Workflow: create 8-pose character turnarounds in Leonardo AI

Short answer
A repeatable pipeline: create one clean reference (face + outfit) → use image-to-image (low strength) + fixed seed to rotate the character in 8 angles (0°, 45°, 90°, … 315°) → upscale and touch up consistently → batch export with standardized filenames. Below are concrete prompts, settings, and a practical checklist.

Why this works
Using a single reference image and a fixed seed keeps facial features and clothing consistent. Image-to-image keeps identity while changing pose/angle when you keep strength low and only change the angle descriptor. Upscale last to preserve details and avoid generating artifacts during composition steps.

Recommended pipeline (step‑by‑step)
1) Prepare a reference sheet: front-facing, neutral expression, solid background, full body + closeup of face. Save at 1024+ px.
2) Base prompt template (use for img2img):
"[CharacterName], full body, neutral expression, consistent clothing: [brief clothing descriptors], face: [eye color, nose shape, jawline, scars/tattoos], camera: 50mm, full-body frame, soft 3-point lighting, photorealistic / stylized [choose], angle: [ANGLE], neutral pose. High detail, consistent proportions, no extra limbs, no text."
Negative prompt: "blurry, deformed hands, extra fingers, watermark, text, mutated anatomy, lowres."
3) Img2Img settings: strength 0.25–0.45 (lower preserves identity), exact same seed for all 8 renders, sampling steps 20–30, sampler: DDIM/PLMS (or default Leonardo sampler), CFG/scale 6–9.
4) Generate each angle by replacing [ANGLE] with: 0° (front), 45°R, 90°R (three-quarter), 135°R, 180° (back), 225°R, 270°R (left three-quarter), 315°R. Keep camera, lighting, clothing tokens identical.
5) If limbs or perspective break, use targeted inpainting on those areas (keep seed) to repair without changing face.
6) Upscale: use Leonardo’s built-in upscaler at the end or export the 1× image and use Topaz/Gigapixel/ESRGAN for a cleaner 2×–4× upscale. Apply slight sharpening only after upscaling.
7) Batch export: name files with a standard convention: CharacterName_seed_angle_v1.png (e.g., Mara_12345_045.png). Export PNG/TIFF. Keep a JSON or CSV log with seed + prompt used per render for reproducibility.

Decision criteria
- Budget: built-in upscaler is fastest/free on platform; external paid upscalers give cleaner detail for print.
- Skill level: beginners should stick to one reference + low img2img strength. Advanced users can generate separate body pose variations and composite the consistent face via inpainting.
- Output goal: for game/animation you may need a 3D turnaround—not suitable to rely on AI-only if you need rigging.

Best-for / Avoid-if
- Best for concept art, character sheets, pitch visuals, and 2D turnarounds.
- Avoid if you need a riggable 3D model, exact photogrammetry, or frame-perfect animation — use a 3D pipeline instead.

Practical checklist
- [ ] Create 1 clean face + full-body reference (1024+ px)
- [ ] Lock seed and character tokens
- [ ] Use img2img strength 0.25–0.45; generate eight angles
- [ ] Repair with inpainting if needed
- [ ] Upscale (built-in or external) and sharpen lightly
- [ ] Export with standardized filenames and log prompts/seeds

If you want, I can convert the base template into 8 ready prompts or set of seeds to paste into Leonardo AI. Recommendation: try one pass with the built-in upscaler, review consistency, then do a second pass with an external upscaler only if you need higher-res prints.

Compare Leonardo AI and Midjourney