fal.ai Models Reference

Image-to-Video

Kling O3 Standard (RECOMMENDED)

• Model: fal-ai/kling-video/o3/standard/image-to-video
• Docs: https://fal.ai/models/fal-ai/kling-video/o3/standard/image-to-video/llms.txt
• Price: $0.084/s (no audio), $0.112/s (with audio)
• Duration: "3" to "15" (string)
• Params: image_url, prompt, duration, generate_audio, end_image_url, negative_prompt, enable_safety_checker
• Best balance of quality and price

Kling v3 Pro

• Model: fal-ai/kling-video/v3/pro/image-to-video
• Docs: https://fal.ai/models/fal-ai/kling-video/v3/pro/image-to-video/llms.txt
• Price: $0.112/s (no audio), $0.168/s (audio), $0.196/s (audio+voice)
• Duration: "3" to "15" (string)
• Params: same as O3 plus elements, negative_prompt, cfg_scale (0-1)
• Supports elements for character/video references (max 8s with elements)

Kling v2.5 Turbo Pro (cheapest)

• Model: fal-ai/kling-video/v2.5-turbo/pro/image-to-video
• Docs: https://fal.ai/models/fal-ai/kling-video/v2.5-turbo/pro/image-to-video/llms.txt
• Price: $0.07/s
• Fastest generation, lowest cost

Seedance 1.5 Pro (ByteDance)

• Model: fal-ai/bytedance/seedance/v1.5/pro/image-to-video
• Docs: https://fal.ai/models/fal-ai/bytedance/seedance/v1.5/pro/image-to-video/llms.txt
• Price: ~$0.26 per 5s 720p with audio
• Duration: STRING "4" to "12" (NOT integer!)
• Resolution: "480p", "720p", "1080p"
• Aspect ratio: "21:9", "16:9", "4:3", "1:1", "3:4", "9:16", "auto"
• Generates audio by default, supports end_image_url

Motion Control

Kling v3 Pro Motion Control

• Model: fal-ai/kling-video/v3/pro/motion-control
• Docs: https://fal.ai/models/fal-ai/kling-video/v3/pro/motion-control/llms.txt
• Price: $0.126/s
• Required: image_url, video_url, character_orientation ("image" or "video")
• Optional: prompt, keep_original_sound, elements
• video orientation: up to 30s, better for complex motion
• image orientation: up to 10s, better for preserving framing

Kling v3 Standard Motion Control

• Model: fal-ai/kling-video/v3/standard/motion-control
• Docs: https://fal.ai/models/fal-ai/kling-video/v3/standard/motion-control/llms.txt
• Same params, lower cost

Speaking Avatar

HeyGen Avatar4

• Model: fal-ai/heygen/avatar4/image-to-video
• Docs: https://fal.ai/models/fal-ai/heygen/avatar4/image-to-video/llms.txt
• Price: $0.10/s
• Params: image_url, prompt (speech text), talking_style ("stable" / "expressive"), resolution, aspect_ratio, voice, audio_url
• Resolution: "360p" / "480p" / "540p" / "720p" / "1080p"
• Aspect ratio: "16:9" / "9:16" / "1:1"

Image Edit / Upscale

Seedream v5 Lite

• Model: fal-ai/bytedance/seedream/v5/lite/edit
• Docs: https://fal.ai/models/fal-ai/bytedance/seedream/v5/lite/edit/llms.txt
• Price: $0.035/image
• Default output: auto_2K
• Params: image_urls (list, up to 10), prompt, image_size, enable_safety_checker

Seedream v4.5

• Model: fal-ai/bytedance/seedream/v4.5/edit
• Docs: https://fal.ai/models/fal-ai/bytedance/seedream/v4.5/edit/llms.txt
• Price: $0.04/image
• Output: up to 4096x4096
• Multi-image input: can reference multiple images (Figure 1, Figure 2, etc.)
• Great for replacing characters while preserving pose

xAI Grok Imagine Video

Grok Imagine Text to Video

• Model: xai/grok-imagine-video/text-to-video
• Docs: https://fal.ai/models/xai/grok-imagine-video/text-to-video/api
• Params: prompt, duration, aspect_ratio, resolution
• Resolution: 480p, 720p
• Aspect ratio: 16:9, 4:3, 3:2, 1:1, 2:3, 3:4, 9:16
• No documented safety-checker toggle in the fal schema

Grok Imagine Image to Video

• Model: xai/grok-imagine-video/image-to-video
• Docs: https://fal.ai/models/xai/grok-imagine-video/image-to-video/api
• Params: prompt, image_url, duration, aspect_ratio, resolution
• Good fit when you want to pull the first frame from a reference video and animate from that still
• No documented generate_audio, negative_prompt, end_image_url, or safety-checker toggle in the fal schema

Grok Imagine Edit Video

• Model: xai/grok-imagine-video/edit-video
• Docs: https://fal.ai/models/xai/grok-imagine-video/edit-video/api
• Params: prompt, video_url, resolution
• Input video is resized to a maximum area of 854x480 and truncated to 8 seconds according to the fal schema
• Best fit for transforming an existing clip; not a dedicated "extend video timeline" endpoint

Content Policy

• enable_safety_checker: False disables client-side filter only
• Server-side content filter remains active on all models
• For Grok Imagine Video on fal.ai, the public schema does not currently expose enable_safety_checker, so there is nothing explicit to disable from the client side
• Blocked words in suggestive context: explicit anatomy, "squeezes", "revealing", "sensual"
• Safe alternatives: "decollete", "fashion editorial", "adjusts", "touches", "graceful movements"

Orientation Guide (Motion Control)

• "video": dancing, full-body motion, complex movement, up to 30s
• "image": preserve still image framing/pose, up to 10s
• Match the source image pose to the first frame of the video for best results