File size: 5,663 Bytes
126aa72
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
"""Batch-generate RGBA object images from a running ComfyUI server.

Reads a prompt per line, submits the API workflow, waits for completion and saves
the PNG (alpha preserved) into --out.

Usage:
    python generate_images.py --server http://127.0.0.1:8188 \
        --workflow ../workflows/qwen_2.1_turbo_api.json \
        --objects ../prompts/objects.txt --template ../prompts/prompt_template.txt \
        --out ../samples/images --size 1024 --seed 42
"""
import argparse
import json
import os
import time
import urllib.parse
import urllib.request
import uuid


def post_json(server, path, payload):
    data = json.dumps(payload).encode()
    req = urllib.request.Request(server + path, data=data,
                                 headers={"Content-Type": "application/json"})
    with urllib.request.urlopen(req, timeout=60) as r:
        return json.loads(r.read())


def get_json(server, path):
    with urllib.request.urlopen(server + path, timeout=60) as r:
        return json.loads(r.read())


def get_bytes(server, query):
    with urllib.request.urlopen(server + "/view?" + query, timeout=120) as r:
        return r.read()


def wait_for(server, prompt_id, timeout=1800, poll=2.0):
    t0 = time.time()
    while time.time() - t0 < timeout:
        hist = get_json(server, f"/history/{prompt_id}")
        if prompt_id in hist:
            entry = hist[prompt_id]
            status = entry.get("status", {})
            if status.get("completed") or entry.get("outputs"):
                return entry
            if status.get("status_str") == "error":
                raise RuntimeError(f"ComfyUI error: {json.dumps(status)[:800]}")
        time.sleep(poll)
    raise TimeoutError(f"prompt {prompt_id} did not finish in {timeout}s")


def build(base, node_ids, prompt, negative, width, height, seed, prefix):
    wf = json.loads(json.dumps(base))
    wf[node_ids["text_encode"]]["inputs"]["prompt"] = prompt
    wf[node_ids["text_encode"]]["inputs"]["negative_prompt"] = negative
    wf[node_ids["text_encode"]]["inputs"]["resolution"] = width
    wf[node_ids["latent"]]["inputs"]["width"] = width
    wf[node_ids["latent"]]["inputs"]["height"] = height
    wf[node_ids["noise"]]["inputs"]["noise_seed"] = seed
    wf[node_ids["save"]]["inputs"]["filename_prefix"] = prefix
    return wf


def main():
    ap = argparse.ArgumentParser()
    ap.add_argument("--server", default="http://127.0.0.1:8188")
    ap.add_argument("--workflow", required=True)
    ap.add_argument("--objects", required=True)
    ap.add_argument("--template", required=True)
    ap.add_argument("--out", required=True)
    ap.add_argument("--size", type=int, default=1024)
    ap.add_argument("--seed", type=int, default=42)
    ap.add_argument("--negative", default="")
    ap.add_argument("--timeout", type=int, default=1800)
    ap.add_argument("--only", default=None, help="comma separated subset of object names")
    ap.add_argument("--keep-loaded", action="store_true",
                    help="do not unload models between images (needs enough RAM)")
    args = ap.parse_args()

    base = json.load(open(args.workflow, encoding="utf-8"))
    node_ids = base.pop("_node_ids")
    template = open(args.template, encoding="utf-8").read().strip()
    objects = [l.strip() for l in open(args.objects, encoding="utf-8") if l.strip()]
    if args.only:
        wanted = {s.strip().lower() for s in args.only.split(",")}
        objects = [o for o in objects if o.lower() in wanted]

    os.makedirs(args.out, exist_ok=True)
    client_id = str(uuid.uuid4())
    results = []

    for i, obj in enumerate(objects):
        name = obj.lower().replace(" ", "_")
        target = os.path.join(args.out, f"{name}.png")
        if os.path.exists(target):
            print(f"[{i+1}/{len(objects)}] skip {obj} (exists)")
            results.append({"object": obj, "file": target, "skipped": True})
            continue
        prompt = template.format(object=obj)
        wf = build(base, node_ids, prompt, args.negative, args.size, args.size,
                   args.seed + i, f"idol3d/{name}")
        t0 = time.time()
        resp = post_json(args.server, "/prompt", {"prompt": wf, "client_id": client_id})
        entry = wait_for(args.server, resp["prompt_id"], timeout=args.timeout)
        images = []
        for node_out in entry.get("outputs", {}).values():
            images.extend(node_out.get("images", []))
        if not images:
            raise RuntimeError(f"no image returned for {obj}")
        img = images[0]
        query = urllib.parse.urlencode({"filename": img["filename"],
                                        "subfolder": img.get("subfolder", ""),
                                        "type": img.get("type", "output")})
        data = get_bytes(args.server, query)
        with open(target, "wb") as f:
            f.write(data)
        dt = time.time() - t0
        print(f"[{i+1}/{len(objects)}] {obj}: {dt:.1f}s -> {target}")
        results.append({"object": obj, "file": target, "seconds": round(dt, 2),
                        "seed": args.seed + i})
        if not args.keep_loaded:
            # 8 GB cards: unloading avoids the text-encoder reload crash on 16 GB RAM
            try:
                post_json(args.server, "/free", {"unload_models": True, "free_memory": True})
            except Exception:
                pass

    meta = os.path.join(args.out, "generation_log.json")
    log = json.load(open(meta)) if os.path.exists(meta) else []
    log.extend(results)
    json.dump(log, open(meta, "w", encoding="utf-8"), indent=2)
    print(f"\n{len(results)} images -> {args.out}")


if __name__ == "__main__":
    main()