Google Gemini vs Runway

Why tested: A strong reference frame gives later Runway image-to-video tests clear identity, object and background markers. This task measures whether Runway can create that production-ready starting image from a detailed prompt.

Test setup: A new image generation was started in the Runway desktop web application. No image or style reference was uploaded. The prompt was entered as text, the aspect ratio was set to 16:9, Nano Banana Pro was selected, Quality optimization was active and the 20-credit job was manually approved. The first generated image was downloaded and evaluated without regeneration, inpainting, cropping or retouching.

Input used: Text-only input. The prompt requested a woman in a red coat beside a yellow vintage car on a European street, with a blue bicycle, three terracotta plant pots and a café with green window frames as consistency markers. The downloaded output measured 2048 × 1143 pixels.

Prompt / instruction: Create a photorealistic cinematic image of a woman around 30 years old standing beside a yellow vintage car on a quiet European street. She has shoulder-length dark brown hair, a clearly visible face and is wearing a long red coat, black trousers and white sneakers. Her full body and both hands must be visible. One hand rests naturally on the car door. The street contains a blue bicycle leaning against a brick wall, three small plants in identical terracotta pots and a café with green window frames in the background. Use soft overcast daylight, realistic proportions, natural skin texture and sharp details. Camera: eye-level medium-wide shot, 35mm lens, subject centered slightly to the left. No motion blur, no text, no logos, no extra people, no distorted hands, no cropped limbs.

Evaluation criteria: Prompt adherence; facial and hand anatomy; full-body visibility; correct clothing and object placement; one bicycle and three plant pots; realistic materials and lighting; composition; absence of extra people; absence of cropped limbs; and compliance with the no-text and no-logo instruction.

Result: The first-pass Runway image is highly usable as a reference frame. The woman, red coat, black trousers, white sneakers, yellow vintage car, green café windows, blue bicycle and three plant pots are all present. Her full body and both hands are visible, and the hands, face, car geometry, lighting and street perspective look convincing. The medium-wide eye-level composition also follows the request well. The main failures are strict rather than structural: a Mercedes emblem is clearly visible despite the no-logo instruction, and the café boards and bicycle contain small unreadable or brand-like lettering despite the no-text instruction. The hand rests on the mirror and door area rather than clearly on the door itself, and the three pots are similar but not perfectly identical. Overall, the image is a strong first result, but it would need a small cleanup pass before publication under the original no-text and no-logo requirements.

Sources and notes: ToolLions editorial Runway test evidence from 2026-07-19: prompt and interface screenshot, Quality-setting screenshot, approval panel and original downloaded output. Add the relevant official Runway documentation URL before publication.