#!/usr/bin/env python3 """ Gemini comparison test — Generate 10 multi-angle images for side-by-side evaluation against Qwen Image Edit 2511 + Multi-Angle LoRA. Uses GeminiEngine from batch_generate_refs.py for consistency with the existing pipeline. Usage: GOOGLE_API_KEY=xxx python3 test_gemini_comparison.py Output: - 10 angle images in gemini_test/ - gemini_test_review.html contact sheet - comparison.html (side-by-side with Qwen v2 results) """ import os import sys import json import time from pathlib import Path from datetime import datetime from dataclasses import field # Add parent dirs to path so we can import from batch_generate_refs SCRIPT_DIR = Path(__file__).resolve().parent sys.path.insert(0, str(SCRIPT_DIR)) from batch_generate_refs import GeminiEngine, GenerationJob # ── Config ────────────────────────────────────────────────────────── HERO_IMAGE = SCRIPT_DIR.parent.parent / \ "leviathan/visual/refs/characters/heroes/Jinx_Hero.jpeg" OUTPUT_DIR = SCRIPT_DIR.parent.parent / \ "leviathan/visual/lora_candidates/JINX/gemini_test" QWEN_V2_DIR = SCRIPT_DIR.parent.parent / \ "leviathan/visual/lora_candidates/JINX/qwen_test_v2" # Matching the same 10 angles from the Qwen test. # Gemini uses text prompts for angle control (no numeric params). TEST_ANGLES = [ { "name": "01_front_eye_medium", "desc": "Front, eye level, medium", "prompt_prefix": ( "Single photorealistic full-body photograph, front-facing view. " "Subject standing upright, arms relaxed at sides, neutral pose. " "Camera at eye level. Full body visible from head to feet." ), }, { "name": "02_three_quarter_right_eye_medium", "desc": "3/4 right, eye level, medium", "prompt_prefix": ( "Single photorealistic full-body photograph, three-quarter view from the right. " "Subject's body rotated 45 degrees — the left shoulder partially hidden behind the torso. " "Arms relaxed at sides, neutral standing pose. Full body visible from head to feet." ), }, { "name": "03_three_quarter_left_eye_medium", "desc": "3/4 left, eye level, medium", "prompt_prefix": ( "Single photorealistic full-body photograph, three-quarter view from the left. " "Subject's body rotated 45 degrees — the right shoulder partially hidden behind the torso. " "Arms relaxed at sides, neutral standing pose. Full body visible from head to feet." ), }, { "name": "04_profile_right_eye_medium", "desc": "Profile right, eye level, medium", "prompt_prefix": ( "Single photorealistic full-body photograph, right profile 90-degree side view. " "Subject's body turned so the right shoulder faces the camera. " "Arms relaxed at sides, neutral standing pose. Full body visible from head to feet." ), }, { "name": "05_profile_left_eye_medium", "desc": "Profile left, eye level, medium", "prompt_prefix": ( "Single photorealistic full-body photograph, left profile 90-degree side view. " "Subject's body turned so the left shoulder faces the camera. " "Arms relaxed at sides, neutral standing pose. Full body visible from head to feet." ), }, { "name": "06_back_eye_medium", "desc": "Back view, eye level, medium", "prompt_prefix": ( "Single photorealistic full-body photograph, rear view. " "Subject facing DIRECTLY away from camera — back of head visible, face NOT visible. " "Subject must NOT turn or look over shoulder. Head facing forward, away from camera. " "Arms relaxed at sides, neutral standing pose. Full body visible from head to feet." ), }, { "name": "07_front_eye_closeup", "desc": "Front, eye level, close-up", "prompt_prefix": ( "Tight headshot photograph framed from mid-chest to top of head. " "ONLY the face, neck, and upper shoulders are visible in frame. " "NO full body. NO legs. NO waist. NO hands. NO belt. NO tools. " "The face fills most of the frame. Shot on 85mm f/1.4, shallow depth of field." ), }, { "name": "08_front_low_medium", "desc": "Front, low angle, medium", "prompt_prefix": ( "Single photorealistic full-body photograph, front-facing view from a low camera angle. " "Camera positioned below eye level, looking upward at the subject. " "Subject appears powerful and imposing. Full body visible from head to feet." ), }, { "name": "09_front_elevated_wide", "desc": "Front, elevated, wide (full body)", "prompt_prefix": ( "Single photorealistic full-body photograph, front-facing view from an elevated camera angle. " "Camera positioned above eye level, looking slightly downward at the subject. " "Wide framing shows complete figure with space around the subject. " "Full body visible from head to feet." ), }, { "name": "10_three_quarter_right_eye_closeup", "desc": "3/4 right, eye level, close-up", "prompt_prefix": ( "Tight headshot photograph from three-quarter right angle, framed from mid-chest to top of head. " "Subject's face rotated 45 degrees — the far cheek partially hidden. " "ONLY the face, neck, and upper shoulders are visible in frame. " "The face fills most of the frame. Shot on 85mm f/1.4, shallow depth of field." ), }, ] # Character description for Jinx (simplified for this test) CHARACTER_DESC = ( "Young woman with sharp angular features, intense dark eyes, and choppy dark hair. " "Athletic lean build. Determined expression. " "Pure white #FFFFFF studio background. Studio lighting only." ) def build_gemini_contact_sheet(output_dir: Path, results: list): """Build HTML contact sheet for Gemini results.""" html = """ Gemini 2.5 Flash Comparison Test — JINX """ html += "

Gemini 2.5 Flash — JINX Comparison Test

\n" html += f'

Generated: {datetime.now().strftime("%Y-%m-%d %H:%M")}' html += f' | Model: gemini-2.5-flash-image | Aspect: 1:1

\n' html += '

Reference image:

\n' html += '\n' total_time = sum(r.get("time_s", 0) for r in results) html += f'
' html += f'Total: {len(results)} images | ' html += f'Time: {total_time:.1f}s ({total_time/len(results):.1f}s avg)' html += '
\n' html += '
\n' for r in results: img_name = r["filename"] html += f'
\n' html += f' \n' html += f'
\n' html += f'
{r["name"]}
\n' html += f'
{r["desc"]}
\n' html += f'
{r.get("time_s", 0):.1f}s
\n' html += f'
\n' html += f'
\n' html += '
\n' html += """ """ sheet_path = output_dir / "gemini_test_review.html" sheet_path.write_text(html) return sheet_path def build_comparison_html(qwen_dir: Path, gemini_dir: Path, output_path: Path): """Build side-by-side comparison HTML.""" html = """ Qwen vs Gemini — Side-by-Side Comparison — JINX """ html += "

Qwen vs Gemini — LoRA Training Candidate Comparison

\n" html += f'

Generated: {datetime.now().strftime("%Y-%m-%d %H:%M")} | Character: JINX

\n' html += '

Reference image:

\n' html += '\n' html += '
\n' html += ' ■ Qwen Image Edit 2511 + Multi-Angle LoRA (fal.ai, ~$0.035/img)\n' html += ' ■ Gemini 2.5 Flash (Google, token-based pricing)\n' html += '
\n' for angle in TEST_ANGLES: name = angle["name"] desc = angle["desc"] qwen_img = f"../qwen_test_v2/{name}.png" gemini_img = f"../gemini_test/{name}.png" html += f'

{desc}

\n' html += '
\n' # Qwen card html += '
\n' html += f' \n' html += f'
Qwen Multi-Angle
\n' html += f'
Numeric: h={angle.get("h", "?")}, v={angle.get("v", "?")}, z={angle.get("z", "?")}
\n' html += '
\n' # Gemini card html += '
\n' html += f' \n' html += f'
Gemini 2.5 Flash
\n' html += f'
Text prompt: {desc}
\n' html += '
\n' html += '
\n' html += """ """ output_path.write_text(html) return output_path def main(): if not os.environ.get("GOOGLE_API_KEY"): print("ERROR: GOOGLE_API_KEY not set", file=sys.stderr) print("Get a key at: https://aistudio.google.com/apikey", file=sys.stderr) sys.exit(1) if not HERO_IMAGE.exists(): print(f"ERROR: Hero image not found: {HERO_IMAGE}", file=sys.stderr) sys.exit(1) # Create output dir OUTPUT_DIR.mkdir(parents=True, exist_ok=True) print("=" * 60) print("GEMINI 2.5 FLASH — COMPARISON TEST") print("=" * 60) print(f"Hero image: {HERO_IMAGE.name}") print(f"Output dir: {OUTPUT_DIR}") print(f"Model: gemini-2.5-flash-image") print(f"Test images: {len(TEST_ANGLES)}") print(f"Rate limit: 4.5s between requests (15 RPM)") print("=" * 60) engine = GeminiEngine(model="gemini-2.5-flash-image") results = [] for i, angle in enumerate(TEST_ANGLES): print(f"\n[{i+1}/{len(TEST_ANGLES)}] {angle['desc']}") output_path = OUTPUT_DIR / f"{angle['name']}.png" # Build the full prompt full_prompt = f"{angle['prompt_prefix']}\n\n{CHARACTER_DESC}" job = GenerationJob( asset_type="character", asset_key="JINX", variant=None, angle=angle["name"], prompt=full_prompt, output_path=output_path, aspect_ratio="1:1", hero_image_path=HERO_IMAGE, ) t0 = time.time() result = engine.generate(job) elapsed = time.time() - t0 if result.success: print(f" Saved: {angle['name']}.png ({elapsed:.1f}s)") results.append({ "name": angle["name"], "desc": angle["desc"], "filename": f"{angle['name']}.png", "time_s": elapsed, }) else: print(f" FAILED: {result.error}") results.append({ "name": angle["name"], "desc": angle["desc"], "filename": "FAILED", "time_s": elapsed, "error": result.error, }) # Rate limiting (skip after last image) if i < len(TEST_ANGLES) - 1: wait = max(0, engine.rate_limit_delay - elapsed) if wait > 0: print(f" Rate limit wait: {wait:.1f}s") time.sleep(wait) # Write manifest manifest = { "test": "gemini_comparison", "model": "gemini-2.5-flash-image", "hero_image": str(HERO_IMAGE), "generated_at": datetime.now().isoformat(), "results": results, } manifest_path = OUTPUT_DIR / "test_manifest.json" manifest_path.write_text(json.dumps(manifest, indent=2)) # Build Gemini contact sheet sheet_path = build_gemini_contact_sheet(OUTPUT_DIR, results) # Build comparison HTML (only if Qwen v2 results exist) comparison_dir = OUTPUT_DIR.parent / "comparison" comparison_dir.mkdir(parents=True, exist_ok=True) comparison_path = comparison_dir / "comparison.html" if QWEN_V2_DIR.exists(): build_comparison_html(QWEN_V2_DIR, OUTPUT_DIR, comparison_path) print(f"\nComparison: {comparison_path}") else: print(f"\nNote: Qwen v2 results not found at {QWEN_V2_DIR} — skipping comparison HTML") # Summary success = [r for r in results if r["filename"] != "FAILED"] failed = [r for r in results if r["filename"] == "FAILED"] total_time = sum(r["time_s"] for r in results) print("\n" + "=" * 60) print("TEST COMPLETE") print("=" * 60) print(f"Success: {len(success)}/{len(results)}") if failed: print(f"Failed: {len(failed)}") for f in failed: print(f" - {f['name']}: {f.get('error', 'unknown')}") print(f"Time: {total_time:.1f}s total ({total_time/len(results):.1f}s avg)") print(f"\nGemini review: {sheet_path}") if comparison_path.exists(): print(f"Side-by-side: {comparison_path}") print(f"Output: {OUTPUT_DIR}") if __name__ == "__main__": main()