> ## Documentation Index > Fetch the complete documentation index at: https://docs.unifically.com/llms.txt > Use this file to discover all available pages before exploring further. # SeeDance 2.0 Pro RFR > Highest-quality video generation — real-face restricted Generate videos using ByteDance SeeDance 2.0 Pro RFR — the highest quality tier. Supports text-to-video, first/last frame, and multimodal reference workflows with image, video, and audio references. Output durations are 4–15 seconds at resolutions up to 4K (10-bit H.265). **RFR — Real Face Restriction.** This model rejects reference images and videos that contain real human faces, unless the likeness is authorized on the provider side. Use the non-RFR [SeeDance 2.0 Pro](/models/video/bytedance/seedance-2.0-pro) if your references include real people. ## Model ``` bytedance/seedance-2.0-pro-rfr ``` ## Request types | Request type | Required inputs | Optional inputs | Notes | | -------------------- | ---------------------------------------------------------------- | ------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------ | | Text-to-video | `prompt` | `ratio`, `duration`, `resolution`, `generate_audio` | No frame or reference URLs | | First/last frame | `image_urls` (1 or 2 entries) | `prompt`, `ratio`, `duration`, `resolution`, `generate_audio` | 1 URL = first frame; 2 URLs = first + last frame. `prompt` optional | | Multimodal reference | At least one of `reference_image_urls` or `reference_video_urls` | `prompt`, `reference_audio_urls`, `ratio`, `duration`, `resolution`, `generate_audio` | Generate, edit, or extend video from reference media. Mutually exclusive with `image_urls` | ## Parameters | Parameter | Type | Required | Default | Description | | ---------------------- | --------- | ----------- | ------------ | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | `prompt` | string | Conditional | — | Text prompt (max 6,000 chars; ≤500 Chinese chars or ≤1,000 English words recommended). **Required** when no frame images or reference media are provided. Supports Chinese, English, Japanese, Indonesian, Spanish, Portuguese. Dialogue in double quotes drives generated speech | | `image_urls` | string\[] | No | `null` | Frame image URLs. 1 = first frame; 2 = first + last frame. Formats: jpeg, png, webp, bmp, tiff, gif, heic, heif. Max 30 MB each. Sides: 300–6000 px. Aspect ratio: 0.4–2.5. Mutually exclusive with `reference_*` fields | | `reference_image_urls` | string\[] | No | `null` | Reference image URLs (max 9) for multimodal-reference generation. Same format/size constraints as `image_urls`. Must not contain real human faces | | `reference_video_urls` | string\[] | No | `null` | Reference video URLs (max 3) — enables video editing and extension. Formats: mp4, mov. Max 200 MB each. Duration: 2–15 s per clip, combined total ≤ 15 s. Must be publicly reachable URLs. Must not contain real human faces | | `reference_audio_urls` | string\[] | No | `null` | Reference audio URLs (max 3). Formats: wav, mp3. Max 15 MB each. Duration: 2–15 s per clip, combined total ≤ 15 s. Cannot be the only input — include at least one reference image or video. Must be publicly reachable URLs | | `ratio` | string | No | `"adaptive"` | `adaptive`, `1:1`, `3:4`, `4:3`, `9:16`, `16:9`, `21:9`. `adaptive` picks the closest ratio from the prompt or input media | | `duration` | integer | No | `5` | Output duration in whole seconds (4–15), or `-1` to let the model choose the length (billed by actual duration) | | `resolution` | string | No | `"720p"` | `480p`, `720p`, `1080p`, `4k`. 4K output is 10-bit H.265 with tight upstream capacity — expect queueing | | `generate_audio` | boolean | No | `true` | Generate synchronized audio (voices, sound effects, background music). Output audio is mono | This model does not accept `seed` or `camera_fixed`. Output is mp4 at 24 fps. ## Example - Text-to-Video ```bash theme={null} curl -X POST https://api.unifically.com/v1/tasks \ -H "Content-Type: application/json" \ -H "Authorization: Bearer YOUR_API_KEY" \ -d '{ "model": "bytedance/seedance-2.0-pro-rfr", "input": { "prompt": "Cinematic slow-motion shot of a dancer performing on a rain-soaked stage", "ratio": "16:9", "duration": 10, "resolution": "1080p" } }' ``` ## Example - First & Last Frame ```bash theme={null} curl -X POST https://api.unifically.com/v1/tasks \ -H "Content-Type: application/json" \ -H "Authorization: Bearer YOUR_API_KEY" \ -d '{ "model": "bytedance/seedance-2.0-pro-rfr", "input": { "prompt": "The car accelerates and drives away into the sunset", "duration": 12, "image_urls": [ "https://example.com/start.jpg", "https://example.com/end.jpg" ] } }' ``` ## Example - Multimodal Reference ```bash theme={null} curl -X POST https://api.unifically.com/v1/tasks \ -H "Content-Type: application/json" \ -H "Authorization: Bearer YOUR_API_KEY" \ -d '{ "model": "bytedance/seedance-2.0-pro-rfr", "input": { "prompt": "Bring the illustrated character to life in the style of the reference clip", "duration": 8, "reference_image_urls": [ "https://example.com/character.jpg" ], "reference_video_urls": [ "https://example.com/animation-style.mp4" ] } }' ``` Frame `image_urls` are fetched server-side, so they only need to be reachable by Unifically. Reference **video/audio** URLs are forwarded to the provider directly and must be publicly downloadable. ## Response ```json theme={null} { "code": 200, "success": true, "data": { "task_id": "abc123def456", "status": "processing" } } ```