revid ai media types and models.md
Revid AI Media Types, AI Models, and Style Presets
How visuals work in Revid.ai: media types (stock, AI images, AI video, your own), model tiers including Google Veo3 and OpenAI Sora 2, 40+ style presets, and consistent AI characters.
Media types
Chosen via "Choose media type" in the app, or media.type in the API:
| Type | API value | Description |
|---|---|---|
| Stock videos | stock-video |
Professional pre-recorded footage matched to your content |
| Moving AI images | moving-image |
AI images with motion (media.animation: none / soft / dynamic / depth) |
| AI video | ai-video |
Fully AI-generated clips per scene |
| Motion graphics | motion-graphics |
AI-directed animated motion graphics scenes (explainers, data stories) |
| Your own media | custom |
Upload media; AI auto-places it (or use it exclusively) |
| Static / Gameplay | — | Static backgrounds or gameplay footage (Subway Surfers, Minecraft...) |
Web-content tools can also use scraped media from the source page, optionally mixed with stock.
AI video model tiers
Credits are per 5 seconds of generated video, billed pro rata per second.
| Model | Credits/5s | What you get |
|---|---|---|
| Low | 15 | Cost-effective motion |
| Pro | 50 | Better quality and prompt adherence |
| Ultra | 100 | Seedance 2.5 & MiniMax H3 at 720p (1080p: 250, native 4K: 500) |
| Veo3 (Google) | 100 | ~8s cinematic clips — ASMR, street interviews, product demos |
| Sora 2 (OpenAI) | 130 | 4/8/12s clips with accurate physics and synchronized audio |
API: media.videoModel: base | pro | ultra | veo3 | sora2. Veo3/Sora2 are disabled by default on most tools and must be explicitly selected where supported.
For media.type: motion-graphics the video model picks the scene director instead: base (fast director, 30 credits/scene) or pro (premium director with better composition and coherence, 60 credits/scene). Higher models (ultra/veo3/sora2) are automatically mapped to pro.
AI image model tiers
| Model | Credits/image | API value |
|---|---|---|
| Low | 1 | cheap |
| Pro | 4 | good |
| Ultra | 16 | ultra |
Or set media.quality (standard/pro/ultra) as a convenience profile covering both image and video models.
Style presets (40+)
Presets restyle all generated visuals. Highlights: DEFAULT (photorealistic), GHIBLI, PIXAR, ANIME, REALISM / ULTRA REALISM, FLAT ANIMATION, SKETCH_COLOR / SKETCH_BW, INK (Japanese ink wash), 3D RENDER, LOW POLY, LEGO, MINECRAFT, NINTENDO, SCIFI, CLAYMATION, FANTASY, RETRO CARTOON, Spider-verse, Kids Book, New Yorker Cartoon, Renaissance Fresco, and more. Underlying models include Flux, Seedream, GPT-Image, and Recraft depending on tier.
API: media.mediaPreset. Discover slugs with revid presets --json or revid presets --search cinematic --json.
Guidance and consistency
- Mood: a free-text "Specify a general mood" field (e.g. "dark mood") keeps all visuals tonally consistent.
- Per-scene direction:
[bracket notes]in the script steer individual scenes. - Consistent characters: train custom characters (LoRAs, 200 credits) or define described avatars, then select up to 5–6 per video. Characters persist across videos for brand consistency. API:
characterIds[]; manage viaGET/POST/DELETE https://www.revid.ai/api/public/v3/consistent-charactersor thelist_characters/create_character/delete_characterMCP tools. - Style transfer: transform your uploaded images into the selected preset style (4 credits/image, images only, subscription required).
- Turn images into videos: animate each uploaded image with AI motion (choose base/pro/ultra). Applies when your media is placed in the video (custom / real-photo media); with an AI media type your uploads serve as references for the generated shots instead.
Your own media, precisely controlled
API fields under media: provided[] (your media items), useOnlyProvided (no AI/stock additions), useProvidedInOrder (map provided items to script sections in order), backgroundVideo, mergeVideos, addAudioToVideos, turnImagesIntoVideos, `applyStyleTransfer.
Got questions? Quick answers
Which media type should I pick?
Stock for realistic b-roll on a budget; moving images for stylized narration (best cost/quality ratio); AI video for full generative motion; Veo3/Sora2 for hero clips where realism matters.
How many clips does a video use?
Roughly one per 4 seconds; control with media density (low/medium/high) or media.maxItems.
Can I regenerate a single scene later?
Yes — in the editor, "Regenerate from prompt" re-creates the media of any slide. See https://www.revid.ai/llm/revid-ai-video-editor-guide.md