ComfyUI node explorer • Sharing AI workflows • Diffusing pixels and conditioning latent space 🫡
San Francisco, CA
- Tweets 1,020
- Following 548
- Followers 6,363
- Likes 5,106
The best open source 3D AI generation model now runs natively in ComfyUI. The workflow runs 100% locally, on just 6GB of VRAM.
No more complex setup or dependency errors. TRELLIS.2 and Pixal3D work right out of the box, with one click.
Not to mention, it's FULLY COMMERCIAL-SAFE. The team re-implemented and removed every strictly-licensed library (thanks to the legend Kijai).
Drop in an image and click run, it's that easy to start. Go play with it.
Free workflow below!
40
136
1,280
74,448
Big day for open source.
MiniMax Music 3 is undeniably a state of the art open weights Music Generation model and a real alternative to Suno.
My favorite part about open weights: the best is yet to come. Once the community starts playing with the model and training LoRAs, this model only gets better and allows for more control.
🎵MiniMax-Music3
Next-Generation Open-Weights Production-Ready & Versatile Music Model
huggingface.co/MiniMaxAI/Min…
82
232
2,908
323,045
MiniMax H3 is the SOTA open weight model... but the large model comes with slow gens (depending on your hardware). You can be waiting for minutes just for the generation to miss the mark
Here's a ComfyUI workflow to get the most out of H3 without the wait.
→ H3 Turbo LoRA (4–8 steps)
→ Model Preview Override + Tiny AutoEncoder H3 for a fast composition preview
→ Bad preview? stop it, tweak the prompt or seed
→ Save your precious gpu time
Workflow and model links below 👇
16
36
462
27,734
It can run on a toaster.
Total memory footprint reduced by 66%. Live in ComfyUI!
MiniMax-H3 Is Now Publicly Available
huggingface.co/MiniMaxAI/Min…
5
2
68
4,683
Get your caffeine and GPUs ready.
Open source video models are taking a huge step forward tomorrow, thanks to Minimax H3. Truly a game changer for local users.
The cracked team at ComfyUI has optimized this huge model to the point where it can run comfortably on a toaster.
This generation was one shot and came out exactly how I imagined it. Used two reference images and an input audio that generated identically.
I can't wait to see what the community does with the open weights.
5
6
81
7,334
Open source video models are taking a huge step forward tomorrow, thanks to Minimax H3. Truly a game changer for local users.
The cracked team at ComfyUI has optimized this huge model to the point where it can run comfortably on a toaster.
This generation was one shot and came out exactly how I imagined it. Used two reference images and an input audio that generated identically.
I can't wait to see what the community does with the open weights.
23
14
307
30,363
Hailou MiniMax H3 vs. Seedance 2.0 with 9 Reference images.
Only one model used all 9 images... and only one model generated a better output at 1/3 the cost...
8
6
134
13,814
Short form algorithms better get ready for FLUX 3.
"A timelapse video of a man in rugged worn clothing in a jungle clearing building a primitive stone portal from clay bricks and river metal, working with only a machete and bare hands. Full timelapse from digging the pit to the final stone — at which point a swirling purple portal ignites inside the ring, glowing and rippling, and he steps through it and vanishes. Continuous shot with no cuts, no music, no narration. Jungle sounds only, with a low hum as the portal opens."
6
4
60
4,697
Loving the native audio from this model
Introducing FLUX 3.
One multi-modal model for Image, Video, Audio and Action-Prediction. Creations are truer to life in every kind of style.
FLUX 3 Video is now available in early access (link below).
Jointly trained in one unified architecture, our model can be extended to predict actions for robotics. See our work with mimic and Audi in the thread.
3
5
99
7,312
Only one image model can pull off these one shot generations...
Identity transfer + 2 style LoRAs + perfect prompt adherence
And it's open source
2
1
31
3,365
The best part about open source is how it compounds.
Raw image model weights get open sourced → someone trains an outpainting LoRA on top → someone else wraps it for ComfyUI → I use it.
And the result is an outpainting workflow with Krea 2 that is super high quality + fast.
9
11
108
7,227
Here's how to get 100% consistent product ads from one seedance 2.0 generation.
I did it all in a single chat using the Comfy MCP. The real control here comes from calling my existing workflows (shared below) instead of the agent improvising a pipeline.
I directed the agent to call my ComfyUI workflow for cinematic product ads. I specified the close-up shot of the sprite animating, the bezel turn flipping the screen, the display changing to the time 10:04.
Now for the consistent variations. The driving video does the heavy lifting but you need to get it right
→ depthanything v3 pass blended with canny edge lines to show the fine detail... it's why the tiny debossed logo is there
→ the initial sprite outline lived in those edge lines too, and every gen kept inheriting it. claude suggested a sam3 mask over the screen to hide it (s/o the agent)
→ with the screen masked, the new star sprite is just prompting: one gpt-image-2 still to generate a reference, one extra line in the seedance 2.0 prompt, and it animates
oh and the whole process is a claude skill now.
16
27
322
20,885
Claude Fable running ComfyUI workflows through the Comfy MCP is a cheat code.
→ Claude pulled the shot from the web, auto-detected the scene cuts + trimmed it
→ ran my saved depthanything v3 + openpose workflow to build the driving reference video
→ scrubbed every frame for the sharpest one, then character swapped me in with gpt-image-2 exactly how I directed it: wearing a tuxedo, liminal space environment, no background characters
→ wrote the seedance 2.0 prompts + ran the gens
→ stitched together the 3-panel comparison you're watching
The preprocessing + prompt writing is usually most of the work on a clip like this. With the MCP I didn't touch any of it. Didn't need to look at a single node lol.
30
92
628
31,227
if you need a precise style AND a model that follows your prompt... don't use Midjourney
Krea 2 trained a style I couldn't get into any other model. Midjourney included. It's about time an open source model got there.
I've been sitting on this dataset forever. It's of full-page plates from an early-1900s book of watercolor illustrations (via Public Domain Review). The texture and faded tonal gradients with soft bleeding edges tripped up every model I tried... none could get the aesthetic right with decent prompt adherence.
Krea 2 nailed it.
14
17
636
47,056
Krea 2 trained a style I couldn't get into any other model. Midjourney included. It's about time an open source model got there.
I've been sitting on this dataset forever. It's of full-page plates from an early-1900s book of watercolor illustrations (via Public Domain Review). The texture and faded tonal gradients with soft bleeding edges tripped up every model I tried... none could get the aesthetic right with decent prompt adherence.
Krea 2 nailed it.
31
34
635
69,460
S/o @0xInk_ for the reference character, I used gpt-image-2 to add some realism + background. Def give him a follow for the best AI character design out there!
2
11
1,507
Testing SCAIL-2, an open source model for motion transfer. I wanted to see how it can handle driving videos with fast, dynamic movement.
The part that really impressed me was how the model retained details from the input character's outfit (especially the straps on the shorts). I attached the input character below for you to compare against.
The model also has a replacement mode that swaps the character into the driving video's scene, but here I used animation mode, which keeps the reference image's scene instead.
18
55
474
29,234
Exploring tools with @ComfyUI
Ideogram V4 for image of Messi → ComfyCloud MCP to generate the other players → MCP + Grok Imagine 1.5 i2v → @ltx_model 2.3 + @thesystms FLW LoRA for vid transitions → Speed ramp in Davinci → TimeSlice node → Suno for audio
4
7
71
3,906
The only thing worse than looking at a blank canvas is prompting with JSON.
Here's a ComfyUI workflow to fix that.
Upload your image and an LLM automatically creates bounding boxes + structured JSON for Ideogram V4. The 'Prompt Builder' node draws the bboxes and scaffolds the prompt. From there you just refine and tweak. Change the prompts, bbox positions, color palettes — then generate and iterate.
Prompt below ⬇️
10
19
254
11,663
Controlling layouts using bounding boxes with Ideogram V4 opens a completely new paradigm for image generation.
→ Tweaked the bbox layouts and refined the prompt in ComfyUI using 'Ideogram 4 Prompt Builder' node
→ Brought the structured JSON prompt into Claude Code
→ ComfyCloud MCP to generate variations with new subjects, colors (hex values) and descriptions
→ Layout held to the exact pixel across all of them
Zoom in to check out the accuracy of the bboxes and text rendering (examples are one shot btw)
3
5
40
2,336
The best way to bring the composition from your head into an image → Ideogram V4 + drawing bounding boxes in Comfy.
The control here is quite unique. The model uses structured JSON so drawing bounding boxes to get the exact placement works very well.
The model only needs 12 steps (turbo), so iterating with different seeds + very impressive text rendering capability leads me to say this is a state of the art open source image model right now.
Using the 'Ideogram 4 Prompt Builder' node by Kijai.
11
37
312
26,588
Testing VOID, Netflix's inpainting/object removal model.
For these POV shots, the real test was getting accurate masks with SAM3. With very simple prompts VOID handled the removal very well.
Using the default workflow on ComfyUI, 5 second video is ~110 seconds on an RTX PRO 6000.
1
9
1,521
Comparison of Omni, Seedance 2 and LTX 2.3 at video outpainting.
Surprisingly, Omni failed at this task (I tried a variety of reference videos and prompts). Might work better with realism…
Unsurprisingly, Seedance 2.0 nailed it and LTX did incredibly well at a fraction the cost.
Formal challenge to get video outpainting to consistently work with Omni
7
1
44
3,585
Search up "Pyramids of Egypt" and you'll see just how impressive the world knowledge is...
> a recording from a the back of a Camel in the outskirts of Cairo, a jerky zoom into something in the distance and then refocusing (with a bit of back and forth) (no timestamp or dialog)
Gemini Omni Flash:
> a recording from a capsule on the london eye, a jerky zoom into something in the distance and then refocusing (with a bit of back and forth) (no timestamp or dialog)
Note the world knowledge of London’s landscape, and the way the video is gently moving like the capsules do.
1
7
1,048
Wrapped this technique into a simple ComfyUI workflow.
Upload a video + character image and watch the nodes work their magic.
You can also easily prompt for variation - in the vid below I prompted "extra emphasis on the rubberhose animated movement"
Below I will teach you how to reverse engineer any 15s video you see
You will need to tweak it a bit but I will explain in this thread exactly how to make these if you are ever curious
Thread below 👇
4
2
43
7,981
Save your credits and use open source models when you can.
The new LTX 2.3 lipdub LoRA paired with Chatterbox TTS voice cloning model is the best workflow for lip syncing, change my mind.
19
24
340
39,745
Here's a SOTA Seedance 2.0 feature which I haven't seen anyone show off...
Seamlessly extend videos using this ComfyUI workflow
→ Upload driving video
→ Prompt the next shots camera movement and scene action
→ Select the frame that most closely matches driving videos end frame
→ Workflow stitches audio and video
Link to the workflow below!
4
5
55
6,650
Combine the right open source models with Seedance 2.0 and you can have so much control over your generations... the face accuracy and detail is insane
Here’s the process
→ extract video first frame + gpt-images-2 to head swap
→ sapiens2 + depthanything3 to create control reference videos
→ seedance 2.0 real human comfyui workflow (to bypass realistic human guardrails)
→ upload seedance gen into ltx2.3 HDR lora workflow
→ color grade the video in comfy or davinci resolve
will drop a guide with exact workflows if there's enough interest
11
11
122
11,604
George Costanza is a gigachad for this take
The best face swap method is also open source
I trained this @ltx_model LTX 2.3 LoRA of George Costanza at home on my 5090 in about a day with AI Toolkit. I generated this 30 second video with @ComfyUI on my 5090 in 6 minutes. Open source is, always has been, and always will be, the future of generative AI. (SOUND ON)
13
24
391
42,756
Trajectories! Wan ATI is such a fun model to play around with
Quick demo to show how easy it is to draw your own paths - template is live in Comfy!
these are trajectories!
just draw paths to steer video generation, and it looks like bending reality.
new project with @nvidia 💚
3
9
90
8,638
Kling 3.0's Multi Shot feature is underrated.
I built this ComfyUI workflow to have balance between creative control & automation.
Here's how it works:
Input images of product/character → Select total duration and # of scenes → LLM gens all prompts and timing → You refine manually → Generate multi scene video
Workflow attached 👇
13
22
255
15,700
Nodes have always been a huge hurdle for two groups: non-technical creatives wanting to try ComfyUI, and builders with complex workflows they can't easily hand off.
App Mode fixes both. Now you get the full power of node-based workflows without looking at a single node.
Here's a quick demo to show you how easy it is to go from nodes → app
Two massive updates for the ComfyUI ecosystem today:
1️⃣ App Mode: The power of the node graph, now behind an easy-to-use interface. Turn complex workflows into custom apps.
2️⃣ ComfyHub: A brand new home to discover, run, and share community workflows and apps instantly via URL. Try ComfyHub preview via links.comfy.org/4dke0ki
Create in App Mode. Share on ComfyHub.
Learn more here: links.comfy.org/4bAOjuz
22
23
391
41,535
Now YOU are the prompt, lora and controlnet
Hard to think this would’ve been possible 2 years ago, Nano Banana 2 is impressive
And now we wait for opensource to catch up (it will)
8
8
118
10,602
You can push Seedance 2.0 in some very trippy ways...
Can't wait to pipe together some workflows with it in Comfy
11
8
192
12,426
Yeah and because the visuals are already audio reactive you don't even need audio conditioning
Below is an img and audio input + prompting (no depth/canny info), it takes ~60secs for 50seconds of video at 480p on an rtx6000 (the switch up at second 27 goes hard lol)
4
5
549
Here's the start frame and prompt in alt for the legend that tests this <3
ALT The woman (softly in a concerned tone), "Hey legend! Are you sure that's legal?" The man responds, he has a sly smile, (in a cool and confident deep southern voice), "These undisclosed paid ads will never see the light of day." He closes the trunk, pitch black.
1
4
3,425
Kling 3.0 performs really well with simple prompts.
Can someone please test this generation using Seedance 2.0 for me??
2
25
63,188
The most entertaining outcome is the most likely
You thought the fun was over? 🏈
This weekend, video takes center stage on the timeline.
We’re awarding $1M, $500K, and $250K to the top three videos about @grok, created with Imagine 1.0.
11
2,409
The quality, cost, and control you can achieve for upscaling + fixing plastic AI skin with open source models still amazes me...
Models used:
→ Z-image-turbo for image gen (~3s)
→ SDXL + Lora for skin texture (~15s)
→ SeedVR2 for upscaling (~40s)
27
58
690
42,025
Great prompt, but fine details won't survive the grid.
Nano Banana Pro generates each in ~1k resolution.
Product text gets lost.
This workflow fixes that:
→ Automatically extracts all 9 grid image
→ You select your favorite
→ Upscales to 4K using your input as reference
8
33
517
84,278
So many ways you can modify the contact sheet workflow, the possibilities are endless (and it's fun)
I rebuilt my original contact sheet workflow for fashion style videos end to end.
Just drop in any character and outfit and it generates the full suite of assets. Workflow and links all right here.
4
6
95
12,070
By far the best way to generate images at different angles is this 3D Camera Control node.
Instead of describing angles with words, you just click the view you want. Simple UI, and the Qwen Image Edit 2511 multi-angle LoRA keeps things consistent across generations.
Workflow below 👇
4
26
210
61,492
There are still many undiscovered use cases for Nano Banana Pro.
Here's a fun one.
Upload your profile picture → Prompt for your theme → Generate 8x8 grid → Select favorite → Upscale to restore details
64 variations from 1 prompt. Consistency holds well across the entire grid.
+ there are definitely more use cases for this workflow...
Give it a try! 👇
5
13
2,413
Ash Ketchum for the Pokémon x North Face winter drip collection.
These fashion contact sheet videos are ridiculously easy to make with the following workflow.
All you need is
→ character image
→ outfit image
→ a bit of manual prompting
→ a good idea
Full workflow 👇
Pokémon × The North Face Custom Jacket 🔥⚡ A one-of-a-kind TNF Nuptse Retro 700 customized by LISE LAB ❄️
Level up your winter drip — ready to flex?
#Pokemon #TheNorthFace
21
100
1,399
250,719
This workflow works very well for character consistency in AI UGC
→ Upload character + product input image
→ Prompt each grid position individually (e.g. "Position 1,1: Close up of the main subject drinking from the can")
→ Automatically saves each grid image individually
→ Select desired grid image
→ Additional Nano Banana Pro generation for adding back product details to selected image
Prompt in alt and link to try the workflow below!
ALT Prompt: Create a 3×3 grid for a 1980's commercial aesthetic ad featuring only the following product: can of diet pepsi. Generate as one 3×3 grid with no margin, gutter or borders. Every grid image should follow the aesthetic, colors, lighting and color grading of input image 1. Follow the prompts for each grid image. If left blank use your creativity: Position 1,1: Close up of the main subject drinking from the can Position 1,2: Product pouring out liquid onto the black boxing Position 1,3: Extreme close up of the main subject holding the product close to her face, fierce expression Position 2,1: The subject is in a dynamic boxing pose with the can crushed between her teeth, medium shot Position 2,2: Action shot of the Can product being hit by the boxing glove. Position 2,3: Shot of the subject drinking from the can, sitting in on a stool in the corner of the ring Position 3,1: Product in sharp focus resting on a ring rope, the subject is out of focus standing in the ring b
One prompt to create an amazing Ad campaign, use this prompt with any product you have and share your results
Prompt:
Create a 3×3 grid in
3:4 aspect ratio for a high-end commercial marketing campaign using the uploaded product as the central subject.
Each frame must present a distinct visual concept while maintaining perfect product consistency across all nine images.
Grid Concepts (one per cell):
1. Iconic hero still life with bold composition
2. Extreme macro detail highlighting material, surface, or texture
3. Dynamic liquid or particle interaction surrounding the product
4. Minimal sculptural arrangement with abstract forms
5. Floating elements composition suggesting lightness and innovation
6. Sensory close-up emphasizing tactility and realism
7. Color-driven conceptual scene inspired by the product palette
8. Ingredient or component abstraction (non-literal, symbolic)
9. Surreal yet elegant fusion scene combining realism and imagination
Visual Rules:
Product must remain 100% accurate in shape, proportions, label, typography, color, and branding
No distortion, deformation, or redesign of the product
Clean separation between product and background
Lighting & Style:
Soft, controlled studio lighting
Subtle highlights, realistic shadows
High dynamic range, ultra-sharp focus
Editorial luxury advertising aesthetic
Premium sensory marketing look
Overall Feel:
Modern, refined, visually cohesive
High-end commercial campaign
Designed for brand websites, social grids, and digital billboards
Hyperreal, cinematic, polished, and aspirational
8
26
370
34,596
The biggest downside to Nano Banana Pro is the cost ($0.25/image) and slow generation speed.
Here's a workflow that addresses both:
1 prompt = 9 distinct images at 1K resolution (~3 cents per image)
The key is prompting each grid position individually, so you can test different lighting, backgrounds, or angles simultaneously. Iterate on different looks quickly and pick your winner.
The workflow:
→ Extracts all 9 images
→ You select your favorite
→ Upscales it to 4K while referencing your input image to maintain consistency + details
43
60
1,006
107,737
Scroll stopping UGC content is now only limited by your creativity
This workflow generates 4 separate images with 1 prompt - cutting the cost of Nano Banana Pro by 4x
+ you can chain the outputs together for a longer sequence
slideshow prompts to go viral:
{
"subject": {
"description": "A young woman taking a mirror selfie, playfully biting the straw of an iced green drink",
"mirror_rules": "ignore mirror physics for text on clothing, display text forward and legible to viewer, no extra characters",
"age": "young adult",
"expression": "playful, nose scrunched, biting straw",
"hair": {
"color": "brown",
"style": "long straight hair falling over shoulders"
},
"clothing": {
"top": {
"type": "ribbed knit cami top",
"color": "white",
"details": "cropped fit, thin straps, small dainty bow at neckline"
},
"bottom": {
"type": "denim jeans",
"color": "light wash blue",
"details": "relaxed fit, visible button fly"
}
},
"face": {
"preserve_original": true,
"makeup": "natural sunkissed look, glowing skin, nude glossy lips"
}
},
"accessories": {
"headwear": {
"type": "olive green baseball cap",
"details": "white NY logo embroidery, silver over-ear headphones worn over the cap"
},
"jewelry": {
"earrings": "large gold hoop earrings",
"necklace": "thin gold chain with cross pendant",
"wrist": "gold bangles and bracelets mixed",
"rings": "multiple gold rings"
},
"device": {
"type": "smartphone",
"details": "white case with pink floral pattern"
},
"prop": {
"type": "iced beverage",
"details": "plastic cup with iced matcha latte and green straw"
}
},
"photography": {
"camera_style": "smartphone mirror selfie aesthetic",
"angle": "eye-level mirror reflection",
"shot_type": "waist-up composition, subject positioned on the right side of the frame",
“aspect_ratio”: “9:16 vertical”,
"texture": "sharp focus, natural indoor lighting, social media realism, clean details"
},
"background": {
"setting": "bright casual bedroom",
"wall_color": "plain white",
"elements": [
"bed with white textured duvet",
"black woven shoulder bag lying on bed",
"leopard print throw pillow",
"distressed white vintage nightstand",
"modern bedside lamp with white shade"
],
"atmosphere": "casual lifestyle, cozy, spontaneous",
"lighting": "soft natural daylight"
}
}
Community note
This is a reverse prompt of a real person's image using nano banana pro. This person likely did not give approval for her likeness to be cloned.
The result is impressive due to it being based on a real person and not text to image prompt.
x.com/BryceDearden/s
14
54
1,015
144,504
Made this Chrome Extension for all the boomers to help prevent them from getting one shotted.
All they need to know is how to right click and press "IS THIS IMAGE REAL".
Be a hero this Thanksgiving.
78
303
3,958
652,535
We're cooked.
It's impossible to distinguish between what is AI and what is real anymore.
87
63
965
147,157
Nano Banana Pro can take image references of graphic heavy ads + your product and seamlessly integrate the product AND change the copy. Reference images attached below.
(TIP: Generate in 4k if ad has lots of elements for best results)
Prompt:
Objective: Create a composite commercial image by synthesizing layout inputs with subject inputs. Target Product: {insert product description/copy here}
1. Inputs
Reference Input 1 (Style & Layout): Acts as the strict structural blueprint. This dictates composition, graphic elements, lighting direction, and text layout.
Reference Input 2 (Subject): Provides the primary subject matter (product shape, texture, branding).
2. Execution Instructions
Phase A: Scene & Design Analysis
Environmental Scan: Analyze the perspective, depth, and lighting environment of [Input 1].
DNA Deconstruction: Identify font styles, graphic shapes, and specific noise/grain levels.
Phase B: Object Replacement
Removal: Remove the main product currently featured in [Input 1].
Insertion: Insert the product isolated from [Input 2] into the exact focal point and scale defined by the original layout.
Phase C: Text Modification (Content & Style)
Condition Check: Analyze [Input 1] for existing Copy text elements (excluding text physically on the product).
If Copy Text Exists: Replace the existing copy with text relevant to the target product. You must replicate the exact font, weight, kerning, and drop-shadows of the original.
If No Copy Text Exists: Strictly do not generate any text. Preserve the negative space exactly as it appears in [Input 1].
Phase D: Harmonize & Finish
Lighting: Adjust lighting, cast shadows, and reflections on the inserted product to match the environment.
Texture: Apply the original image's texture (film grain, halftone, print artifacts) to the new elements to ensure a cohesive "print ad" look.
3. Final Output
Aspect Ratio: Must match Input 1.
Quality: A seamless, professional advertisement where the design layout is preserved, but the subject and color story are synthesized according to the user's preference.
5
8
50
4,793
Nano Banana Pro opens up so many new creative workflows due to it’s world knowledge and native 4K resolution images
Not only can you generate 4 images at once (and save $$$), but your prompt can “time travel” each image with character and scene consistency
I built a FREE workflow that utilizes the model’s SOTA capabilities
→ Upload reference image
→ Prompt includes “each frame should reference how the input image looks 1,2,3,4 seconds after the {sequence}”
→ Generate a 2x2 grid in 4k
→ Select and automatically extract the frame (supports any aspect ratio)
→ Use as input for the next sequence
→ Repeat with new sequence
Check the workflow and demo below 👇
ALT The sequence is a capybara taking a selfie from different camera angles while holding a can of redbull, it showcases the redbull and takes a sip while showing off the canyon scenery.
ALT The sequence is 4 action shots from different camera angles of the capybara as it base jumps off a cliff and flies through a canyon valley.
ALT The sequence is 4 action shots from different camera angles of the capybara as it base jumps off a cliff and flies through a canyon valley.
ALT The sequence is a capybara landing from a base jump, extremely happy and cheering as it lands on the ground.
9
4
110
9,339
→ Generate base image using Qwen in Comfy
→ Create variations of scene with Multi Angle Lora
→ First and Last frame video gen with Veo
→ Editing tool to cut extra frames (Veo adds frames at the end)
→ Suno v5 for music
All in a couple hours work.
Another win for open-source AI models.
This new Qwen Multiple Angles LoRA perfects scene and subject consistency.
Look at her ring, the records, album covers and turntable buttons. Even the headphone wire and text. Insane.
Test it for yourself - FREE workflow ⬇️
40
54
731
72,648
Out of box results for this image were not great but with additional prompting it turned out well
“Change the picture 1 to a surreal absurd photograph, she has ginger hair, olive skin, long eyelashes and an open mouth smile. She has perfect hands with five fingers and Show more
1
1
1,368
First things first. The workflow and model link are attached for you to test it yourself.
The model is in alpha so it’s not perfect. The most common issue is the output lighting is darker. However, Qwen has very strong prompt adherence and prompting for hair, eye and other colors in the scene help avoid this. Faces tend to lean towards asian, but once again prompting for “caucasian” can help.
If you found this workflow helpful, follow @hellorob for more free AI workflows.
workflow - pastebin.com/HPFRNyxk
model - huggingface.co/lrzjason/Qwen…
1
1
267
The best AI tools are increasingly open-source.
Here’s a new one click solution for what used to require stacking multiple models.
Qwen “Anything2Real” LoRA dropped and it transforms ANY style into photorealistic images.
FREE workflow, more examples and my learnings 👇
2
2
31
3,309
First things first. The workflow and model link are attached for you to test it yourself.
The model is in alpha so it’s not perfect. The most common issue is the output lighting is darker. However, Qwen has very strong prompt adherence and prompting for hair, eye and other colors in the scene help avoid this. Faces tend to lean towards asian, but once again prompting for “caucasian” can help.
If you found this workflow helpful, follow @hellorob for more free AI workflows.
workflow - pastebin.com/HPFRNyxk
model - huggingface.co/lrzjason/Qwen…
4
1,154
Here's what the simple workflow looks like once loaded in ComfyUI.
Not much too it, but worth noting I prefer using an 'Upscale Factor' instead of calculating the shortest edge in pixels. Included is a Note for some tips. I definitely recommend watching the authors video on setting up your settings, it's very easy to follow and in-depth.
Follow @hellorob if you found this helpful and for more easy-to-use AI workflows!
workflow - pastebin.com/MGDc4tYF
tutorial - youtube.com/watch?v=MBtWYXq_…
2
27
1,945
Upscaling has never been so easy.
SeedVR2 got a new update and the speed + quality + control is now the best tool on the market.
Control both high frequency and low frequency details AND great logs to help get the fastest speed on any machine.
FREE workflow + tutorial link ⬇️
7
10
104
8,005
