Kitty AI Studio — Online AI Video & Image Generator

ByteDance Seedance 2 text-to-video. Standard $1.50/5s (720p/1080p, 4K available) or Fast $0.95/5s (480p/720p). Native audio, multi-shot, up to 15s.

Animate images with Seedance 2. Standard $1.50/5s (720p/1080p, 4K) or Fast $0.95/5s (480p/720p). First+last frame, native audio, up to 15s.

Multimodal generation: up to 9 reference images (@Image1), 3 reference videos (@Video1), 3 reference audios (@Audio1), OR first+last frame. Standard $1.50/5s or Fast $0.95/5s. Native audio.

Google Omni Flash — fast multimodal video from a prompt and up to 3 reference images. 720p / 1080p / 4K, 4–10s.

ByteDance Seedance 2.5 — up to 30 seconds in a single shot, 480p or 720p, seven aspect ratios and an optional synchronised audio track. $0.20/s at 480p, $0.40/s at 720p.

Generate high-quality video with a native soundtrack from text using the LTX 2.5 22B distilled model. Optional custom audio track, two LoRA slots, 1–2 megapixel output. From $0.40.

Animate a still image with LTX 2.5. The model writes its own soundtrack, or you can upload your own audio and it lip-syncs to it. Two LoRA slots, 1–2 megapixel output. From $0.40.

Generate a smooth transition between two images with LTX 2.5, complete with a generated soundtrack — or supply your own audio. Two LoRA slots, 1–2 megapixel output.

Generate high-quality video with audio from text prompts. Multi-shot narrative, audio-video sync. 480P, 720P or 1080P, up to 30 seconds in a single pass.

Generate video from first frame, first+last frame, with audio sync or video continuation. Multi-shot narrative. 480P, 720P or 1080P, up to 30 seconds. Audio-sync and video-continuation modes run on the Wan 2.7 engine, which is the only one that supports them.

Create videos with consistent characters from reference images, video clips and audio tracks — up to 10 images, 5 clips and 5 audio tracks. Multi-character interaction, voice timbre replication. 720P or 1080P.

Create high-quality extended videos up to 30 seconds! Improved WAN 2.2 models for superior quality.

Kuaishou Kling 3.0 — HD/1080p video with native audio, multi-shot storyboarding, character consistency. 3-15 seconds.

Google DeepMind's advanced video generation. T2V, I2V, and First/Last Frame modes. $0.10-0.40/sec.

Create smooth video transitions morphing from first to last frame.

Fast image-to-video with optional LoRA (Wan 2.2) and frame interpolation for smoother motion.

Create longer videos up to 30 seconds. For better consistency, try SVI WAN 2.2 Extended Video.

Text to image on GPT Image 2.5 (Sunburst) — realistic lighting, crisp typography, great for posters and product shots. Two quality tiers above High, transparent backgrounds for logos and cut-outs, PNG / JPEG / WebP output, 10 aspect ratios up to 4K.

Ideogram 4.0 open-weights text-to-image model with best-in-class TEXT RENDERING — ideal for posters, logos, typography, and memes. Understands structured JSON prompts for precise control. Two custom LoRA slots. Single quality mode, $0.20 per image. Up to 4 images per generation.

Krea 2 — Krea's aesthetic-control image model. Generate from text (Turbo) or use the Identity Edit tab to repose/restyle a person while keeping their identity. Preinstalled style LoRA dropdown plus custom LoRA slots. Open source.

Open-source Boogu base generation refined by Z-Image Turbo, then upscaled with the NVIDIA PiD pixel-diffusion upscaler. Every run returns three images — base, refined, and PiD upscaled. Orientation select + denoise control. Locked at 1024 base.

Krea 2 with the GonzoLomo v3.0 finetune — raw analog photography, lomo colour, film imperfection. Returns the base render and a PiD 4× upscale, same as Image Turbo. One built-in style LoRA slot plus one custom slot. Open source.

Two sampling passes on a base Krea 2 checkpoint for maximum photographic detail, up to 1920px. Five base models to choose from, one built-in style LoRA plus two custom slots, and a negative prompt on the front. $0.24 per image.

Upload one photo of your character and get a complete LoRA training dataset: angles, expressions, full body and location shots, generated with Nano Banana 2 or GPT Image 2. Auto-captioned and delivered as a ready-to-train ZIP.

Gemini 3.1 Flash Image — pro-level visual intelligence with Flash-speed efficiency. Edit with up to 14 reference images or generate from text.

Generate and edit images with Wan 2.7. Up to 9 input images for editing, fusion, style transfer, and more. Standard model, up to 2K.

Generate images with Instagram-perfect aesthetic. Optimized for portraits with the "Instagirl" style.

Generate highly realistic images with optimized settings and special realism-focused LoRA.

Single-pass Z-Image Base generation with NVIDIA PiD pixel-diffusion upscaler — faster and sharper than two-stage. Always returns both base and upscaled outputs (multiplier ×1–×4, max base resolution 1024). Up to 3 custom LoRAs. Full control over denoise, steps, and CFG.

Generate images from a reference image with NVIDIA PiD pixel-diffusion upscaler. AI analyzes the reference and creates the prompt automatically, then generates and upscales in a single pass — both images returned (multiplier ×1–×4, max base resolution 1024). Up to 3 custom LoRAs. Full control over denoise, steps, and CFG.

Generate high-quality images from text prompts with two optional custom LoRA slots (WAN 2.2 High/Low Noise).

Ultra-fast image generation with NVIDIA PiD pixel-diffusion upscaler always on. Every generation returns two images: base and upscaled (multiplier ×1–×4). Max base resolution 1024. Results in seconds!

Two-stage Z-Image. Get both base and turbo-refined outputs with LoRA presets and up to 3 custom LoRAs.

Generate 8 different camera angles (close-up, wide, 45°, 90°, aerial, low angle) from a single character image using Qwen AI.

Generate different camera angles of your image using interactive 3D controls. Adjust horizontal angle (0-360°), vertical angle (-30° to 60°), and zoom level.

Generate 6 different camera angles of your image at once. Configure each angle with horizontal, vertical, and zoom controls for comprehensive character sheets or product views.

Image editing on GPT Image 2.5 (Sunburst) — natural-language instructions with up to 16 reference images. Paint a mask and it edits only what you painted, leaving the rest of the picture untouched. Transparent backgrounds for logos and cut-outs, PNG / JPEG / WebP output, and two extra quality tiers above High. Crisp typography, photorealistic composites, 10 aspect ratios.

Open-source Boogu multi-image editor — upload up to 4 photos and blend, fuse or restyle them with a text instruction. Reference your uploads in the prompt as @image1–@image4. Inputs auto-resized to 1280px.

Open-source face inpainting with Z-Image Turbo. Automatically detects the face and repaints only that region behind a feathered mask for a seamless blend — no visible box. Denoise control plus the same Z-Image Turbo LoRA presets and a custom LoRA slot. Locked at 1280px.

Edit videos with text instructions. Object replacement, style transfer. Standard $1.50/5s or Fast $0.95/5s. Use @image1 in prompt to reference uploaded images.

Say what to mask in plain words — "blonde hair", "the red car" — and SAM3 finds it, then Krea 2 repaints only that area. No brush, no manual masking. One custom LoRA slot. $0.22 per image.

Edit videos with text instructions. Style transfer, object replacement, scene changes. Optional reference images. 480P, 720P or 1080P, up to 30 seconds.

Professional image editing with Wan 2.7 Pro. Thinking mode for better composition, 4K support, up to 9 input images.

Edit images with text instructions using Qwen AI model.

Change clothes on people in images with consistent LoRA style.

Open-source text-guided image editing with state-of-the-art identity consistency. Upload 1-3 reference images and describe the edit. Supports clothing changes, style transfer, makeup, photo restoration, virtual try-on, and more. 20B parameter model by Xiaohongshu/RedNote.

Paint over areas you want to change, then describe what should replace them. Perfect for object removal, replacement, or adding new elements.

Remove background from any video using AI matting. Outputs green screen video with clean edges.

Give it a photo and it starts talking. Add your own voice track and the mouth follows your words exactly — your audio is written into the video untouched, not imitated. Leave the audio out and the model writes its own soundtrack. The output keeps your photo's shape at the resolution the model was trained on, so nothing gets letterboxed or stretched. Up to 15 seconds, realism LoRA on by default, two slots for your own.

Generate talking head videos from a face image and audio. Max 7 min audio, 1024px image. Powered by Wan 2.1 InfiniteTalk.

SCAIL-2 (Wan 2.1 14B) — state-of-the-art pose-driven character animation. Upload one character image and a driving performance video; SCAIL-2 transfers full-body motion, hands, and facial expression onto your character with rock-solid identity — now loopable up to 30 seconds. $0.08/second.

Transfer motion from reference video to character image. Dance, choreography, character animation.

Straightforward ESRGAN upscaling — no prompt, no re-generation, nothing invented. The picture stays exactly as it was, only bigger and cleaner. Pick 2× for a safe enlargement or one of the 4× models for posters and prints. $0.04 per image.

Feed in a clip and LTX 2.5 re-renders it sharper: a light refinement pass at the original content, a temporal upscale that doubles the frame rate through the sampler for smoother motion, then RTX super-resolution. Keeps the original audio. From $0.44 — same duration tiers as LTX 2.5 Image to Video plus 4c.

Enhance videos up to 30 seconds with smart batch processing and seamless frame blending.

Upscale images to 4K resolution using SeedVR2 model.

Quick image upscaling with SeedVR2 for everyday use.

Enhance and upscale images with optional custom LoRA for style control.

Enhance video quality. Upscale resolution and boost details frame by frame.

Upscale videos to HD resolution using SeedVR2 model.

Add authentic film grain texture to your images. Adjust intensity and saturation for vintage look.

Auto-detect and double your video frame rate using RIFE AI interpolation. Smoother motion!

High-definition magnification trained on Qwen-Image-Edit-2511. Losslessly enlarges images to approximately 2K size. Add your own LoRA for custom styles.

FlashVSR-powered video detail restoration. Restores hair, skin, textures while preserving face identity. Optional 2x upscale.

NVIDIA RTX Video & Image AI Upscaler — powered by RTX Video Super Resolution. Upscale videos up to 4x and images to ultra-high resolution.

Open-source Anima model for high-quality anime art. Base mode features aesthetic-highres LoRA control with optional DBZ Style and Pixel Art LoRAs plus a custom LoRA slot. Real mode runs Anima base with a Z-Image photorealistic refine pass for stunning semi-real renders. Base mode $0.15/generation, Real mode $0.18.