> ## Documentation Index
> Fetch the complete documentation index at: https://julius.ai/docs/llms.txt
> Use this file to discover all available pages before exploring further.
# Images & Videos
> Generate, edit, and transform images and videos directly in chat — from text prompts, uploaded files, or animated HTML components.
Julius can generate, edit, and transform images and videos directly in chat: from text prompts, from files you upload, or from animated HTML components.
***
## Overview
Julius offers four media-generation capabilities inside any conversation:
Create images from text prompts, or edit uploaded images with masks, restyling, and transformations.
Text-to-video, image-to-video, reference-based video, plus editing and extending existing MP4 clips.
Pixel-perfect product demos and explainers rendered from HTML/CSS/GSAP code, not an AI video model.
Generate spoken narration (MP3) to pair with videos, with a choice of built-in voices.
***
## Quick Start
You don't need any special syntax — just ask in plain English:
> Generate an image of a sunset over a mountain lake in watercolor style
> Take the photo I uploaded and remove the background
> Create a 10-second video of waves crashing on a rocky shore at dusk
> Animate the logo image I uploaded so it fades in and pulses
> Make me a 15-second product demo video for my analytics app
Generated media is saved to your conversation's sandbox and displayed inline. You can download any file from the Files page, or reference it in follow-up requests (e.g. "now extend that video by 5 seconds").
***
## Generating Images
Text-to-image generation uses OpenAI's image model. Describe what you want — subject, style, lighting, composition, mood — and Julius handles the rest.
### Examples of what you can create
* **Photorealistic scenes** - "a candid photo of a barista pouring latte art, shallow depth of field"
* **Illustrations & art** - "flat vector illustration of a remote team collaborating, warm palette"
* **Product mockups** - "a sleek smartwatch on a marble surface, studio lighting"
* **Diagrams & graphics** - charts with custom styling, banners, thumbnails, icons
* **Logos & brand assets** - "minimal geometric logo for a coffee brand called Ember"
### Follow-up edits
Images are iterative. After generating one, you can refine it conversationally:
> Make the lighting warmer
> Remove the person in the background
> Regenerate it as a JPEG in landscape
### Image Parameters
You usually don't need to set these since Julius picks sensible defaults, but you can request them explicitly:
| Parameter | Options | Notes |
| --------------- | ----------------------------------------- | --------------------------------------------------------------------------------------------------------------- |
| `size` | `auto`, or dimensions like `1024x1024` | Edges must be multiples of 16; max edge 3840px; aspect ratio at most 3:1; total pixels between \~0.65M and 8.3M |
| `output_format` | `png` (default), `jpeg`, `webp` | Use `jpeg` for smaller files, `png` for transparency |
| `quality` | `auto` (default), `low`, `medium`, `high` | `high` for final assets; `low` / `medium` for fast drafts |
***
## Editing Images
You can edit images you've uploaded or images Julius generated earlier in the conversation. Supported edit types include:
* **Inpainting / removal** - erase objects, people, watermarks, or text
* **Restyling** - turn a photo into a painting, sketch, cartoon, 3D render, etc.
* **Transformations** - change backgrounds, relight, recolor, change seasons or time of day
* **Masked edits** - provide a mask image to target edits to a specific region
* **Multi-image composition** - combine or reference multiple source images in one edit
When asking for an edit, refer to the file by name (e.g. `product-photo.png`) or upload it with the + button. Julius uses the exact file from your conversation, not a URL.
***
## Generating Videos
Julius can generate short AI videos from scratch. Describe the scene, camera movement, and mood:
> Generate a 10-second cinematic video of fog rolling over a pine forest at sunrise
### Capabilities
* **Text-to-video** - describe the scene and get a video clip
* **Image-to-video** - animate a still image; the image becomes the first frame
* **Reference-to-video** - provide up to 7 images as visual references for style/subjects
* **Video-to-video editing** - transform an existing MP4 (restyle, modify content) while keeping its duration, resolution, and aspect ratio
* **Video extension** - continue a clip from its final frame, adding 2-10 seconds
### Image-to-Video & Editing
**Animating a still image**
> Animate this product photo so the camera slowly orbits around it
The uploaded (or previously generated) image is used as the first frame. Julius matches the output aspect ratio to your source image automatically.
**Editing an existing video**
> Take the clip I uploaded and make it look like it was shot on vintage film
Video-to-video editing preserves the source's **duration, aspect ratio, and resolution**. Source videos must be MP4 and no larger than 200 MB.
**Extending a video**
> Extend that video by 5 more seconds
Extension continues motion naturally from the last frame. Extensions support 2-10 seconds per request.
Image inputs and video inputs can't be combined in a single generation. Image inputs must be PNG or JPEG, each under 20 MB (Julius can downscale oversized files for you automatically).
### Video Parameters
| Parameter | Options | Notes |
| -------------- | ------------------------------------------------- | -------------------------------------------------------------- |
| `duration` | 1-15 seconds | Required for extensions (2-10s); optional otherwise |
| `aspect_ratio` | `16:9`, `9:16`, `1:1`, `2:3`, `3:2`, `3:4`, `4:3` | Text/image-to-video only; ignored for video inputs |
| `resolution` | `480p`, `720p` | Text/image-to-video only |
| Source images | 1-7 PNG/JPEG, max 20 MB each | First image = first frame; extras = visual references |
| Source video | MP4, max 200 MB | For editing or extension; cannot be combined with image inputs |
For social content: use `9:16` for Reels/TikTok/Shorts, `16:9` for YouTube, and `1:1` for feed posts. Say it in your prompt, e.g. "vertical 9:16 video", and Julius sets it for you.
***
## HTML / Animated Component Videos
For product demos, launch videos, explainers, title sequences, and animated UI mockups, Julius uses a different approach: instead of an AI video model, it authors the animation in HTML, CSS, and GSAP, then renders it deterministically to MP4 with headless Chromium and FFmpeg.
### When to use which
| Use AI video generation for... | Use animated components for... |
| --------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------- |
| Real-world scenes, nature, people, cinematic footage, artistic motion | Product UI demos, screen recordings with callouts, kinetic typography, data visualizations, branded motion graphics |
| Organic, unpredictable motion | Pixel-perfect, frame-accurate, on-brand motion |
| Short clips up to 15s | Longer compositions with narration and scenes |
### How it works
1. Julius writes a narration script and generates voiceover audio (optional).
2. It builds a single self-contained HTML file with one paused, seekable GSAP timeline; scenes, transitions, and motion are all code.
3. The composition is rendered frame-by-frame in headless Chromium and encoded to MP4 with FFmpeg.
4. The same source can also be delivered as an **interactive HTML player** with the narration audio driving the timeline.
Because the timeline is deterministic code, revisions are precise: *"make the headline slide in faster at 4 seconds"* is a one-line change, not a re-roll of an AI model.
***
## Audio / Text-to-Speech
Julius can generate spoken narration as MP3 files to accompany videos, presentations, or standalone use:
> Generate a friendly voiceover that says: "Welcome to Acme Analytics..."
Available built-in voices: `alloy` (default), `ash`, `ballad`, `cedar`, `coral`, `echo`, `fable`, `marin`, `nova`, `onyx`, `sage`, `shimmer`, `verse`. Scripts up to 4,096 characters per generation.
When combining narration with video, Julius inspects the audio duration and syncs scene timing to it, extending or looping visuals if the narration runs long.
***
## Prompt Guide
### Anatomy of a good image prompt
`[subject] + [action/pose] + [environment] + [style/medium] + [lighting] + [camera/mood]`
> "A golden retriever puppy sitting in a field of wildflowers, soft watercolor painting, warm morning light, gentle and dreamy mood."
### Anatomy of a good video prompt
`[shot type] + [subject + motion] + [environment] + [camera movement] + [lighting] + [duration + aspect ratio]`
> "Medium shot of a chef flipping a pancake in a rustic kitchen, slow dolly-in, warm tungsten lighting, cozy cinematic feel, 8 seconds, 16:9."
### Do's and don'ts
* Be specific about motion for videos: *what moves, how fast, in what direction*
* Name the style explicitly: *cinematic, anime, 3D render, film noir, flat illustration...*
* Iterate in small steps; one change per follow-up gives cleaner results
* Don't cram multiple unrelated subjects into one short video clip
* Don't expect legible small text in AI-generated images/videos — for text-heavy visuals, use animated HTML components instead
***
## Limits & Constraints
| Feature | Limits |
| ---------------------- | ------------------------------------------------------------------------------------ |
| Image generation | Max edge 3840px; aspect ratio max 3:1; 0.65M-8.3M total pixels; PNG/JPEG/WebP output |
| Video generation | 1-15s per generation; 480p or 720p; 7 aspect ratios |
| Video editing | MP4 input, max 200 MB; keeps source duration/resolution/aspect |
| Video extension | 2-10s per extension, from the clip's final frame |
| Image inputs for video | 1-7 images, PNG/JPEG, max 20 MB each; not combinable with video input |
| TTS audio | Max 4,096 characters per generation; MP3 output; 13 voices |
Need a longer video? Chain extensions ("extend it again") or stitch multiple generated clips together. Julius can concatenate them with FFmpeg right in the sandbox.
***
## FAQ
Yes. Just refer to it ("the logo you made earlier" or the filename). Julius keeps generated media in the conversation sandbox and can use it as a source for future edits or video animation.
Yes. Upload a PNG or JPEG with the + button and ask Julius to animate it. It becomes the first frame, and you can add up to 6 more images as style/subject references.
Yes. Julius can generate narration with TTS and mux it into the video with FFmpeg. For animated component videos, narration timing drives the whole composition. AI-generated video clips can also be combined with a narration track afterward.
For UI mockups, text, and branded motion, code-rendered animation is sharper and infinitely more editable than AI video. Julius picks the right tool for the job, but you can always ask for a specific approach.
Everything is saved to your conversation and appears on the Files page for download. Generated HTML artifacts can be viewed inline and published.
***
## Related
* [Artifacts overview](/docs/artifacts/overview)
* [Interactive artifacts](/docs/artifacts/interactive-artifacts)
* [Slides](/docs/artifacts/slides)
* [Reports](/docs/artifacts/reports)