger Words: Anne-Tayanita Cree, young woman
【B】Yuffie Kisaragi Voice (Eng) - Final Fantasy 7 - Audio Only [Minimax H3] (Commission) · Minimax H3 v1.0
LoRA · New Model · Author OneMoreLurker · Published 11-08 07:33 · Downloads 33
This Lora was trained on English language audio clips of Yuffie from the FF7 Remake trilogy.
Image
Reasons for Recommendation
- Published 3 hours ago, already has 33 downloads and 4 likes; download speed is higher than 96%% of H3 models from the last 7 days
- Description is relatively complete (approx. 494 characters)
- Free download, license allows on-site rental, rental, and sale
How to Use
- Load this LoRA in the H3 workflow (142 MB · BF16); the author has not registered trigger words, refer to the example prompts to describe the effect
Risks and Limitations
- All examples are sensitive-level (6 hidden), so the effect cannot be previewed
- No trigger words registered
- The card image is an AI-generated schematic, not the actual model output; the author has no publicly available non-sensitive examples
Data: Civitai public API (base model MiniMax H3), inclusion window 11-07 20:31 → 11-08 10:03 (Beijing Time). Reasons for recommendation are automatic drafts; download counts only represent attention.