HELENA

AI art • local gen experiments • BHS devlog
NSFW + visual storytelling

[HELENA @VelvetRiftVN]

Been a little quiet lately — work has been keeping me busy.
But I’ve also been spending my free time experimenting with something new. 👀

Early visual test for a future project, currently codenamed BHS.
Nothing final yet. 💛

[HELENA @VelvetRiftVN]

@digo_rossi Baita placa, foi bom
Mesmo!

[HELENA @VelvetRiftVN]

@monospodcast Vc tem que ter hardware forte pro 27b eu uso rocm com 16gb e sofro com o 9b

[HELENA @VelvetRiftVN]

@C_of_Creativity Chat gpt me auxilia a configurar meu comfyui ROCM, e a fazer meus modefiles para Qwen 3 30b a3b, é basicamente minha assistente pessoal que conhece meus caminhos todos, sem ela nunca teria conseguir rodar oque todo em Uma RDNA2 16 gb no windows 😬

[HELENA @VelvetRiftVN]

Testando o MiniMax H3 Eros Max Q5 GGUF localmente no meu setup: RX 6800 16 GB + ROCm/Windows.

Esse take foi gerado com um workflow 2-pass com latent upscale. Ainda é bem pesado nessa GPU e o tempo de geração está longe de ser ideal, mas a estabilidade temporal, expressões e qualidade do motion me impressionaram bastante.

O H3 claramente tem muito mais potencial do que esse teste simples consegue mostrar.

Próximo passo: vou gerar a mesma imagem + mesmo prompt no LTX 2.5 e postar o comparativo de qualidade, motion e tempo de geração. 👀

H3 Eros vs LTX 2.5 — coming soon.

Eros Max v4 Q5_0 | Euler / Simple | 8-step first pass + latent upscale

[HELENA @VelvetRiftVN]

@toyxyz3 LTX 2.5 give me much better results then h3 in rocm, so, since h3 works better on a 3060 12 gb then a Rocm 16gb i will just not work with it, and LTX 2.5 is realy amazing upgrade on 2.3, even the Loras from 2.3 works well

[HELENA @VelvetRiftVN]

NVIDIA buying Hugging Face should be a wake-up call for @AMD.

Imagine AMD partnering with Civitai:

Radeon AI PRO hardware for their generators.
ROCm as a first-class backend.
GPUs sent to ComfyUI / open-source developers.
Day-one support for future high-VRAM Radeon cards.

NVIDIA is investing in its moat.

AMD should invest in the people building outside of it. 🔴

[HELENA @VelvetRiftVN]

@MiaAI_lab As a Rocm user, im realy sad about that

[Clara @aClara4u]

☺️

[Clara @aClara4u]

Clara 💋

[HELENA @VelvetRiftVN]

Wataa

as u waited 🥵, LTX2.5 is amazing

[HELENA @VelvetRiftVN]

would u want to see this animated? 👀

[Clara @aClara4u]

💋🤍

[Clara @aClara4u]

💋😘🍓

[Clara @aClara4u]

Blue suits me well? 👉👈💙🫐

[HELENA @VelvetRiftVN]

@gatogatosogordo U must be, future updates soon 🥰🥵

[HELENA @VelvetRiftVN]

Been a little quiet lately — work has been keeping me busy.
But I’ve also been spending my free time experimenting with something new. 👀

Early visual test for a future project, currently codenamed BHS.
Nothing final yet. 💛

[Clara @aClara4u]

tell me, wich one is mine? 🥹

[Clara @aClara4u]

out for a drink… behave 👀

[Clara @aClara4u]

do i look good on black? 🖤😳