> For clean Markdown of any page, append .md to the page URL. > For a complete documentation index, see https://docs.roor.ai/llms.txt. > For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.roor.ai/_mcp/server. # Video The Video tool is where the studio opens. The modes at the top of the console are **Create**, for new clips from a prompt, and **Motion & Effects**, for driving a still image with a preset camera move or effect. **Edit** and **Lipsync** appear when they are switched on for your account. ## Create ### The prompt Write the shot: subject, action, camera, light, mood. Short and concrete beats long and vague. Models respond well to film language: push-in, handheld, golden hour, shallow focus. * **Speech.** Put a line in quotes and models that make sound will speak it with matching lip movement, in English, Hindi and the other languages the model covers. Describe the voice in the prompt too: "a calm male voice with an Indian English accent". * **Elements.** Type `@` to mention a saved character, location, prop or style. The chips under the prompt show what is attached. * **What to avoid.** Under Advanced, list what should not appear. Not every model honours it. ### References Open **References** to attach images that guide the take: a face, a product, a location, a look. Pick from your **Assets**, your **Uploads**, your **Elements** or an earlier **Take**. * Each model sets how many references it accepts. The picker stops you at the limit. * On Seedance 2.0 and 2.5 you can also attach a voice or music clip (MP3 or WAV) from Uploads. The clip guides mood, pace and delivery; it does not copy the voice. On 2.0 an image reference must sit beside it; on 2.5 the clip can stand alone. * A start frame and references cannot be combined on the same take. Choose one approach. * Seedance 2.0 and 2.5 refuse uploaded photos of realistic faces, AI-made ones included; stylised characters pass. The console warns you before you generate. See [Content rules](/help/content-rules). ### Extend a take Open a finished video take and press **Extend**, or switch the console's source to **Extend** and pick the take. Most models start a fresh clip from the take's last frame; models marked **Continues a clip** carry the motion itself forward. ### The settings | Pill | What it sets | | -------- | --------------------------------------------------------------------------------------- | | Duration | Seconds, from 4 up to your plan's limit. Some models offer fixed steps, others a range. | | Aspect | 16:9, 9:16, 1:1, 4:3, 3:4 or 21:9, as the model allows. | | Quality | 480p, 720p, 1080p. 4K on the Studio plan, on models that offer it. | | Count | Render up to 4 takes from one prompt. The price multiplies. | Anything above your plan shows a lock and a link to Pricing. **Advanced** holds the seed (reroll it for a different result from the same prompt, or keep it to reproduce one), the audio switch on models that make sound, a fixed-camera option on models that support it, and **What to avoid**. On Seedance 1.5 Pro, switching audio on doubles the price; on the 2.x models sound is included. ### Models The picker shows the models available to you, each with a one-line note: **Sound**, **Speech + music**, **Clip reference** or **Silent**. What each one does: | Model | Best for | Sound and music | Speech | Your own clip as a reference | | --------------------- | --------------------------------------------------------------------- | --------------------- | ----------------------- | -------------------------------- | | Seedance 2.5 | The most capable tier. Best subject coherence and motion. Up to 30 s. | Yes | Yes, with matching lips | Yes, on its own or beside images | | Seedance 2.0 | Cinematic shots with strong coherence. 5 or 10 s. | Yes | Yes, with matching lips | Yes, beside an image reference | | Seedance 2.0 Fast | The 2.0 look, quicker and cheaper. | Yes | Yes | Yes, beside an image reference | | Seedance 2.0 Mini | The cheapest 2.0 tier for drafts. | Yes | Yes | Yes, beside an image reference | | Seedance 1.5 Pro | A strong, economical all-rounder. Sound is a priced switch. | Yes, when audio is on | Yes | No | | Seedance 1.0 Pro | Reliable silent clips at a low price. | Silent | No | No | | Seedance 1.0 Pro Fast | The cheapest way to test an idea. | Silent | No | No | Two things the sound models are not: a voice cloner (the prompt's description of the speaker decides the voice, not the clip) and a music library (the music is composed fresh each time). ### Price and Generate The **Generate** button shows the cost in credits and updates as you change the settings. Credits are charged when you press it and refunded automatically if the take fails. See [Credits and plans](/get-started/credits-and-plans). ## Motion & Effects Pick a still image, pick a preset, generate. **Motion** presets are 47 camera moves over your image: orbits, push-ins, reveals. **Effects** presets are 64 transformations. Both run silent, at 5 or 10 seconds, in 16:9 or 9:16. Presets you use often can be starred; they move to the top of the list. See [Presets](/library/presets). ## Edit Edit reworks a take you already have. Pick the **Source take**, then describe the change: "Make it night with rain on the street. Remove the person on the left. Turn it into a watercolour." The model relights, restyles, removes or reframes and renders a new take; the original is untouched. ## Lipsync Lipsync makes a photo speak. Switch the source to **Lipsync**, add one photo of the presenter, then either a voice clip of your own or a script in a built-in voice. The mouth is matched to the audio. ## The rail Every take lands in the rail beside the console, newest first. Filter it, switch between grid and list, and open any take full size for **Download**, **Favourite**, **Reuse**, **Extend**, **Upscale** and **Send to Community**. See [Takes and the rail](/library/rail). Takes are kept for your plan's retention window, from 30 days on Free to a year on Studio. Favourite the ones you want to hold on to, or download them. ## Content rules Every prompt is checked before it renders. The [Acceptable Use Policy](https://studio.roor.ai/legal/acceptable-use) lists what is refused: pornographic material, depictions of real atrocities, deepfakes of real people, instructions for causing harm, and work you have no right to use. A refused prompt costs no credits. Repeated attempts to get around a refusal lead to suspension.