feature: image edit gifs and demo-reels

README.md

@@ -61,6 +61,17 @@ with prompt-based masking.
 <img src="assets/bowl_of_fruit_strawberries.jpg" height="256"><img src="assets/freckled_woman_cyborg.jpg" height="256"><br>
 <img src="assets/girl_with_a_pearl_earring_clown_makeup.jpg" height="256"><img src="assets/mona-lisa-headshot-photo.jpg" height="256"><br>
 
+Want just quickly have some fun? Try `--suprise-me` to apply some pre-defined edits.
+```bash
+>> aimg edit --gif --suprise-me pearl_girl.jpg
+>> aimg edit --gif --suprise-me mona-lisa.jpg
+>> aimg edit --gif --suprise-me luke.jpg
+>> aimg edit --gif --suprise-me spock.jpg
+```
+<img src="assets/girl_with_a_pearl_earring_suprise.gif" height="256"><img src="assets/mona-lisa-suprise.gif" height="256"><br>
+<img src="assets/luke-suprise.gif" height="256"><img src="assets/spock-suprise.gif" height="256"><br>
+<img src="assets/gg-bridge-suprise.gif" height="256"><img src="assets/shire-suprise.gif" height="256"><br>
+
 
 ### Prompt Based Masking  [by clipseg](https://github.com/timojl/clipseg)
 Specify advanced text based masks using boolean logic and strength modifiers. 
@@ -274,6 +285,8 @@ docker run -it --gpus all -v $HOME/.cache/huggingface:/root/.cache/huggingface -
 
 **8.0.0**
 - feature: 🎉 edit images with instructions alone!
+- feature: when editing an image add `--gif` to create a comparision gif
+- feature: `aimg edit --suprise-me --gif my-image.jpg` for some fun pre-programmed edits
 - feature: prune-ckpt command also removes the non-ema weights
 
 **7.6.0**

imaginairy/api.py

@@ -13,7 +13,11 @@ from imaginairy.enhancers.clip_masking import get_img_mask
 from imaginairy.enhancers.describe_image_blip import generate_caption
 from imaginairy.enhancers.face_restoration_codeformer import enhance_faces
 from imaginairy.enhancers.upscale_realesrgan import upscale_image
-from imaginairy.img_utils import pillow_fit_image_within, pillow_img_to_torch_image
+from imaginairy.img_utils import (
+    make_gif_image,
+    pillow_fit_image_within,
+    pillow_img_to_torch_image,
+)
 from imaginairy.log_utils import (
     ImageLoggingContext,
     log_conditioning,
@@ -55,6 +59,8 @@ def imagine_image_files(
     record_step_images=False,
     output_file_extension="jpg",
     print_caption=False,
+    make_comparison_gif=False,
+    return_filename_type="generated",
 ):
     generated_imgs_path = os.path.join(outdir, "generated")
     os.makedirs(generated_imgs_path, exist_ok=True)
@@ -74,6 +80,7 @@ def imagine_image_files(
         draw.text((10, 10), str(description))
         img.save(destination)
 
+    result_filenames = []
     for result in imagine(
         prompts,
         precision=precision,
@@ -97,9 +104,25 @@ def imagine_image_files(
             )
             result.save(filepath, image_type=image_type)
             logger.info(f"🖼  [{image_type}] saved to: {filepath}")
+            if image_type == return_filename_type:
+                result_filenames.append(filepath)
+        if make_comparison_gif and prompt.init_image:
+            subpath = os.path.join(outdir, "gif")
+            os.makedirs(subpath, exist_ok=True)
+            filepath = os.path.join(subpath, f"{basefilename}.gif")
+            resized_init_image = pillow_fit_image_within(
+                prompt.init_image, prompt.width, prompt.height
+            )
+            make_gif_image(
+                filepath,
+                imgs=[result.images["generated"], resized_init_image],
+                duration=1750,
+            )
         base_count += 1
         del result
 
+    return result_filenames
+
 
 def imagine(
     prompts,

imaginairy/cmds.py

@@ -11,6 +11,7 @@ from imaginairy.enhancers.prompt_expansion import expand_prompts
 from imaginairy.log_utils import configure_logging
 from imaginairy.samplers import SAMPLER_TYPE_OPTIONS
 from imaginairy.schema import ImaginePrompt
+from imaginairy.suprise_me import create_suprise_me_images
 from imaginairy.train import train_diffusion_model
 from imaginairy.training_tools.image_prep import (
     create_class_images,
@@ -220,6 +221,9 @@ logger = logging.getLogger(__name__)
     is_flag=True,
     help="Print the version and exit.",
 )
+@click.option(
+    "--gif", "make_gif", default=False, is_flag=True, help="Generate a gif of the edit."
+)
 @click.pass_context
 def imagine_cmd(
     ctx,
@@ -255,6 +259,7 @@ def imagine_cmd(
     model_config_path,
     prompt_library_path,
     version,  # noqa
+    make_gif,
 ):
     """Have the AI generate images. alias:imagine."""
     return _imagine_cmd(
@@ -291,6 +296,7 @@ def imagine_cmd(
         model_config_path,
         prompt_library_path,
         version,  # noqa
+        make_gif
     )
 
 
@@ -487,8 +493,22 @@ def imagine_cmd(
     is_flag=True,
     help="Print the version and exit.",
 )
+@click.option(
+    "--gif",
+    "make_gif",
+    default=False,
+    is_flag=True,
+    help="Generate a gif comparing the original image to the modified one.",
+)
+@click.option(
+    "--suprise-me",
+    "suprise_me",
+    default=False,
+    is_flag=True,
+    help="make some fun edits to the provided image",
+)
 @click.pass_context
-def edit_image(
+def edit_image(  # noqa
     ctx,
     init_image,
     prompt_texts,
@@ -521,8 +541,21 @@ def edit_image(
     model_config_path,
     prompt_library_path,
     version,  # noqa
+    make_gif,
+    suprise_me,
 ):
     init_image_strength = 1
+    if suprise_me and prompt_texts:
+        raise ValueError("Cannot use suprise_me and prompt_texts together")
+
+    if suprise_me:
+        if quiet:
+            log_level = "ERROR"
+        configure_logging(log_level)
+        create_suprise_me_images(init_image, outdir=outdir, make_gif=make_gif)
+
+        return
+
     return _imagine_cmd(
         ctx,
         prompt_texts,
@@ -557,6 +590,7 @@ def edit_image(
         model_config_path,
         prompt_library_path,
         version,  # noqa
+        make_gif,
     )
 
 
@@ -593,7 +627,8 @@ def _imagine_cmd(
     model_weights_path,
     model_config_path,
     prompt_library_path,
-    version,  # noqa
+    version=False,  # noqa
+    make_gif=False,
 ):
     """Have the AI generate images. alias:imagine."""
     if ctx.invoked_subcommand is not None:
@@ -670,6 +705,7 @@ def _imagine_cmd(
         output_file_extension="jpg",
         print_caption=caption,
         precision=precision,
+        make_comparison_gif=make_gif,
     )
 
 

imaginairy/img_utils.py

@@ -70,3 +70,15 @@ def pillow_img_to_model_latent(model, img, batch_size=1, half=True):
             model.encode_first_stage(init_image.half())
         )
     return model.get_first_stage_encoding(model.encode_first_stage(init_image))
+
+
+def make_gif_image(filepath, imgs, duration=1000, loop=0):
+
+    imgs[0].save(
+        filepath,
+        save_all=True,
+        append_images=imgs[1:],
+        duration=duration,
+        loop=loop,
+        optimize=False,
+    )

imaginairy/modules/diffusion/ddpm.py

@@ -32,6 +32,7 @@ from imaginairy.modules.diffusion.util import (
 )
 from imaginairy.modules.distributions import DiagonalGaussianDistribution
 from imaginairy.modules.ema import LitEma
+from imaginairy.paths import PKG_ROOT
 from imaginairy.samplers.kdiff import DPMPP2MSampler
 from imaginairy.utils import instantiate_from_config
 
@@ -47,7 +48,7 @@ def log_txt_as_img(wh, xc, size=10):
     for bi in range(b):
         txt = Image.new("RGB", wh, color="white")
         draw = ImageDraw.Draw(txt)
-        font = ImageFont.truetype("data/DejaVuSans.ttf", size=size)
+        font = ImageFont.truetype(f"{PKG_ROOT}/data/DejaVuSans.ttf", size=size)
         nc = int(40 * (wh[0] / 256))
         lines = "\n".join(
             xc[bi][start : start + nc] for start in range(0, len(xc[bi]), nc)

imaginairy/suprise_me.py

@@ -0,0 +1,128 @@
+"""
+
+
+aimg.
+"""
+
+import os.path
+
+from PIL import ImageDraw, ImageFont
+
+from imaginairy import ImaginePrompt, LazyLoadingImage, imagine_image_files
+from imaginairy.enhancers.facecrop import detect_faces
+from imaginairy.img_utils import make_gif_image, pillow_fit_image_within
+from imaginairy.paths import PKG_ROOT
+
+generic_prompts = [
+    ("make it anime style", 18, ""),
+    ("make it pen and ink style", 20, ""),
+    ("make it a thomas kinkade painting", 20, ""),
+    ("make it pixar style", 20, ""),
+    ("make it look like a marble statue", 15, ""),
+    ("make it look like a golden statue", 15, ""),
+    ("make it look like a snowstorm", 20, ""),
+    ("make it night", 15, ""),
+    ("make it a sunny day", 15, ""),
+    ("put fireworks in the background", 15, ""),
+    ("make it in a forest", 15, ""),
+]
+
+person_prompts = [
+    ("make the person close their eyes", 10, ""),
+    ("make the person wear clown makeup", 10, ""),
+    ("make the person a cyborg", 14, ""),
+    ("make the person wear shiny metal clothes", 14, ""),
+    ("make the person wear a tie-dye shirt", 7.5, ""),
+    ("make the person wear a suit", 7.5, ""),
+    ("make the person bald", 7.5, ""),
+    ("change the hair to pink", 7.5, ""),
+    (
+        "make it a color professional photo headshot. Canon EOS, sharp focus",
+        10,
+        "old, ugly",
+    ),
+    ("make the face smiling", 5, ""),
+    # ("make the person sad", 20, ""),
+    # ("make the person angry", 20, ""),
+    ("make the person look like a celebrity", 10, ""),
+    ("make the person younger", 10, ""),
+    ("make the person older", 10, ""),
+    ("make the person a disney cartoon character", 7.5, ""),
+]
+
+
+def suprise_me_prompts(img, person=None):
+    prompts = []
+    if isinstance(img, str):
+        if img.startswith("http"):
+            img = LazyLoadingImage(url=img)
+        else:
+            img = LazyLoadingImage(filepath=img)
+
+    if person is None:
+        person = bool(detect_faces(img))
+
+    if person:
+        for prompt_text, strength, neg_prompt_text in person_prompts:
+            prompts.append(
+                ImaginePrompt(
+                    prompt_text,
+                    init_image=img,
+                    prompt_strength=strength,
+                    negative_prompt=neg_prompt_text,
+                    model="edit",
+                    steps=20,
+                )
+            )
+
+    for prompt_text, strength, neg_prompt_text in generic_prompts:
+        prompts.append(
+            ImaginePrompt(
+                prompt_text,
+                init_image=img,
+                prompt_strength=strength,
+                negative_prompt=neg_prompt_text,
+                model="edit",
+                steps=20,
+            )
+        )
+
+    return prompts
+
+
+def create_suprise_me_images(img, outdir, person=None, make_gif=True):
+    if isinstance(img, str):
+        if img.startswith("http"):
+            img = LazyLoadingImage(url=img)
+        else:
+            img = LazyLoadingImage(filepath=img)
+    prompts = suprise_me_prompts(img, person=person)
+    generated_filenames = imagine_image_files(
+        prompts,
+        outdir=outdir,
+        record_step_images=False,
+        output_file_extension="jpg",
+        print_caption=False,
+        make_comparison_gif=make_gif,
+    )
+    if make_gif:
+        imgs_path = os.path.join(outdir, "compilations")
+        os.makedirs(imgs_path, exist_ok=True)
+        base_count = len(os.listdir(imgs_path))
+        new_filename = os.path.join(imgs_path, f"suprise_me_{base_count:03d}.gif")
+        simg = pillow_fit_image_within(img, prompts[0].width, prompts[0].height)
+        gif_imgs = [simg]
+        for prompt, filename in zip(prompts, generated_filenames):
+            gen_img = LazyLoadingImage(filepath=filename)
+            draw = ImageDraw.Draw(gen_img)
+
+            font_size = 16
+            font = ImageFont.truetype(f"{PKG_ROOT}/data/DejaVuSans.ttf", font_size)
+
+            x = 15
+            y = 485
+
+            draw.text((x, y), prompt.prompt_text, font=font, fill=(255, 255, 255))
+            gif_imgs.append(gen_img)
+
+        make_gif_image(new_filename, gif_imgs)

setup.py

@@ -25,6 +25,7 @@ setup(
     package_data={
         "imaginairy": [
             "configs/*.yaml",
+            "data/*.*",
             "enhancers/phraselists/*.txt",
             "vendored/clip/*.txt.gz",
             "vendored/clipseg/*.pth",