> For the complete documentation index, see [llms.txt](https://docs.atlas.design/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.atlas.design/atlas-ai-studio-overview/release-notes/release-0.49.0.md).

# Release 0.49.0

## Run API-based Labs on desktop

Use the desktop Lab runner to enter a description or photo, follow the applicable workflow, preview and download results, and use Open as graph where the Lab provides a workflow. Desktop runs the full applicable chain; step prices are estimates, and failed-step retry reuses completed steps.

When a Lab includes a prepared example, **Try the free example** replays its steps without starting server generation or using credits. You can skip to the result or replay it. Reopening a Lab restores its latest saved run in that workspace: finished results come back, and an in-progress step can be followed without starting completed steps again.

![A prepared fantasy-lantern example progressing through the desktop Lab steps.](https://3654894688-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2F1vrJnbrc1VBQjgywC3Fj%2Frelease049-lab-prepared-example-progress.jpg?alt=media)

A prepared fantasy-lantern example playing step by step.

![The prepared lantern result with Replay available.](https://3654894688-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2FqzHV2pPZJWtrQr217oL7%2Frelease049-lab-prepared-example-result.jpg?alt=media)

Skip to the prepared result, then replay the example.

### A forest-observatory product shot

Product shot places the same game-prop reference in a lantern-lit forest-observatory setting. Use Open as graph to continue working with the saved inputs and result.

![Product shot — the supplied chest and its completed forest-observatory scene. The workspace billing name is redacted.](https://3654894688-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2FZiVVLyZAeQgFjv3sp1L3%2Frelease-049-lab-privacy-product-shot-SAME-FIRST-photo-result.jpg?alt=media)

![Open as graph retains the photo, setting and result in an editable workflow. The displayed template has Gemini 3.1 Flash Lite settings; the retained result was generated by the frozen Nano Banana 2.1 API, rather than by rerunning this template.](https://3654894688-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2F0KJzYijvnvynj9G6u38H%2Frelease-049-lab-clean-private-Open-as-graph-ZOOM1.jpg?alt=media)

## Read Lab documentation inside Atlas

Supported framed Labs can show Documentation in the desktop top bar. Switch between the guide and Lab while preserving the loaded Lab; documentation mode supports a direct link and guides can open shared workflows in a new tab. Availability requires an allowed per-Lab documentation URL.

API-based Labs with a configured guide also offer **Documentation** over their desktop runner. Switching back preserves the runner’s inputs and run state.

![The Sketch to Asset documentation guide for creating a fantasy-lantern game prop.](https://3654894688-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2FGSBp8CanSla0MzOqQFa1%2Frelease049-sketch-asset-documentation.jpg?alt=media)

## Create a Blender continuation from saved progress

Create continuation adds an idle Blender agent connected to saved Scene and Conversation files, including usable partial progress from failed or cancelled runs. Creation does not run it; subsequent execution is separately charged. Both files and usable conversation context are required.

![Existing partial-result QA example: the saved sports-car Scene and Conversation connect to one new idle continuation.](https://3654894688-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2Fggp4kPTZ0X2tKUIhAffv%2Frelease-0.49.0-blender-sports-car-continuation-1080.jpg?alt=media)

## Choose how to start from Home

When enabled for eligible first-time workspace owners, a first-run Home offers **Labs**, **Your agent (MCP)** and **Workflows**, alongside an assistant prompt and an empty-canvas start. An assistant-started project opens with automatic node execution disabled; opening a blank canvas or template does not run nodes.

![Choose Labs, your MCP agent or Workflows for a forest-ranger equipment pipeline.](https://3654894688-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2Fu4CA9Y7TG8438RThYIS1%2Frelease049-home-start-720p.jpg?alt=media)

Starting with an empty canvas opens the assistant panel for the current session without sending a prompt or running nodes. Your saved panel preference stays unchanged.

## Animate image loops with Pixelcut

Pixelcut Looping Video adds image-based 5–15 second loops with Subtle or Spin motion and optional audio. The input shape is retained. Subtle loop seams can jump; Spin invents unseen sides, so results need visual review.

![Pixelcut workflow with image input, Subtle motion, audio off and the saved monster loop.](https://3654894688-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2FSRzxJTzIq23FX1H4gclW%2Frelease049-pixelcut-monster-canvas.jpg?alt=media)

{% embed url="<https://3654894688-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2FGpX7MEY6Xnp6l7FhFthi%2Frelease049-pixelcut-monster-original.mp4?alt=media>" %}
A pixel-art creature animated from one image.
{% endembed %}

## Choose MiniMax H3 Base for video generation

The full Text + Image to Video node adds MiniMax H3 Base for text, first/last-frame and multimodal-reference video. It supports native audio and 5–15 second durations; 2K/4K settings upscale a 768P result. Target soundtrack alignment trims or pads without changing playback speed.

![H3 Base workflow for The Verdant Meridian woodland-observatory reveal.](https://3654894688-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2FowG5Id8nbk1N7fpsNd0T%2Frelease049-verdant-meridian-h3-base-graph.jpg?alt=media)

{% embed url="<https://3654894688-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2FwhjQj2df0xrXjsqul5kO%2Frelease049-verdant-meridian-h3-base-original.mp4?alt=media>" %}
A woodland-observatory reveal for The Verdant Meridian, generated with H3 Base at native 768P with native audio. Original generated video.
{% endembed %}

Referance to Video Example: Carry the observatory architecture, camera movement and musical direction into another game-cinematic moment by connecting image, video and audio references together.&#x20;

![H3 Base receives image, video and audio references together in the complete game-cinematic workflow.](https://3654894688-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2F3T2VbWOgTHDdg1Ukh1ms%2Frelease049-h3-multimodal-native-reference-graph.jpg?alt=media)

{% embed url="<https://3654894688-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2F34tEl5QerQtNmu2CVIPr%2Frelease049-h3-multimodal-native-reference-original.mp4?alt=media>" %}
Original H3 Base multimodal-reference result with native audio, generated at 768P.
{% endembed %}

## Create PixVerse V6 first/last-frame transitions

Connect First Image and End Image to generate a PixVerse V6 transition. Transitions use both connected frames and ignore the ordinary aspect-ratio/outpainting settings; end-only input is rejected before generation.

<figure><img src="https://3654894688-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2FV3NJ0zjN2AbXT4ZasGtk%2Fimage.png?alt=media&amp;token=e765dfa0-058c-42ef-b3e9-9a46ed46a48f" alt=""><figcaption></figcaption></figure>

{% embed url="<https://3654894688-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2FNN8wc8akxQV7T5H2ZLeI%2Frelease049-pixverse-observatory-awakening-original.mp4?alt=media>" %}
Original generated PixVerse V6 transition with native audio, 5.04 seconds. The original video binary is preserved.
{% endembed %}

## Score video and create music variations with Sonilo

Sonilo adds up to three music variations, video-based music scoring with Prompt Influence, and video-synchronized sound effects. Score the whole video or a shorter selected duration. Video SFX has an optional Keep Original Speech control; audition the result to assess dialogue preservation.

<figure><img src="https://3654894688-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2Fd0J90scHGOq5dn9eyrfK%2Fimage.png?alt=media&amp;token=a3720de5-79d2-49c1-b9f0-f58e2a88c584" alt=""><figcaption></figcaption></figure>

## Clean speech with VEED Clean Audio

Edit audio adds VEED Clean Audio to reduce background noise in audio or a video audio track. Adjust Strength and normalization. The result is audio-only 48 kHz mono MP3; music can be suppressed and peak limiting can prevent an exact loudness target.

<figure><img src="https://3654894688-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2FRnhQIylAWGPUKMNaYz8i%2Fimage.png?alt=media&amp;token=c8b6ce90-1051-40c6-bf2e-7ebfc9008353" alt=""><figcaption></figcaption></figure>

## Use more speech engines

Text to speech adds AssemblyAI TTS 1, Voxtral Mini TTS 3.0, Fish Audio Drama 3 and Microsoft MAI-Voice 2.1. AssemblyAI, Voxtral and MAI offer model-specific voice selections; Drama 3 does not expose a voice selector.

<figure><img src="https://3654894688-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2FrvuTyaOU2Pq5uv6QrXs3%2Fimage.png?alt=media&amp;token=4fa23c0e-8a40-497d-abde-045c8d9b83e1" alt=""><figcaption></figcaption></figure>

<figure><img src="https://3654894688-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2F0j8YGG5Fu3uDGr4oYqKp%2Fimage.png?alt=media&amp;token=d16150a3-12ba-478f-b321-a2b2092b7b31" alt=""><figcaption></figcaption></figure>

<figure><img src="https://3654894688-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2FuWHwgPv5OwzVF7aefm2l%2Fimage.png?alt=media&amp;token=b563aa39-9dcb-4c1c-a9e8-2e59c6109d1d" alt=""><figcaption></figcaption></figure>

HeyGen Voice and Voice Turbo add 24 preset voices or reference-audio cloning, automatic or explicit language selection, and Expressiveness controls for character dialogue and narration. Text can be up to 5,000 characters; output remains the original WAV.

<figure><img src="https://3654894688-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2FWv5rX1JBsPzk28Vv5YT9%2Fimage.png?alt=media&amp;token=6ad7d60a-bd14-4fa7-a5c0-29c51680ea06" alt=""><figcaption></figcaption></figure>

## Texture with Meshy N1 and preserve textures through Auto Split

Texture Mesh adds Meshy N1 with one to four image references, fixed 2K base color and optional original-UV preservation. Auto Split adds Keep Texture for eligible project-owned textured Meshy tasks. Eligibility and task lifetime still apply.

<figure><img src="https://3654894688-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2FMsi6F2vBoMUOW7trhcGH%2Fimage.png?alt=media&amp;token=41f6f0a3-c83f-420f-a013-95ba0db11d5d" alt=""><figcaption></figcaption></figure>

<figure><img src="https://3654894688-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2FDCUl5QqP0Fx3O8Q3RuYw%2Fimage.png?alt=media&amp;token=4d951a14-da73-4762-9981-823fce75ee42" alt=""><figcaption><p>Before Retexture</p></figcaption></figure>

<figure><img src="https://3654894688-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2FcWHnMnwb2LFGXv8WjpmO%2Fimage.png?alt=media&amp;token=b8174901-bd26-4a27-9f3e-5b8e51e94682" alt=""><figcaption><p>After Retexture</p></figcaption></figure>

<figure><img src="https://3654894688-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2Fz4B3AWCntrHCJ24ddPI8%2Fimage.png?alt=media&amp;token=62892768-b523-4207-a0be-e70463b2587e" alt=""><figcaption></figcaption></figure>

<figure><img src="https://3654894688-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2FKEW2ug0MmRnhMX2aOYj3%2Fimage.png?alt=media&amp;token=4a8539dd-f5e2-4f5e-a226-920436bb7e93" alt=""><figcaption></figcaption></figure>

Split the eligible textured source into assembled pieces, with Keep Texture off or on. These independent splits produced different part counts and boundaries.

## Rig humanoids with Tripo v3.0

Rig Tripo Mesh adds the v3.0 humanoid rig while keeping v1.0 as the default. Connect its Rig Metadata into animation to use the 11 supported core presets with v3.0.

<figure><img src="https://3654894688-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2FciQSftCnehQForPXXHVv%2Fimage.png?alt=media&amp;token=fcfbf83a-5130-4701-a7e5-ea30fbe3e3f8" alt=""><figcaption></figcaption></figure>

<figure><img src="https://3654894688-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2FB6ChPBtvnv0DKcu6U2AK%2Fimage.png?alt=media&amp;token=996b4810-f5c8-4c3f-b869-05a5ee74e158" alt="" width="436"><figcaption><p>v 3.0</p></figcaption></figure>

<figure><img src="https://3654894688-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2F16PlCHAObJ6CxeCs7cJZ%2Fimage.png?alt=media&amp;token=2cbe179c-f98d-4b34-812c-8db17bd4e028" alt="" width="371"><figcaption><p>v 1.0</p></figcaption></figure>

## Generate and convert native Ideogram vectors

Generate SVG adds native Ideogram Vector output, image references, quality/aspect controls and up to eight variations. Convert Image to SVG adds Ideogram Default and Simple detail. SVGs remain vector files; detail differences depend on the input.

<figure><img src="https://3654894688-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2FQNRPwCGaZ91HZ9d8DpfQ%2Fimage.png?alt=media&amp;token=7207e615-2def-46cd-9474-0a1e93bfcf3e" alt=""><figcaption></figcaption></figure>

![Reference-guided generation ](https://3654894688-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2Fgu9juLmmaoYJfbtDedYI%2Frelease049-svg-guild-high-square-graph.jpg?alt=media)

![Example: Silverpine Guild:previewed from its original SVG.](https://3654894688-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2FGBrmmWLvHsToaQeSLqtu%2Frelease049-svg-guild-high-square-native-2-native-preview.jpg?alt=media)


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation by asking a question.

Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters:

```
GET https://docs.atlas.design/atlas-ai-studio-overview/release-notes/release-0.49.0.md?ask=<question>&goal=<user_goal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case.

The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.
