{ "keyframe": { "model": "fal-ai/nano-banana-pro/edit", "input": { "aspect_ratio": "9:16", "resolution": "2K", "num_images": 1, "output_format": "jpeg" }, "fields": "prompt, image_urls (the cropped before photo first, then up to 2 more references such as a product photo), aspect_ratio, resolution (1K | 2K | 4K), num_images, output_format", "approx_cost_usd": 0.15, "alternative": { "model": "openai/gpt-image-2/edit", "fields": "prompt, image_urls, image_size { width: 864, height: 1536 }, quality high, num_images, output_format; no aspect_ratio or resolution", "approx_cost_usd": 0.27 } }, "video": { "draft": { "model": "bytedance/seedance-2.0/mini/image-to-video", "input": { "resolution": "480p", "duration": "5", "aspect_ratio": "auto", "generate_audio": false }, "fields": "prompt, image_url, resolution (480p | 720p), duration (4 to 15, as a string), aspect_ratio, generate_audio; no negative_prompt", "approx_cost_per_second_usd": 0.0721, "purpose": "Tests the after motion for 5 s at 480p before the final. It is a different model from the final, so judge whether the action reads, not the physics. It refuses a start frame with a clearly visible realistic face, so the people version skips it." }, "final": { "default": "kling", "kling": { "model": "fal-ai/kling-video/v3/pro/image-to-video", "input": { "duration": "5", "generate_audio": false }, "fields": "prompt, start_image_url (not image_url), duration (3 to 15, as a string), generate_audio, negative_prompt, cfg_scale, end_image_url, elements, multi_prompt, shot_type (keep customize); no aspect_ratio or resolution (the keyframe sets them)", "approx_cost_per_second_usd": { "silent": 0.112, "with_audio": 0.168 }, "why": "The default. It keeps the subject's shape and markings from the start frame (in the pet and ASMR samples it held them where Veo did not), accepts a realistic face for the people version, and a silent 5 s final is $0.56." }, "veo": { "model": "fal-ai/veo3.1/fast/image-to-video", "input": { "aspect_ratio": "9:16", "duration": "6s", "resolution": "720p", "generate_audio": false }, "fields": "prompt, image_url (720p or more, 16:9 or 9:16), aspect_ratio (auto | 16:9 | 9:16), duration (4s | 6s | 8s), resolution (720p | 1080p | 4k), generate_audio, negative_prompt, seed, auto_fix, safety_tolerance (1 to 6, default 4)", "approx_cost_per_second_usd": { "silent": 0.1, "with_audio": 0.15 }, "why": "A cheaper, livelier alternative for pets and rooms. It cannot make 5 s (4, 6 or 8 only), so the after part would be 6 s and the reveal 8 s. Check markings frame by frame." }, "seedance": { "model": "bytedance/seedance-2.5/image-to-video", "input": { "resolution": "720p", "duration": "5", "aspect_ratio": "auto", "generate_audio": false, "bitrate_mode": "high", "codec": "H264" }, "fields": "prompt, image_url, resolution (480p | 720p | 1080p), duration (4 to 30, as a string), aspect_ratio auto, generate_audio (same price), bitrate_mode, codec; no negative_prompt", "approx_cost_per_second_usd": 0.473, "why": "Premium option, about four times the price of Kling silent. It refuses a visible realistic face." } }, "negative_prompt": "text, subtitles, captions, watermark, logo, loading bar, progress bar, user interface, white flash, fade, cut, scene change, morphing, slow motion, extra people, extra animals, extra limbs, deformed paws, distorted face" }, "assemble": { "size": "720x1280", "fps": 24, "hold_seconds": 2, "after_seconds": 5, "flash_clear_seconds": 0.25, "flash_lead_frames": 2, "label": "loading", "panel_centre_y": 0.66, "audio": "none, by design", "beat": "the first fully white frame, at hold_seconds; written to the edit list and set as an MP4 chapter" }, "cost_note": "fal list prices checked 2026-11-13: Nano Banana Pro edit $0.15 an image; Seedance 2.0 mini 480p $0.0721 a second (5 s about $0.36); Kling 3 pro silent $0.112 a second (5 s $0.56, confirmed on the model's llms.txt). One reveal: after still $0.15 + draft $0.36 + final $0.56 = about $1.07. The people version has no draft: about $0.71. Ask before every paid step." }