Skip to content
Python

Photodump starts each photo from its own vertical canvas with its own framing

neuregex941417980d98
diff --git a/apps/photodump/app.json b/apps/photodump/app.json
index d3b5aeb..55be209 100644
--- a/apps/photodump/app.json
+++ b/apps/photodump/app.json
@@ -11,7 +11,7 @@
"source": {
"template": "image_flux3_klein_distilled_image_edit.json",
"comfyui": "0.42.0",
- "notes": "apps/edit's pipeline four times over the same photo: four prompts (the aesthetic select writes all four, nodes 4 and pd_t2-4), four ReferenceMultiLatent and four samplers with the same seed, joined with ImageBatch into one SaveImage. The prompts follow apps/character's wording (\"The same person as in <ref_image_0>, <scene>\"): with \"A candid phone photo of the same person\" the first run (20261026-002124-l40s) pasted the photo, same pose and framing, on four backgrounds."
+ "notes": "apps/edit's pipeline four times over the same photo: four prompts (the aesthetic select writes all four, nodes 4 and pd_t2-4), four ReferenceMultiLatent and four samplers with the same seed, joined with ImageBatch into one SaveImage. Each of the four starts from its own empty 2:3 latent (EmptyFlux2LatentImage, node ff_size) with its own framing: full body, half body, from the side, a mirror selfie. Runs 20261026-002124-l40s and -003153-l40s, which started from the photo's latent shape, kept the photo's pose and framing in two to four of the photos."
},
"inputs": [
{
@@ -43,22 +43,22 @@
{
"node": "4",
"field": "text",
- "value": "The same person as in <ref_image_0>, on a neon-lit street at night, rain on the pavement, in a new pose and clothes that suit the scene, seen from a new angle: a candid phone snapshot. Keep the face, hairstyle and build identical. No text, no logos. {details}"
+ "value": "The same person as in <ref_image_0>, on a neon-lit street at night, rain on the pavement, full body, walking toward the camera, in clothes that suit the scene: a candid vertical phone snapshot. Keep the face, hairstyle and build identical. No text, no logos. {details}"
},
{
"node": "pd_t2",
"field": "text",
- "value": "The same person as in <ref_image_0>, at a rooftop bar at dusk with the city lights behind, in a new pose and clothes that suit the scene, seen from a new angle: a candid phone snapshot. Keep the face, hairstyle and build identical. No text, no logos. {details}"
+ "value": "The same person as in <ref_image_0>, at a rooftop bar at dusk with the city lights behind, half body, sitting and laughing, looking away from the camera, in clothes that suit the scene: a candid vertical phone snapshot. Keep the face, hairstyle and build identical. No text, no logos. {details}"
},
{
"node": "pd_t3",
"field": "text",
- "value": "The same person as in <ref_image_0>, on an empty subway platform at night, a train arriving with motion blur, in a new pose and clothes that suit the scene, seen from a new angle: a candid phone snapshot. Keep the face, hairstyle and build identical. No text, no logos. {details}"
+ "value": "The same person as in <ref_image_0>, on a subway platform at night, a train arriving with motion blur, seen from the side, looking at something out of frame, in clothes that suit the scene: a candid vertical phone snapshot. Keep the face, hairstyle and build identical. No text, no logos. {details}"
},
{
"node": "pd_t4",
"field": "text",
- "value": "The same person as in <ref_image_0>, in a diner booth late at night, a milkshake on the table, direct flash, in a new pose and clothes that suit the scene, seen from a new angle: a candid phone snapshot. Keep the face, hairstyle and build identical. No text, no logos. {details}"
+ "value": "The same person as in <ref_image_0>, in a diner restroom mirror late at night, direct flash, taking a mirror selfie, the phone partly covering the chin, in clothes that suit the scene: a candid vertical phone snapshot. Keep the face, hairstyle and build identical. No text, no logos. {details}"
}
]
},
@@ -69,22 +69,22 @@
{
"node": "4",
"field": "text",
- "value": "The same person as in <ref_image_0>, walking barefoot along the shoreline at golden hour, holding sandals, in a new pose and clothes that suit the scene, seen from a new angle: a candid phone snapshot. Keep the face, hairstyle and build identical. No text, no logos. {details}"
+ "value": "The same person as in <ref_image_0>, on the shoreline at golden hour, holding sandals, full body, walking toward the camera, in clothes that suit the scene: a candid vertical phone snapshot. Keep the face, hairstyle and build identical. No text, no logos. {details}"
},
{
"node": "pd_t2",
"field": "text",
- "value": "The same person as in <ref_image_0>, at a beach bar table with a fresh coconut drink, laughing, in a new pose and clothes that suit the scene, seen from a new angle: a candid phone snapshot. Keep the face, hairstyle and build identical. No text, no logos. {details}"
+ "value": "The same person as in <ref_image_0>, at a beach bar table with a fresh coconut drink, half body, sitting and laughing, looking away from the camera, in clothes that suit the scene: a candid vertical phone snapshot. Keep the face, hairstyle and build identical. No text, no logos. {details}"
},
{
"node": "pd_t3",
"field": "text",
- "value": "The same person as in <ref_image_0>, on a small boat with the sea behind, wind in the hair, in a new pose and clothes that suit the scene, seen from a new angle: a candid phone snapshot. Keep the face, hairstyle and build identical. No text, no logos. {details}"
+ "value": "The same person as in <ref_image_0>, on a small boat with the sea behind, wind in the hair, seen from the side, looking at something out of frame, in clothes that suit the scene: a candid vertical phone snapshot. Keep the face, hairstyle and build identical. No text, no logos. {details}"
},
{
"node": "pd_t4",
"field": "text",
- "value": "The same person as in <ref_image_0>, in a sunny old-town street with whitewashed walls, eating an ice cream, in a new pose and clothes that suit the scene, seen from a new angle: a candid phone snapshot. Keep the face, hairstyle and build identical. No text, no logos. {details}"
+ "value": "The same person as in <ref_image_0>, in a beach house mirror, sun-bleached light, taking a mirror selfie, the phone partly covering the chin, in clothes that suit the scene: a candid vertical phone snapshot. Keep the face, hairstyle and build identical. No text, no logos. {details}"
}
]
},
@@ -95,22 +95,22 @@
{
"node": "4",
"field": "text",
- "value": "The same person as in <ref_image_0>, curled up on a couch under a knitted blanket, reading a book, in a new pose and clothes that suit the scene, seen from a new angle: a candid phone snapshot. Keep the face, hairstyle and build identical. No text, no logos. {details}"
+ "value": "The same person as in <ref_image_0>, in a small kitchen, carrying a pot of pasta, full body, walking toward the camera, in clothes that suit the scene: a candid vertical phone snapshot. Keep the face, hairstyle and build identical. No text, no logos. {details}"
},
{
"node": "pd_t2",
"field": "text",
- "value": "The same person as in <ref_image_0>, cooking pasta in a small kitchen, steam rising, in a new pose and clothes that suit the scene, seen from a new angle: a candid phone snapshot. Keep the face, hairstyle and build identical. No text, no logos. {details}"
+ "value": "The same person as in <ref_image_0>, on a couch under a knitted blanket with a book, half body, sitting and laughing, looking away from the camera, in clothes that suit the scene: a candid vertical phone snapshot. Keep the face, hairstyle and build identical. No text, no logos. {details}"
},
{
"node": "pd_t3",
"field": "text",
- "value": "The same person as in <ref_image_0>, holding a coffee mug by a rainy window in the morning, in a new pose and clothes that suit the scene, seen from a new angle: a candid phone snapshot. Keep the face, hairstyle and build identical. No text, no logos. {details}"
+ "value": "The same person as in <ref_image_0>, by a rainy window in the morning, holding a coffee mug, seen from the side, looking at something out of frame, in clothes that suit the scene: a candid vertical phone snapshot. Keep the face, hairstyle and build identical. No text, no logos. {details}"
},
{
"node": "pd_t4",
"field": "text",
- "value": "The same person as in <ref_image_0>, sitting on the floor among house plants, sorting vinyl records, in a new pose and clothes that suit the scene, seen from a new angle: a candid phone snapshot. Keep the face, hairstyle and build identical. No text, no logos. {details}"
+ "value": "The same person as in <ref_image_0>, in a bathroom mirror in the morning light, taking a mirror selfie, the phone partly covering the chin, in clothes that suit the scene: a candid vertical phone snapshot. Keep the face, hairstyle and build identical. No text, no logos. {details}"
}
]
},
@@ -121,22 +121,22 @@
{
"node": "4",
"field": "text",
- "value": "The same person as in <ref_image_0>, on a forest hiking trail with a backpack, in a new pose and clothes that suit the scene, seen from a new angle: a candid phone snapshot. Keep the face, hairstyle and build identical. No text, no logos. {details}"
+ "value": "The same person as in <ref_image_0>, on a forest hiking trail with a backpack, full body, walking toward the camera, in clothes that suit the scene: a candid vertical phone snapshot. Keep the face, hairstyle and build identical. No text, no logos. {details}"
},
{
"node": "pd_t2",
"field": "text",
- "value": "The same person as in <ref_image_0>, at a picnic on the grass in a park, friends out of focus, in a new pose and clothes that suit the scene, seen from a new angle: a candid phone snapshot. Keep the face, hairstyle and build identical. No text, no logos. {details}"
+ "value": "The same person as in <ref_image_0>, at a picnic on the grass in a park, friends out of focus, half body, sitting and laughing, looking away from the camera, in clothes that suit the scene: a candid vertical phone snapshot. Keep the face, hairstyle and build identical. No text, no logos. {details}"
},
{
"node": "pd_t3",
"field": "text",
- "value": "The same person as in <ref_image_0>, riding a bicycle along a lakeside path, in a new pose and clothes that suit the scene, seen from a new angle: a candid phone snapshot. Keep the face, hairstyle and build identical. No text, no logos. {details}"
+ "value": "The same person as in <ref_image_0>, on a lakeside path next to a bicycle, seen from the side, looking at something out of frame, in clothes that suit the scene: a candid vertical phone snapshot. Keep the face, hairstyle and build identical. No text, no logos. {details}"
},
{
"node": "pd_t4",
"field": "text",
- "value": "The same person as in <ref_image_0>, by a campfire at dusk, wrapped in a flannel shirt, in a new pose and clothes that suit the scene, seen from a new angle: a candid phone snapshot. Keep the face, hairstyle and build identical. No text, no logos. {details}"
+ "value": "The same person as in <ref_image_0>, in a cabin mirror, a flannel shirt on, taking a mirror selfie, the phone partly covering the chin, in clothes that suit the scene: a candid vertical phone snapshot. Keep the face, hairstyle and build identical. No text, no logos. {details}"
}
]
},
@@ -147,22 +147,22 @@
{
"node": "4",
"field": "text",
- "value": "The same person as in <ref_image_0>, on a colorful street in Coyoacán with papel picado overhead, in a new pose and clothes that suit the scene, seen from a new angle: a candid phone snapshot. Keep the face, hairstyle and build identical. No text, no logos. {details}"
+ "value": "The same person as in <ref_image_0>, on a colorful street in Coyoacán with papel picado overhead, full body, walking toward the camera, in clothes that suit the scene: a candid vertical phone snapshot. Keep the face, hairstyle and build identical. No text, no logos. {details}"
},
{
"node": "pd_t2",
"field": "text",
- "value": "The same person as in <ref_image_0>, on a trajinera boat in Xochimilco, flowers on its arch, in a new pose and clothes that suit the scene, seen from a new angle: a candid phone snapshot. Keep the face, hairstyle and build identical. No text, no logos. {details}"
+ "value": "The same person as in <ref_image_0>, on a trajinera boat in Xochimilco, flowers on its arch, half body, sitting and laughing, looking away from the camera, in clothes that suit the scene: a candid vertical phone snapshot. Keep the face, hairstyle and build identical. No text, no logos. {details}"
},
{
"node": "pd_t3",
"field": "text",
- "value": "The same person as in <ref_image_0>, at a market stall full of fruit, holding a mango on a stick, in a new pose and clothes that suit the scene, seen from a new angle: a candid phone snapshot. Keep the face, hairstyle and build identical. No text, no logos. {details}"
+ "value": "The same person as in <ref_image_0>, at a market stall full of fruit, holding a mango on a stick, seen from the side, looking at something out of frame, in clothes that suit the scene: a candid vertical phone snapshot. Keep the face, hairstyle and build identical. No text, no logos. {details}"
},
{
"node": "pd_t4",
"field": "text",
- "value": "The same person as in <ref_image_0>, on a terrace at sunset with the city behind, in a new pose and clothes that suit the scene, seen from a new angle: a candid phone snapshot. Keep the face, hairstyle and build identical. No text, no logos. {details}"
+ "value": "The same person as in <ref_image_0>, in an elevator mirror of an old Roma Norte building, taking a mirror selfie, the phone partly covering the chin, in clothes that suit the scene: a candid vertical phone snapshot. Keep the face, hairstyle and build identical. No text, no logos. {details}"
}
]
},
@@ -173,22 +173,22 @@
{
"node": "4",
"field": "text",
- "value": "The same person as in <ref_image_0>, against a white wall, full body, direct flash, in a new pose and clothes that suit the scene, seen from a new angle: a candid phone snapshot. Keep the face, hairstyle and build identical. No text, no logos. {details}"
+ "value": "The same person as in <ref_image_0>, against a white wall, direct flash, full body, walking toward the camera, in clothes that suit the scene: a candid vertical phone snapshot. Keep the face, hairstyle and build identical. No text, no logos. {details}"
},
{
"node": "pd_t2",
"field": "text",
- "value": "The same person as in <ref_image_0>, in a close-up with direct flash and a hard shadow behind, in a new pose and clothes that suit the scene, seen from a new angle: a candid phone snapshot. Keep the face, hairstyle and build identical. No text, no logos. {details}"
+ "value": "The same person as in <ref_image_0>, on a stool, direct flash and a hard shadow behind, half body, sitting and laughing, looking away from the camera, in clothes that suit the scene: a candid vertical phone snapshot. Keep the face, hairstyle and build identical. No text, no logos. {details}"
},
{
"node": "pd_t3",
"field": "text",
- "value": "The same person as in <ref_image_0>, sitting on a stool, legs crossed, direct flash, in a new pose and clothes that suit the scene, seen from a new angle: a candid phone snapshot. Keep the face, hairstyle and build identical. No text, no logos. {details}"
+ "value": "The same person as in <ref_image_0>, against a red wall, direct flash, seen from the side, looking at something out of frame, in clothes that suit the scene: a candid vertical phone snapshot. Keep the face, hairstyle and build identical. No text, no logos. {details}"
},
{
"node": "pd_t4",
"field": "text",
- "value": "The same person as in <ref_image_0>, looking over the shoulder against a red wall, direct flash, in a new pose and clothes that suit the scene, seen from a new angle: a candid phone snapshot. Keep the face, hairstyle and build identical. No text, no logos. {details}"
+ "value": "The same person as in <ref_image_0>, in a studio mirror, direct flash, taking a mirror selfie, the phone partly covering the chin, in clothes that suit the scene: a candid vertical phone snapshot. Keep the face, hairstyle and build identical. No text, no logos. {details}"
}
]
}
diff --git a/apps/photodump/workflow.api.json b/apps/photodump/workflow.api.json
index c7e7f58..a6b8b2d 100644
--- a/apps/photodump/workflow.api.json
+++ b/apps/photodump/workflow.api.json
@@ -58,8 +58,8 @@
1
],
"latent_image": [
- "17",
- 2
+ "ff_size",
+ 0
],
"seed": 563228241303057,
"steps": 16,
@@ -212,8 +212,8 @@
1
],
"latent_image": [
- "pd_r2",
- 2
+ "ff_size",
+ 0
],
"seed": 563228241303057,
"steps": 16,
@@ -303,8 +303,8 @@
1
],
"latent_image": [
- "pd_r3",
- 2
+ "ff_size",
+ 0
],
"seed": 563228241303057,
"steps": 16,
@@ -394,8 +394,8 @@
1
],
"latent_image": [
- "pd_r4",
- 2
+ "ff_size",
+ 0
],
"seed": 563228241303057,
"steps": 16,
@@ -462,5 +462,16 @@
"_meta": {
"title": "The four photos"
}
+ },
+ "ff_size": {
+ "class_type": "EmptyFlux2LatentImage",
+ "inputs": {
+ "width": 832,
+ "height": 1248,
+ "batch_size": 1
+ },
+ "_meta": {
+ "title": "Phone photo, 2:3"
+ }
}
}
diff --git a/lab/cases.json b/lab/cases.json
index 3e4db58..107f7a6 100644
--- a/lab/cases.json
+++ b/lab/cases.json
@@ -529,7 +529,7 @@
"note": "AI Influencer, character type Bold: the example card of that option. Does the type show and still read as a photo?"},
{"id": "aiinf-extreme", "app": "ai-influencer", "values": {"kind": "a person with extremely exaggerated, cartoonish features and proportions, still photographed for real", "gender": "A man", "hair": "Short hair.", "facialhair": "A mustache.", "feature": "", "style": "Wearing a tailored suit."},
"note": "AI Influencer, character type Extreme: the example card of that option. Does the type show and still read as a photo?"},
- {"id": "aiinf-insect", "app": "ai-influencer", "values": {"kind": "a character whose whole head is a grasshopper's head, with large compound eyes and long antennae, on a human body: no human face", "style": "Wearing gothic black clothes."},
+ {"id": "aiinf-insect", "app": "ai-influencer", "values": {"kind": "a character whose whole head is a grasshopper's head, with large compound eyes and long antennae, on a human body: no human face", "style": "Wearing gothic black clothes.", "hair": "", "haircolor": "", "feature": ""},
"note": "AI Influencer, character type Insect: the example card of that option. Does the type show and still read as a photo?"},
{"id": "aiinf-frog", "app": "ai-influencer", "values": {"kind": "a character whose whole head is a green frog's head, with a wide mouth and big round eyes, on a human body: no human face", "gender": "A man", "hair": "Short hair.", "facialhair": "", "feature": ""},
"note": "AI Influencer, character type Frog: the example card of that option. Does the type show and still read as a photo?"},
@@ -539,7 +539,7 @@
"note": "AI Influencer, character type Dog: the example card of that option. Does the type show and still read as a photo?"},
{"id": "aiinf-rodent", "app": "ai-influencer", "values": {"kind": "an anthropomorphic capybara character: a capybara head on a human body", "gender": "A man", "hair": "Short hair.", "facialhair": "", "feature": "", "style": "Wearing 1970s retro clothes."},
"note": "AI Influencer, character type Rodent: the example card of that option. Does the type show and still read as a photo?"},
- {"id": "aiinf-bird", "app": "ai-influencer", "values": {"kind": "a character whose whole head is a puffin's head, black and white feathers and a bright orange beak, on a human body: no human face", "style": "Wearing sporty athletic clothes."},
+ {"id": "aiinf-bird", "app": "ai-influencer", "values": {"kind": "a character whose whole head is a puffin's head, black and white feathers and a bright orange beak, on a human body: no human face", "style": "Wearing sporty athletic clothes.", "hair": "", "haircolor": "", "feature": ""},
"note": "AI Influencer, character type Bird: the example card of that option. Does the type show and still read as a photo?"},
{"id": "create-look-disposable", "app": "create-image", "values": {"prompt": "Two friends laughing at a taco stand at night in Mexico City", "expand": false, "look": "Shot on a cheap disposable film camera with its built-in flash: harsh flash, slight vignette, film grain, a warm color cast.", "aspect": "3:2", "resolution": 1.5},
"note": "Create Image with a look, 3:2 at 1.5K (1872 x 1248): does the look show, and does the scaled latent come out sharp at the new size?"},
diff --git a/tools/build_apps.py b/tools/build_apps.py
index 2650c28..f2e6026 100644
--- a/tools/build_apps.py
+++ b/tools/build_apps.py
@@ -621,6 +621,10 @@ def recipes(templates: Path, ltx: Path, info: dict, reshoot: Path | None = None)
pd["pd_b3"] = {"class_type": "ImageBatch", "inputs": {"image1": ["pd_b2", 0], "image2": ["pd_d3", 0]}}
pd["pd_b4"] = {"class_type": "ImageBatch", "inputs": {"image1": ["pd_b3", 0], "image2": ["pd_d4", 0]}, "_meta": {"title": "The four photos"}}
put(pd, "9", "images", ["pd_b4", 0])
+ # its own vertical canvas: started from the photo's latent shape, two of four photos kept the photo's framing (run 20261026-003153-l40s)
+ pd["ff_size"] = {"class_type": "EmptyFlux2LatentImage", "inputs": {"width": 832, "height": 1248, "batch_size": 1}, "_meta": {"title": "Phone photo, 2:3"}}
+ for ks in ("7", "pd_k2", "pd_k3", "pd_k4"):
+ put(pd, ks, "latent_image", ["ff_size", 0])
out["photodump"] = pd
ip = json.loads(json.dumps(e))