# Krea 2 - Wulver
Full fine-tune of Krea 2 Raw (12.8B DiT) — anthro & furry, with strong anime / kemono
range and true multi-character composition. Trained on 1.5M+ curated images with
multi-format captions.
> In Shetland folklore, the Wulver is a wolf-headed being that was never cruel —
> it fished the lochs and left its catch on the windowsills of those in need.
This is a v0.1 *beta** — expect rough edges. An aesthetic pass is in active development
and updates will roll out as the base training matures.*
## What it does
- Anthro / furry as its home turf — species knowledge far beyond generalist models
- Anime & kemono styles — not the usual western-model anime approximation
- Multi-character scenes — characters interacting, not merged into one blob
- Artist styles via @artistname prefixes (Curently under development)
- Best results with detailed descriptive prose — see prompting tips below
## Settings
| | |
|---|---|
| Steps | 14 |
| CFG | 1.5 |
| Sampler / Scheduler | euler / simple |
| Resolution | 1024 native |
| CLIPLoader type | krea2 |
## ComfyUI setup
1. Wulver_v0.1_fp8.safetensors → ComfyUI/models/diffusion_models/
2. Text encoder qwen3vl_4b_bf16.safetensors from [Comfy-Org/Krea-2](https://huggingface.co/Comfy-Org/Krea-2) → models/text_encoders/
3. VAE qwen_image_vae.safetensors from the same repo → models/vae/
4. Drag the attached workflow into ComfyUI and hit queue.
## Prompting tips
This model was trained on long natural-language captions, so detailed prose gets you
the most out of it. Template that works:
> *A digital painting of [subject & species, appearance details]. [What they are doing,
> pose, expression]. [Clothing / body details]. The background features [setting,
> lighting, atmosphere]. [Artistic style / medium notes].*
- 60–120 words is the sweet spot. Short prompts work, detail works better
- @artistname, at the very start for artist styles
- Keep CFG at 1.0 — this is a guidance-distilled blend; higher CFG burns the image
- Copy the prompts from the sample gallery to see the dialect in action
## Compatibility notes
- LoRAs trained on base Krea 2 may behave differently on Wulver (weights have moved)
- bf16 version available on request
## License
This is a modified version of the Krea 2 model by Krea.
Krea 2 is licensed under the [Krea 2 Community License Agreement](https://krea.ai/krea-2-licensing) —
by using this model you agree to its terms (including [Acceptable Use Policy])(https://www.krea.ai/krea-2-use-policy).
Not affiliated with or endorsed by Krea.
---
by Vaelico — generate online shortly at [vaelico.ai](https://vaelico.ai)
Description
# Wulver v0.1 — Non-Turbo BF16
Original Wulver v0.1 full DiT checkpoint before the Turbo adapter was merged, intended as a base for LoRA training and further fine-tuning.
- Architecture: Krea 2 RAW, 12.8B DiT.
- Precision: BF16, 25.64 GB.
- Full model weights, not a LoRA.
- Preserved sft_crema_epoch2 checkpoint, renamed without altering its bytes.
- Text encoder and VAE are separate: Qwen3-VL-4B and Qwen Image VAE, as used by Wulver.
- Requires a trainer with explicit Krea 2 support. Compatibility with a particular third-party trainer has not been newly tested for this upload.
This is the Non-Turbo counterpart of the existing v0.1 release, not a newly trained Wulver version. The Turbo release's 14-step / low-CFG recommendations do not apply automatically.
File: Wulver_v0.1_non_turbo_bf16.safetensors
SHA256: 82d4f30e2477fe7532ecbbe9396bff284110a72b96694fdaa44e1554081c2514
Hugging Face: https://huggingface.co/Vaelico/Wulver
Modified Krea 2 model. Same Krea 2 Community License, Acceptable Use Policy, and attribution notice as the existing Wulver release; see LICENSE.md and NOTICE in the Hugging Face repository. Not affiliated with or endorsed by Krea.
FAQ
Comments (8)
Any info on the difference between v0.1 and base 0.1v? Are they just difference quants?
Base is the raw version, is mainly for traning and trying things out. For normal generation the v0.1 is recomended.
It's very good so far! I'm loving it! Keep it up! I can't wait for the final release!
Really nice work! Definitely getting more furry concepts than the baseline. Struggling a bit with related ideas bleeding together, but that's to be expected.
I'm no expert at training, but is there a chance that expanding your prompt dialect by adding a tag block from the source booru to the end of your natural language in the training files might help it learn some of the more subtle concepts (like the sizes of things, species, etc.) while keeping the human part of the workload down?
Hi, yes, i will add tags this time too so it is easier to learn some things. And thanks!
@VaelicoAi Really excited to see how that turns out; I'd hope it would help it absorb the furry-specific concepts even faster. This tune has already been pretty magical compared to the (already impressive) base Krea2 for the concepts that took.
Would it be possible to get an int8 convrot version? it runs way better on 3000 series GPUs.
Hi, isn't it already up?
Wulver_v0.1_int8_convrot.safetensors