fofr
@fofrAI
Followers
21,553
Following
913
Media
1,866
Statuses
5,017
Explore trending content on Musk Viewer
Dortmund
• 468624 Tweets
Palestino
• 99600 Tweets
iPad Pro
• 98395 Tweets
Grok
• 83621 Tweets
FOURTH BE MINE 10M🩵
• 81832 Tweets
People In Support
• 75814 Tweets
YAIBA
• 67533 Tweets
Rangers
• 46864 Tweets
ノーリーズン
• 30917 Tweets
Demichelis
• 27614 Tweets
Shai
• 25917 Tweets
backnumber
• 23350 Tweets
Canes
• 18470 Tweets
#SoloLasMas
• 18374 Tweets
ゼウスシルエット
• 16702 Tweets
プレス機
• 14937 Tweets
Chará
• 13917 Tweets
Chet
• 13365 Tweets
ムラサメ改
• 12256 Tweets
リュウケンドー
• 11387 Tweets
Pinned Tweet
A new experiment mixing regional prompting and controlnet to create some smooth non-flickering animations.
If folks are interested I'll put up a workflow video on YouTube.
64
178
2K
I asked Midjourney v5 to '/describe' some logos, to see how it would create prompts for them, and to see what it would create in response.
Starbucks
108
518
6K
🧵 A big
#Midjourney
thread on how to write prompts to get good cinematic images.
In this thread I’ll build up a single prompt with cinematic elements, and show their effects.
Each prompt will use a 16:9 aspect ratio, and to minimise variation I've locked in a seed.
98
439
3K
With
#stablediffusion
img2img, I can help bring my 4yr old’s sketches to life.
Baby and daddy ice cream robot monsters having a fun day at the beach. 😍
#AiArtwork
41
270
2K
The live action Game of Thrones Mario Kart adaptation is wild.
Winter is coming... so are the spiny shells.
42
170
2K
You can also try out to color vectorize an image. Thanks for the heads up
@KireiStudios
.
1. Use Midjourney with a prompt like: "an svg vector image of [subject], white background, logo"
2. upload to
3. get svg, eps, pdf …
🙌😍
20
200
1K
I trained an SDXL fine-tune on really bad early digital photos with poor use of flash and crappy lighting.
50
118
1K
Stable Video Diffusion (SVD) is now on Replicate:
- Make 14 frame or 25 frame videos
- Use any image, it'll be resized to make the best video
- Control degree of movement
Great work putting this together
@lucataco93
, and
@StabilityAI
for the fantastic…
26
141
922
😍 Woah, CoDeF (temporally consistent video processing) looks really powerful! Consistency is fantastic.
Transform a derived canonical image using controlnet then apply to a whole video.
Project:
Code:
Some fave examples in 🧵
19
224
847
Dang, I just can't get over how good this Stable Video Diffusion model is.
39
80
849
face + style = become image
I've now pushed "become-image" as a Replicate model and open sourced the code and ComfyUI workflow on Github. All links in the thread below.
face + style – using IPAdapter for style, InstantID for consistency, prompt for steering and depth controlnet for continuity with original.
This also uses DreamshaperXL Lightning, so these images are in just 4 steps.
Will share details soon.
15
52
489
17
99
729
🧵 Announcing Prompter, an open-source tool (and npm package) for generating and sharing prompts:
Use it with Midjourney, Stable Diffusion, Gen2, anything. It's great for quickly exploring models.
I've shared some of my fave prompts in the thread 👇
28
152
731
👀 face-to-sticker model is now live.
Run it on Replicate.
It's a ComfyUI workflow too.
Details below.
21
81
662
First text-to-video render of Stable Video Diffusion (SVD) from a Midjourney input image.
I'm impressed with
- coherent movement
- video quality
- accuracy with original image
Shame the explosions didn't, uh, explode.
24
62
606
I asked Bard to make some image prompts for MJv5.
Prompts are in the Alt tags.
I iterated and asked:
- Embellish these with dramatic details, using only keywords
- Include the city and year and some cinematic effects
21
66
578
🧵A follow-up thread on how to write prompts to get good animated cinematic images.
🧵 A big
#Midjourney
thread on how to write prompts to get good cinematic images.
In this thread I’ll build up a single prompt with cinematic elements, and show their effects.
Each prompt will use a 16:9 aspect ratio, and to minimise variation I've locked in a seed.
98
439
3K
19
104
562
There's now a usable implementation of DragGAN (that cool demo showing the lion opening and closing its mouth)
Unofficial implementation:
Colab:
Original:
Our team member (Zeqiang Lai) has reproduced
#DragGAN
and integrated it into InternGPT! Have a try! 🏄♂️
Demo:
Code:
14
145
505
8
154
555
I fine-tuned an LLM using Midjourney prompts and
@replicatehq
’s fine-tune API.
The model produces descriptive, keyword heavy prompts out of the box.
Try it here:
Please share your results!
Video guide on fine-tuning LLMs soon.
21
130
536
face + style – using IPAdapter for style, InstantID for consistency, prompt for steering and depth controlnet for continuity with original.
This also uses DreamshaperXL Lightning, so these images are in just 4 steps.
Will share details soon.
15
52
489
Two word prompt. I love stumbling on short prompts that do something magical.
What do you think the words were?
61
28
462
This morning I tried using voice to ask ChatGPT to get some football scores from the web, as now it has all the features baked in.
Seemed like it was working, but then I checked the transcript.
26
23
465
I've put together a new emoji SDXL fine-tune.
Try it with this:
"A TOK emoji of [subject]"
Works great with famous people.
33
57
451
I accidentally discovered that the Tron Legacy style
trained diffusion model can make some truly epic space gardens.
#stablediffusion
#AIart
10
53
432
Experimenting with consistent characters at different angles using InstantID, IPAdapters and RealVisXL
15
36
453
This is an exceptional post on how to use Stable Diffusion (via a1111) with ControlNet to do image restoration.
(It’s an involved process but has stellar results)
10
48
432
A new Midjourney workflow I’m really enjoying:
1. Take a photo
2. Use /describe
3. Append a subject to the very beginning of the prompt given by /describe
(Like a robot, or a cityscape)
18
43
392
Turn yourself into an emoji. I've added my emoji lora to the "face-to-many" model on Replicate.
14
41
396
You can ask ChatGPT to draw you SVGs, and with the API you can instantly render them.
It's like a cute lo-fi image generator. Oh and it can do some popular logos too.
13
59
366
I took a photo of a coffee shop interior, and asked for an isometric vector view from
#midjourney
v4
This is so powerful.
16
43
359
A quick experiment with
@Photoshop
’s new generative fill to combine two Midjourney creations.
Prompt: "a space port"
Scary how fast and easy this was.
13
38
354
🧵 I used image-to-video AI via Pika Labs to bring famous album covers to life.
Nirvana – Nevermind
28
61
344
If you've got the memory, you can make HD videos with SVD, using larger images.
This example is 1408 × 768 (bigger than 720p) straight out of SVD – no upscaling (but I did add interpolation).
Set the number of frames to decode at a time to 1 (decoding_t) to avoid OOM
1/2
25
24
324
Here it is💥!
A video guide to fine-tuning SDXL using Replicate.
- takes ~10 minutes
- needs 5 to 30 training images
- makes awesome results
Video covers:
- picking training images
- starting training
- using model
- prompting tips
- all training settings
Details in 🧵
10
55
312
Trying out typography and elaborate letterforms with
#midjourney
v4, and getting some luck with "letter ..." prompts.
These images spell out my handle, `fofr`.
18
45
306
You can now generate beautifully transparent SDXL images on Replicate, which I'm guessing is using LayerDiffusion:
Prompt: "a photo of a glass bottle"
3
38
305
Now I'm getting the first ChatGPT to generate prompts for Midjourney.
A second ChatGPT evaluates the quality of those prompts and ask for changes.
I give them a starting subject.
11
22
298
I asked ChatGPT how someone from the 15th century might explain an SLR camera. Then I put the response into Midjourney.
"Tis a large and complex contraption"
10
24
289
New ControlNet v1.1
Some things to get excited about in here:
- openpose with face and hands
- "shuffle", which takes an image and a prompt and essentially creates variations ("The model is trained to reorganize images")
- tiling, ie upscale using SD by…
3
50
276
Upscaled to 1080p and interpolated to 24fps, and this AI generation looks like actual video game footage (albeit with time paused).
🤯
First text-to-video render of Stable Video Diffusion (SVD) from a Midjourney input image.
I'm impressed with
- coherent movement
- video quality
- accuracy with original image
Shame the explosions didn't, uh, explode.
24
62
606
23
31
276
This is a neat deepfake detection technique
Intel's FakeCatcher uses a digital version of Photoplethysmography (rPPG) to detect heart flow
this method works by detecting the volume changes in blood vessels, by analyzing color variations in the video pixels that correspond to the blood flow across the face.
10
36
271
8
30
268
Generate transparent images directly with LayerDiffusion. No more background removal tools – it builds transparency into the diffusion process.
It's only on A1111 at the moment:
I need this in ComfyUI! This is the proper way.
10
36
267
Trying out MJ v5.1, the coherency on this is fantastic, all those background faces.
#midjourneyv51
7
14
249
This looks interesting – like AnimateAnyone and MagicAnimate, but with prompting and face transfer.
DreaMoving:
reference image + pose sequence + prompt
4
50
252
An attempt to turn Escher’s Relativity image into a photo using ControlNet and the canny model.
9
38
241
Prompt:
a wide angle photo of a television from a side angle, on the screen a lone news reader is looking at the camera, a CNN banner reports “Breaking news: We are not alone”, there is poor reception, it is a normal living room, a bit messy, day time, there is a window
16
20
234
A comparison of the latest controlnet depth map pre-processors in Automatic1111 stable-diffusion-webui:
1. Leres
2. Leres++
3. Midas (original)
4. Zoe
Leres++ is 🔥
8
36
236
A small thread of interesting things you can do with my become-image Replicate model:
1. You can use animated inputs to reimagine them as real world people, with all of their exaggerated features
9
28
237
I'm blown away by the lip-sync quality of the VideoReTalking model.
- created an image with my SDXL fine-tune
- passed img to
@runwayml
Gen2 to animate
- used video and VideoReTalking model to lip-sync:
6
30
235
Midjourney can make some gorgeous gradients.
Hello new iPhone wallpapers.
8
13
230
Trying out the Runway Gen2 image to video feature – it's taken a lot of attempts to get a video with good motion. Amazing when it works.
Image in thread.
14
22
228
I've been working on a new model on Replicate that lets you run any ComfyUI workflow with an API.
It supports all the popular controlnets, base weights, preprocessors, photomaker, animatediff, LCM, upscalers, IPAdapters.
Details in 🧵
10
32
225
For anyone looking for an open-source video upscaler,
@lucataco93
has added the Real-ESRGAN Video Upscaler to Replicate:
Upscale to Full HD, 2k or 4k
5
41
217
Having 8x upscaling on Midjourney outputs with the newest
@Magnific_AI
is epic. It's like magic✨
10
22
212
I've optimised my latent consistency model on Replicate to make it 3x faster:
text to image, 4 steps at 768x768:
- was: 2s
- now: 0.6s
6
21
206
These all use the Midjourney describe feature, which attempts to describe the image you give it as a prompt.
9
5
202
Image to video model "i2vgen-xl" is now on Replicate – make 16:9 videos at 1280x704 resolution:
Example (which I also interpolated):
It also seems to work well up to 48 frames.
9
25
202
New blog post up – Generate images in one second on your Mac using a latent consistency model.
1 second. Locally. 🤯
7
34
201
If you're interested in AI art, and doing more complex stuff with things like Stable Diffusion – I have a YouTube channel with some tutorials:
5
9
200