Kitty AI Studio is the pay-per-use AI studio on druidcat.com. You pick a tool, upload a picture or type a prompt, and pay for that one job from your credit balance. 1 credit is $0.01 and there is no subscription. Unlike services like Runpod that require technical knowledge, everything runs in your browser with nothing to install.
In November 2026 the studio was rebuilt: a new hub, eight new tools, a faster LTX 2.5, batches of up to 16 pictures and new prices. This guide covers all of it, with the price and typical time of every tool.
Kitty AI Studio at a Glance
- More than 50 tools for video, images, editing, talking avatars, upscaling and audio.
- Two kinds of models: open models on our own GPUs, and partner models such as Veo 3.1, Kling 3.0 and Seedance.
- Pay per job: the tool page shows the exact price before you start. New accounts get a $0.20 welcome credit, valid for 7 days.
- Batches: seven image tools make up to 16 pictures in one run (MOSS Image up to 12), and each extra picture costs less.
- 24-hour storage: inputs and results are deleted from our servers after 24 hours, so download what you want to keep.
- A free desktop editor: the Kitty app puts the same models inside a video editor with an AI agent.
The New Hub
Everything starts at druidcat.com/kitty-ai-studio. The hub has:
- Search: type “upscale”, “lip sync” or a model name.
- Category chips: Video, Images, Edit & Swap, Talking & Motion, Enhance & Upscale and Audio.
- A filter to show only “Our GPUs” or only “Partner models”.
- Light and dark mode.
- A short preview on every card, with the price and the typical time.
Click a card to open the tool. Each tool page has its own inputs and shows the price for your settings on the button before you generate.
Our GPUs vs Partner Models
Our GPUs. Open models that we run ourselves on serverless GPUs of the RTX 5090 class: LTX 2.5, the H3 tools from MiniMax, Z-Image, Qwen Image 2.1, Krea 2, MOSS Image, Anima, SCAIL-2, InfiniteTalk, SeedVR2, the other upscalers and the audio tools. Many of them accept your own LoRA files.
Partner models. Veo 3.1 (Google), Kling 3.0 (Kuaishou), Seedance 2.0 and 2.5 (ByteDance), Wan 3.0 and Wan 2.7 (Alibaba), GPT Image 2.5, Nano Banana 2.1 (Google) and OmniHuman 2.0 run through the partners’ APIs at their usual prices.
One thing to know about our GPUs: they are serverless, so the first job after a quiet period can take a few minutes longer while a GPU starts. The times in this guide are typical times on a warm GPU, not promises.
Credits and Prices
- 1 credit = $0.01. Top-ups start at $5.
- Welcome credit: $0.20 on signup, valid for 7 days. That covers a first picture, for example one Qwen Image 2.1 or Krea 2 Turbo image.
- Paid credits are valid for about 12 months. No monthly fee.
- How a price is built: a fixed part per job plus a part per picture or per second, rounded up to a whole cent. The tool page shows the exact total before you pay.
| Example job | Price |
|---|---|
| One Qwen Image 2.1 picture | $0.14 |
| 16 Z-Image Turbo pictures in one run | $0.50 |
| 5-second LTX 2.5 clip with sound (1 MP) | $0.35 |
| 4-second H3 Reference to Video clip | $0.47 |
| 5-second MiniMax H3 talking video | $0.61 |
| Picture upscaled to 4K with SeedVR2 | $0.36 |
| Song split into dialogue, music and effects | $0.13 |
Batches: Up to 16 Pictures per Run
Seven image tools make several pictures in one run. The fixed part of the price is paid once per run, so the price per picture drops as the batch grows:
| Tool | 1 picture | Biggest batch | Per picture |
|---|---|---|---|
| Z-Image Turbo | $0.17 | 16 for $0.50 | $0.031 |
| Krea 2 Turbo | $0.15 | 16 for $0.66 | $0.041 |
| Krea 2 GonzoLomo | $0.27 | 16 for $0.54 | $0.034 |
| Krea 2 Ultra Realistic | $0.37 | 16 for $4.27 | $0.267 |
| Z-Image Detailed | $0.31 | 16 for $3.31 | $0.207 |
| MOSS Image | $0.26 | 12 for $1.91 | $0.159 |
| Film Grain | $0.11 | 16 for $0.20 | $0.013 |
Qwen Image 2.1, Anima and the edit and upscale tools make one picture per run.
New in November 2026
Eight new tools
- Qwen Person Swap ($0.34): puts the person from one photo into another photo. You give it the scene and one character photo, optionally a second character, choose “body” or “head” mode, and can switch on fast mode or add a prompt. About 2 minutes.
- Image Layers ($0.38): splits one picture into separate layers, for example background, subject and objects, so you can edit or animate them one by one. Input: the picture and a short description. About 2.5 minutes.
- SolarRenderer ($0.93 per clip of about 4.5 seconds): a grey 3D blockout or structure clip plus one picture of the look you want becomes a finished cinematic shot that follows the blockout’s camera and layout. About 7 minutes.
- H3 Reference to Video ($0.47 for the default 4 seconds): one or two reference pictures and a prompt become a 2 to 10 second clip. Optional looks: CinemaGrade (a graded film look) and Commercial 80s (a retro TV-commercial look). About 3 to 4 minutes.
- H3 Extend ($2.30 per extension of about 4.5 seconds): continues an existing clip from its last frames in the same style. About 20 minutes.
- H3 Long Take ($1.42 for an 8-second shot): one continuous shot guided by two reference pictures, with an optional end frame. About 11 minutes.
- Stem Splitter ($0.13 per file): one audio file in, three files out: dialogue, music and effects. Usually done in under a minute.
- Room Acoustics ($0.10 per file): places a dry voice or sound in a realistic room. You get the “wet” audio plus the room’s impulse response, and advanced users can pass their own room recipe. Takes seconds.
Updated tools
- H3 Character Swap swaps the person in a video. New: you choose where in the clip the 5 to 15 second window starts, and you can switch on an automatic anchor frame (+$0.15). The studio then first makes a clean edited first frame with Qwen Image 2.1, which keeps the new character more consistent. You can also upload your own anchor frame.
- LTX 2.5 (text to video, image to video, first and last frame) is faster: 6 sampling steps and hardware video encoding. The LTX 2.5 Upscaler goes up to 2.4 MP. Read the full LTX 2.5 guide.
- Z-Image Turbo, Z-Image Detailed and Krea 2 GonzoLomo: the 4x “PiD” upscaled copy is now an optional switch, off by default, +$0.02 per picture.
- Z-Image Detailed accepts an optional reference photo to match its look. It replaced the separate Z-Image from Reference tool.
- MiniMax H3 Talking Video: the style dropdown moved to H3 Reference to Video.
- Batches go up to 16 pictures per run (see above).
Retired tools and what replaced them
Some older tools are gone. LTX 2.3 is replaced by LTX 2.5, WAN Animate and Wan SCAIL by SCAIL-2, WAN 2.6 by Wan 3.0 and LTX 2.5, Nano Banana Pro by Nano Banana 2.1, Ideogram 4 by GPT Image 2.5 (for text in images), MagiHuman by MiniMax H3 Talking Video, InfiniteTalk and OmniHuman 2.0, and the old Qwen edit tools by Qwen Image 2.1, Qwen Image 2.1 Inpainting and Krea 2 Inpainting. Instagirl, SVI WAN 2.2, the Image and Video Enhancer and a few other older tools were retired too. The Music Video Creator page is retired as well: the free Kitty app replaces it.
All Tools by Category
Prices are for one job. Times for our GPUs are typical times on a warm GPU. For partner models the time is the range shown on the card.
Video
| Tool | What it does | Price | Typical time |
|---|---|---|---|
| LTX 2.5 Text to Video | Prompt to clip with its own soundtrack, up to 20 s, 1 or 2 MP | $0.35 for 5 s, $0.60 for 10 s | about 4 to 5 min |
| LTX 2.5 Image to Video | Picture to clip, optional lip-sync to your audio | $0.35 for 5 s | about 1 min |
| LTX 2.5 First & Last Frame | Two pictures, the motion between them | $0.35 for 5 s | about 2 min |
| H3 Reference to Video | 1 or 2 reference pictures plus a prompt, 2 to 10 s | $0.47 for 4 s, $1.01 for 10 s | about 3 to 4 min |
| H3 Long Take | One continuous 8 s shot from two reference pictures | $1.42 | about 11 min |
| H3 Extend | Continues a clip by about 4.5 s | $2.30 | about 20 min |
| SolarRenderer | Grey 3D blockout plus a look picture to a finished shot | $0.93 | about 7 min |
| Seedance 2.5 (partner) | Up to 30 s, 480p or 720p, optional audio | $0.20/s (480p), $0.40/s (720p) | 2 to 8 min |
| Seedance 2 Text to Video and Image to Video (partner) | Up to 15 s with native audio, Standard or Fast | from $0.18/s | 1 to 5 min |
| Seedance 2 Omni Reference (partner) | Up to 9 images, 3 videos and 3 audio tracks as references | from $0.18/s | 1 to 5 min |
| Veo 3.1 (partner) | Text, image or first and last frame to video | $0.27/s | 1 to 3 min |
| Kling 3.0 (partner) | HD or 1080p with native audio, 3 to 15 s | from $0.19/s | 1 to 3 min |
| Wan 3.0 Text, Image and Reference to Video (partner) | Up to 30 s, 480p to 1080p | $0.10 to $0.40/s | 1 to 5 min |
Images
| Tool | What it does | Price | Typical time |
|---|---|---|---|
| Z-Image Turbo | Fast photoreal pictures, optional 4x PiD copy | $0.17, 16 for $0.50 | under 1 min |
| Z-Image Detailed | Extra detail, up to 3 LoRAs, optional reference photo | $0.31, 16 for $3.31 | about 2 min |
| Qwen Image 2.1 | Native 2K, text to image or edits with up to 5 photos, transparent PNGs | $0.14 | under 1 min |
| Krea 2 Turbo | Realistic photos, 14 style presets, Identity Edit tab | $0.15, 16 for $0.66 | under 1 min |
| Krea 2 GonzoLomo | Analog, lomography-style looks | $0.27, 16 for $0.54 | about 1.5 min |
| Krea 2 Ultra Realistic | Two passes on the Krea 2 Raw base, up to 1920 px | $0.37, 16 for $4.27 | about 2.5 min |
| MOSS Image | Strong composition, built-in prompt writer, up to 2 MP | $0.26, 12 for $1.91 | about 1.5 min |
| Anima | Anime art plus a semi-real mode | $0.17 | under 1 min |
| Nano Banana 2.1 (partner) | Generate, or edit with up to 14 reference pictures | from $0.20 | 10 to 30 s |
| GPT Image 2.5 (partner) | Crisp text in images, transparent backgrounds, up to 4K | from $0.07 | 15 to 45 s |
| Wan 2.7 Image (partner) | Generate and edit with up to 9 pictures, up to 2K | $0.10 to $0.20 | 30 to 60 s |
Edit & Swap
| Tool | What it does | Price | Typical time |
|---|---|---|---|
| Qwen Person Swap | Puts the person from one photo into another, body or head mode | $0.34 | about 2 min |
| Qwen Image 2.1 Inpainting | Paint over an area and describe what goes there, up to 2K | $0.19 | under 1 min |
| Krea 2 Inpainting | Name the area in words, such as “the red car”, and only that is repainted | $0.67 | about 5 min |
| Z-Image Turbo Face Inpaint | Finds the face and repaints only that region | $0.24 | about 1 min |
| Image Layers | Splits a picture into separate layers | $0.38 | about 2.5 min |
| H3 Character Swap | Swaps the person in a video, window of 5 to 15 s | $0.78 for 5 s, $1.45 for 10 s | about 6 min for a 4 s clip |
| GPT Image 2.5 Edit (partner) | Edits with up to 16 reference pictures, optional mask | from $0.07 | 15 to 45 s |
| Wan 2.7 Pro Image Edit (partner) | Edits up to 4K with up to 9 pictures | $0.16 to $0.48 | 30 to 90 s |
| Seedance 2 Video Edit (partner) | Changes a video with a text instruction | from $0.12/s, billed for output plus input length | 1 to 5 min |
| Wan 3.0 Video Editing (partner) | Style transfer, object replacement, up to 30 s | $0.10 to $0.40/s | 1 to 5 min |
Talking & Motion
| Tool | What it does | Price | Typical time |
|---|---|---|---|
| MiniMax H3 Talking Video | Photo plus your voice track to a talking video, up to 15 s | $0.61 for 5 s, $1.11 for 10 s | about 4 to 5 min for 5 s |
| InfiniteTalk | Face plus audio to long talking videos (audio up to 7 min) | $0.45 for 5 s, $0.80 for 10 s | about 7 to 8 min for 12 s |
| SCAIL-2 | Your character performs the motion of a drive video, up to 30 s | $0.53 for 5 s, $0.96 for 10 s | about 9 min for 12 s |
| OmniHuman 2.0 (partner) | Picture plus audio to a talking or singing performer, up to 5 min | $0.143/s | 3 to 15 min |
| Kling 3.0 Motion Control (partner) | Transfers the motion of a video to a character picture | from $0.09/s | 1 to 3 min |
Enhance & Upscale
| Tool | What it does | Price | Typical time |
|---|---|---|---|
| SeedVR2 4K Upscale | Pictures to 4K with new detail | $0.36 | about 2 min |
| SeedVR2 Simple Upscale | Quick, clean picture upscale | $0.14 | under 1 min |
| Simple Upscale | Classic 2x and 4x upscalers, nothing invented | $0.12 | under 1 min |
| RTX AI Upscale | Pictures and video | $0.12 per picture, $0.24 per 5 s of video | seconds for a picture, about 1 to 1.5 min for 5 s of video |
| SeedVR2 HD Video | Video to sharp HD | $0.86 for 5 s | about 6 to 7 min |
| Video Upscale & Detail Restore | Restores hair, skin and texture, optional 2x | $1.25 for 5 s | about 10 min |
| LTX 2.5 Upscaler | Up to 2.4 MP, doubles the frame rate, keeps the audio | $0.74 for 5 s | about 5 to 6 min |
| Frame Interpolation | Doubles the frame rate for smoother motion | $0.27 for 5 s | about 1.5 min |
| Film Grain | Analog film grain on any picture | $0.11, 16 for $0.20 | seconds |
Audio
| Tool | What it does | Price | Typical time |
|---|---|---|---|
| Stem Splitter | One file to three: dialogue, music and effects | $0.13 per file | under 1 min |
| Room Acoustics | Puts a dry voice or sound in a realistic room, plus the impulse response | $0.10 per file | seconds |
Your Files Are Kept for 24 Hours
Everything you upload and every result stays on our servers for 24 hours. After that it is deleted automatically. Download each result you want to keep, and keep your own copies of inputs you plan to reuse, such as character pictures or LoRA files.
Use Your Own LoRA
Many of the tools on our GPUs accept your own LoRA files (.safetensors) for a consistent character or style: LTX 2.5 (two slots), Z-Image Turbo, Z-Image Detailed (up to three), Qwen Image 2.1 (two), the Krea 2 tools, MOSS Image, Anima, the inpainting tools, H3 Character Swap and MiniMax H3 Talking Video. The LoRA has to be made for the model you use. Our LoRA training guide shows how to make one.
The Kitty App: the Studio Inside a Video Editor
The free Kitty app for Windows 10 and 11 is a full video editor with a built-in AI agent, and it uses the same druidcat.com account and credits. Tell the agent what you want and it plans the shots as blocks on a “Kitty Agent” track, shows what each one costs and generates them with the studio’s models when you say OK. A finished generation replaces its block. It can analyse a song’s structure, place clips on the beat, add captions and make kinetic typography reels.
The app replaces the retired Music Video Creator. More in our guides: how to make a music video with the agent, how the AI editing agent works and the free editor’s features.
Tips for Better Results
- Draft cheap, finish sharp. Test a video prompt as a 5-second LTX 2.5 clip at 1 MP ($0.35). Only the keeper goes to 2 MP or through the LTX 2.5 Upscaler.
- Explore with a batch. 16 Z-Image Turbo pictures cost $0.50. Pick the best one and refine it.
- Leave the PiD switch off until you need a large file. It adds $0.02 per picture.
- Keep a character consistent. Start from one good reference picture, then use Qwen Image 2.1 edits (up to 5 photos), Qwen Person Swap, H3 Reference to Video or your own LoRA.
- Need readable text in a picture? Use GPT Image 2.5 for posters, labels and logos.
- Describe the sound in LTX 2.5 prompts. The model writes its own soundtrack from your words.
- Pick the right upscaler. Simple Upscale when nothing should change, SeedVR2 4K Upscale when you want new detail.
- Expect a slower first job after a quiet period, while a GPU starts.
- Download within 24 hours. After that, files are deleted.
Get Started
- Register at druidcat.com and get the $0.20 welcome credit.
- Open the hub at druidcat.com/kitty-ai-studio.
- Pick a tool with the search or the category chips.
- Add your inputs, check the price on the button and generate.
- Download the result, and top up from $5 when you need more credits.
Want ComfyUI scripts, workflows and tutorials? Have a look at our Patreon and the tutorials.
FAQ
Is Kitty AI Studio free?
Signing up is free and you get a $0.20 welcome credit for 7 days. After that you pay per job from your credit balance. There is no subscription and no monthly fee.
How much does a picture or a video cost?
A Qwen Image 2.1 picture costs $0.14, 16 Z-Image Turbo pictures $0.50, a 5-second LTX 2.5 clip with sound $0.35 and a 5-second MiniMax H3 talking video $0.61. Every tool page shows the exact price before you pay.
What is the difference between “Our GPUs” and “Partner models”?
“Our GPUs” are open models we run on our own serverless GPUs (RTX 5090 class). “Partner models” such as Veo 3.1, Kling 3.0, Seedance, Wan, GPT Image 2.5, Nano Banana 2.1 and OmniHuman 2.0 run through the partners’ APIs at their usual prices.
How long are my files stored?
24 hours. Inputs and results are then deleted automatically, so download what you want to keep.
Do credits expire?
The $0.20 welcome credit is valid for 7 days. Paid credits are valid for about 12 months.
Why did my first job take longer?
Our GPUs are serverless. After a quiet period the first job waits a few minutes while a GPU starts. The next jobs run at the typical times.
Is there a desktop app?
Yes. The free Kitty app for Windows is a video editor with a built-in AI agent that uses the same models and the same credits.
