A free, native creative studio for AI-generated media — image, video, music, and speech — that runs entirely on your machine. Bring your own provider keys; your projects live as plain folders on your disk.
Avatica talks to AI providers directly using keys you supply — generation costs go straight to your provider account and nothing else does.
The studio is organized around projects (folders), apps (focused tools that read and write typed assets), and sequences (.seq files you build on a multi-track timeline).
Download the latest installer from the Releases page, or get it from the Mac App Store on macOS.
- Pick a projects folder. On first launch Avatica asks where to store projects. It creates an
Avatica/subfolder in the location you choose (default:~/Movies/Avatica/). Each project is a folder inside it — open it in Finder/Explorer any time. - Add at least one API key. Settings opens automatically if none are configured. Avatica supports three provider families, and you only need one:
- Google Gemini — text, image (Imagen / Nano Banana), video (Veo), music (Lyria), speech. Get a key at aistudio.google.com/apikey.
- xAI — Grok for text, image, video, and TTS. Get a key at console.x.ai.
- OpenAI — GPT Image 2. Get a key at platform.openai.com/api-keys. Keys are stored locally and never leave your machine except as calls to the provider you chose.
- Create a project. Click New Project, give it a name, and you land in the workspace.
Three modes across the top, panels on the sides:
- Create — pick a built-in app from the dropdown and generate. Outputs land in the output grid and on disk in your project folder.
- Compose — multi-track timeline. Drag assets onto tracks, trim, layer text overlays, scrub, and export an MP4.
- Develop — build your own apps and asset types.
Toggle the side panels from the top-right:
- Assets (left) — file browser for the current project. Drag files onto app input slots; double-click a
.seqto open it in Compose. - Log (bottom) — running output from app invocations.
- Chat (right) — the assistant. It sees your project's files and built-in apps as tools and can chain them: generate a character, then a shot from that character, then drop the shot on the timeline.
- Image Generation — text-to-image, or edit an existing image with a prompt. Optional reference images for style.
- Fashion Studio — fashion photoshoot from a reference photo, with style, setting, and shot selections.
- Hair Stylist — hairstyle visualizations from a reference photo, with style, color, and view selections.
- Makeup Artist — makeup looks from a reference photo, with style, skin tone, and view selections.
- Video Generation — text-to-video, image-to-video, frame-to-frame interpolation, or extend an existing clip.
- Cinema Studio — animate a
.shotasset (pre-composed first frame with characters/scene/objects) into a cinematic clip with camera movement and action. - Monologue Studio — turn a shot plus monologue text into a full delivery video. Splits the script into chunks of the target duration, generates each in parallel from the shot's frame, and assembles a sequence.
- Song Generation — music from a text prompt, with optional genre, mood, tempo, and instrument controls.
- Speech Generation — speech audio from text with voice selection.
- Character Creator — a character with portrait, description, and attributes; saved as a
.characterasset other apps can consume. - Object Creator — an object or product from a reference image, description, and attributes; saved as a
.objectasset. - Scene Creator — a scene or location from a reference image, description, and attributes; saved as a
.sceneasset. - Shot Creator — a cinematic first frame from scene + characters + objects (create mode), or a new camera angle that continues a previous shot (continue mode). Output is a
.shotasset that Cinema Studio animates.
- Script Writer — screenplays and scripts for short films, music videos, and AI video content.
- Monologue Writer — standup, storytelling, TED talk, dramatic, pitch, vlog, spoken word, and more.
Drag characters, objects, scenes, or shots from the asset browser onto another app's input slot to chain them together.
Every provider is enabled out of the box once its key is set. The app picks a sensible default per task; you can override the model from the app surface.
| Model | What it does |
|---|---|
| Gemini 3.1 Pro | Text generation (scripts, monologues, reasoning) |
| Gemini 3.1 Flash Lite | Faster/cheaper text generation |
| Nano Banana 2 (Gemini 3.1 Flash Image) | Image generation and edit, up to 4K, full range of aspect ratios |
| Gemini 3.1 TTS | Speech in 18 languages, 30 voices (Zephyr, Puck, Charon, Kore, …) |
| Lyria 3 Clip | Music generation (mp3/wav) |
| Lyria 3 Pro | Higher-quality music generation (mp3/wav) |
| Veo 3.1 Lite | Video generation and image-to-video up to 1080p |
| Veo 3.1 Fast | Video generation, image-to-video, interpolation, and extend up to 4K |
| Veo 3.1 | Highest-quality Veo: generation, image-to-video, interpolation, and extend up to 4K |
| Model | What it does |
|---|---|
| Grok 4.2 | Text generation |
| Grok Imagine | Image generation and edit, 1K/2K, full range of aspect ratios |
| Grok Imagine Pro | Higher-quality image generation and edit |
| Grok Imagine (video) | Video generation, image-to-video, extend, and edit at 480p/720p, 1–15s |
| Grok TTS | Speech in 16 languages, 5 voices (Eve, Ara, Rex, Sal, Leo) |
| Model | What it does |
|---|---|
| GPT Image 2 | Image generation and edit, 1K/2K (4K at 16:9 / 9:16), wide aspect ratio range |
- Projects are real folders. Move, back up, or sync them however you like.
- Double-clicking a
.seqfile in Finder opens it in Compose. - Generated images, video, audio, and JSON assets are named based on the prompt that produced them.
Open Settings from the navbar (gear icon):
- General — API keys, projects root folder, light/dark theme.
- Usage — running cost totals and per-app/per-provider breakdown, estimated from each provider's published pricing.
GPL v3. You're free to use, modify, and distribute Avatica — including for creating commercial content. If you distribute a modified version, the source must be made available under the same license.