Ingi Erlingsson 🪄's banner
Ingi Erlingsson 🪄's profile picture

Ingi Erlingsson 🪄

@ingi_erlingsson9,452 subscribers

Co-founder and CEO of @thesystms. Previously founder @GoldenWolf (acq. 2023) / chief content officer @doodles

Shorts

more glitch exploration with my time slice nodes and Wan 12V + GPT Image 2.0 in ComfyUI ✨

more glitch exploration with my time slice nodes and Wan 12V + GPT Image 2.0 in ComfyUI ✨

99,525 просмотров

By the power of ComfyUI and Wan ⚔️🐅

By the power of ComfyUI and Wan ⚔️🐅

272,594 просмотров

Another test - what I love about this approach is that because there’s no masking involved, the model only replaces what needs to be replaced, so things like the arms and legs in this one stay as they were originally

Another test - what I love about this approach is that because there’s no masking involved, the model only replaces what needs to be replaced, so things like the arms and legs in this one stay as they were originally

51,160 просмотров

Jurassic Bark 🐕 Made in ComfyUI with LTX 2.3 (character swaps), Wan 2.2 Vace (upscaling + refinement) + After Effects (post + comp + sfx) + Suno (music)

Jurassic Bark 🐕 Made in ComfyUI with LTX 2.3 (character swaps), Wan 2.2 Vace (upscaling + refinement) + After Effects (post + comp + sfx) + Suno (music)

17,317 просмотров

I've been working on my first upscaling/detail enhancer LoRA, trained on Flux Klein 9B, which is something I've been wanting to try out for a while. Still needs more training, but so far so good

I've been working on my first upscaling/detail enhancer LoRA, trained on Flux Klein 9B, which is something I've been wanting to try out for a while. Still needs more training, but so far so good

68,502 просмотров

I'm so dope like, yeah We gon' blow like🎈, yeah Made with ComfyUI + Wan Wan Animate + Tongyi Lab Qwen Edit 2511 💥

I'm so dope like, yeah We gon' blow like🎈, yeah Made with ComfyUI + Wan Wan Animate + Tongyi Lab Qwen Edit 2511 💥

138,123 просмотров

I used AI to replace racing grannies with cars

I used AI to replace racing grannies with cars

48,751 просмотров

I've been testing with an unreleased WIP LTX 2.3 LoRA by LDWorks and among other things it enables some of the best performance transfer I've seen so far. It has some ghosting issues, but for a one-shot generation with no masking or pose data it's pretty impressive!

I've been testing with an unreleased WIP LTX 2.3 LoRA by LDWorks and among other things it enables some of the best performance transfer I've seen so far. It has some ghosting issues, but for a one-shot generation with no masking or pose data it's pretty impressive!

32,763 просмотров

Started training my own LTX 2.3 LoRA specifically for character replacement, using it here in combination with D(AI)vid's LoRA to strengthen the effect while avoiding ghosting Nowhere near perfect yet, but given that it's generated in 155s with no controlnets it's not bad

Started training my own LTX 2.3 LoRA specifically for character replacement, using it here in combination with D(AI)vid's LoRA to strengthen the effect while avoiding ghosting Nowhere near perfect yet, but given that it's generated in 155s with no controlnets it's not bad

27,554 просмотров

Round two of gifs to live action with Wan Wan 2.2 + ComfyUI This time exploring whether it can work with traditional cel animation as a base ✏️

Round two of gifs to live action with Wan Wan 2.2 + ComfyUI This time exploring whether it can work with traditional cel animation as a base ✏️

154,572 просмотров

Getting close to a ‘subtle upscale’ version of the LoRA (or 3 LoRAs working together, rather), that stays fairly close to 100% faithful to the identity of the underlying subject - as long as there’s enough pixels to support it

Getting close to a ‘subtle upscale’ version of the LoRA (or 3 LoRAs working together, rather), that stays fairly close to 100% faithful to the identity of the underlying subject - as long as there’s enough pixels to support it

32,852 просмотров

My new favourite thing is using image edit models like 3D shaders, letting you add textures, effects and materials to practically anything We’ve been busy training a whole bunch of them and will be dropping them over on SYSTMS over the coming weeks 🫡

My new favourite thing is using image edit models like 3D shaders, letting you add textures, effects and materials to practically anything We’ve been busy training a whole bunch of them and will be dropping them over on SYSTMS over the coming weeks 🫡

31,958 просмотров

I’ve been obsessively training a Wan 2.2 LoRA for morphing transitions for a couple of months and it feels like I’m getting closer now It’s been deceptively hard training something so conceptual, since you don’t want the model to learn any specific movements or design elements

I’ve been obsessively training a Wan 2.2 LoRA for morphing transitions for a couple of months and it feels like I’m getting closer now It’s been deceptively hard training something so conceptual, since you don’t want the model to learn any specific movements or design elements

31,158 просмотров

I'm working on a video upscaling WF that stays as true to the underlying footage as possible, as I've found that most generative upscalers tend to hallucinate too many details along the way Character likeness still needs work, but given how low res the input was it's not bad

I'm working on a video upscaling WF that stays as true to the underlying footage as possible, as I've found that most generative upscalers tend to hallucinate too many details along the way Character likeness still needs work, but given how low res the input was it's not bad

17,706 просмотров

Videos

Больше нет контента для загрузки