> 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-0.48.0.md).

# Release 0.48.0

Previous Version: **0.47.0**

Edit existing clips with timed references, turn a concept into a low-poly game prop, and plan 3D workflows with input-based estimates.

## Edit existing clips with Runway Aleph 2

Describe a change to an existing clip in **Edit video**, or guide the edit with timed image references.

<figure><img src="https://3654894688-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2FQ6cCj5pjaz1PlisutjdV%2Fimage.png?alt=media&amp;token=c241b062-d04c-49df-a650-7fb0cec4e3a3" alt="" width="254"><figcaption></figcaption></figure>

<figure><img src="https://3654894688-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2F00cjGlR9uI450cy8iqoF%2Fimage.png?alt=media&amp;token=c46a2dd8-fa1e-44ba-aa3e-55a43562096f" alt=""><figcaption></figcaption></figure>

{% embed url="<https://files.gitbook.com/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2FzWNMXIRjIoslnrCr5YFn%2Fgenerated_video%20(12).mp4?alt=media&token=cbfc6d30-fc61-464c-a892-b02562ed0623>" %}
Input Video
{% endembed %}

{% embed url="<https://files.gitbook.com/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2Fnm8BHBdAmvXRypMp0ybN%2Fedited_video%20(2).mp4?alt=media&token=aeb433c3-6b6a-4278-98fd-1a9744d43356>" %}
Edited Video
{% endembed %}

## Create low-poly game assets with Meshy T2.5

**Image to 3D** and **Image to 3D with Fallback** now offer **Meshy T2.5 Smart Topology**. Set an approximate triangle target and choose the texture resolution for your asset.

<figure><img src="https://3654894688-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2FC4xA5Nl5A88grj2yjMQl%2Fimage.png?alt=media&amp;token=3fd03829-995a-4ba1-ac0e-d0c30d2d5bc8" alt=""><figcaption></figcaption></figure>

<figure><img src="https://3654894688-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2Ft5az33xT8mzvZBb0gVld%2Fimage.png?alt=media&amp;token=dedd5e72-a4fc-4f5f-9a40-e1e7afc92707" alt=""><figcaption></figcaption></figure>

## Plan 3D work around a successful generation path

**Image to 3D**, **Multi-view to 3D** and **Image to 3D with Fallback** can use planning estimates that account for a successful eligible generation path and its required processing. Retries and fallbacks remain available under the existing execution limits.

Quotes depend on inputs, settings and pricing policy; provider prices are unchanged. When the primary route may be skipped by eligibility rules, the maximum quote can remain.

![Planning a successful primary route while retaining exceptional retry and backup allowances in the maximum.](https://3654894688-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2FlyxFCtFmsVN0uEDOxQZR%2Frelease048-fallback-planning-compact.jpg?alt=media)

## Download originals from the node surface

Use the visible **Download** control on a media preview to save its original file. Eligible node outputs also show **Download all** for a ZIP, including when you have read-only project access.

![The three original embedded potion material maps, reimported unchanged.](https://3654894688-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2FAaZtqsYnwTaj0e38dcFt%2Frelease048-material-maps-compact.jpg?alt=media)

## Keep a started phone Lab moving between tabs

Where phone continuation is enabled, a started Lab can progress through its remaining steps while you visit other phone tabs. Keep Atlas open, visible and online; finished results remain in Assets.

![The started UI Icon Lab appears on Runs while its Icon step is running.](https://3654894688-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2FhL3ufYyHrncWdm49SHKn%2Frelease048-phone-run-running.jpg?alt=media)

<figure><img src="https://3654894688-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2FfGOHjnJh76EEEMJWw9Hl%2Frelease048-c1-005-c1-005-runs-current-landscape.jpg?alt=media" alt="The Runs tab lists completed historical workflows, including the UI Icon example; no new workflow was started."><figcaption><p>The Runs tab lists completed historical workflows, including the UI Icon example; no new workflow was started.</p></figcaption></figure>

A Lab paused at Draft ready waits for your choice, including while you visit another phone tab. Only Finish it starts its remaining steps.

## Open the desktop handoff from the Labs library

After a completed phone result, the library card offers **Send me a link** to continue on desktop with the same account and assets.

<figure><img src="https://3654894688-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2F0JtnA8RdwwMhmK5mZyFl%2Frelease048-c1-006-c1-006-eligible-library-card-current.jpg?alt=media" alt="The eligible mobile Labs card shows Send me a link. No email was sent or delivery checked."><figcaption><p>The eligible mobile Labs card shows Send me a link. No email was sent or delivery checked.</p></figcaption></figure>

## Dub video speech with ElevenLabs Dubbing v2 (Alpha)

Edit video adds target-language, source-language and voice-similarity controls, returning a dubbed video, MP3 audio and complete WebVTT subtitle text while copying the original video stream.

![Elevenlabs dubbing controls and results.](https://3654894688-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2FnwcNvW6Jq9WdRs2F7LMH%2Frelease048-c1-007-c1-007-elevenlabs-dubbing-controls-and-results.jpg?alt=media)

{% embed url="<https://3654894688-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2FMlwEMIkfBhbPArvt5eSr%2Frelease048-c1-007-game-prototype-english.mp4?alt=media>" %}
English game-prototype source clip.
{% endembed %}

{% embed url="<https://3654894688-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2FWZFF6Ijl1s2Jx5K1M4Hg%2Frelease048-c1-007-elevenlabs-output_video.mp4?alt=media>" %}
ElevenLabs Dubbing v2 Alpha Spanish game-prototype video.
{% endembed %}

<figure><img src="https://3654894688-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2FHD5oey0peuXiYdYEafq3%2Frelease048-c1-007-seven-helper-02.jpg?alt=media" alt="Dubbing v2: English to Spanish, voice similarity 7, with saved dubbed audio."><figcaption><p>Dubbing v2: English to Spanish, voice similarity 7, with saved dubbed audio.</p></figcaption></figure>

## Choose smaller Blender Agent run budgets

Maximum Credits is visible among the main inputs and accepts caps from 50 credits, with the existing 1,000-credit default. When credits block a run, Edit Maximum Credits focuses the setting so you can choose a smaller run.

![Blender 50 credit result and controls.](https://3654894688-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2FS0PyQ2swNWLpf9i9DRMn%2Frelease048-c1-008-c1-008-blender-50-credit-result-and-controls.jpg?alt=media)

![Blender blocked guidance.](https://3654894688-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2FVDqD18zhTkrBdFoYnLOj%2Frelease048-c1-008-c1-008-blender-blocked-guidance.jpg?alt=media)

## See when the assistant follows up automatically

Chat marks follow-ups started after watched nodes finish and identifies those nodes.

<figure><img src="https://3654894688-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2FuHryzWL84Q3V6Oecqd4M%2Fimage.png?alt=media&amp;token=c145a5f6-356d-4d75-a7ff-84bd8309d8b6" alt=""><figcaption></figcaption></figure>

## Extend a track with ElevenLabs Music v2.5

Modify music can keep reference audio and compose an intro, an outro, or both. Set styles to include or avoid, section text/lyrics, Continuity and Seed. Each added section is 0 (skip) or 3–120 seconds; at least one must be enabled, the reference must be at least 10 seconds, and the complete output must fit within 10 minutes. Existing Modify music selections migrate without switching engines.

![ElevenLabs Music v2.5 Extend connected to an input track, with Extend After set to 20 seconds and both waveforms visible. Retained example; no new generation.](https://3654894688-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2Fg2328mOmtqYOLPwyMDTb%2Frelease048-c1-010-music-graph.jpg?alt=media)

## Report the result shown on a node

The node menu adds Report bad result when feedback is enabled and a recorded run is selected, including in read-only views. The form opens as Bad result, identifies the displayed run and lets you turn off its Result attachment. Drafts are kept separately for each reported run.

![Select a recorded run and choose Report bad result from the node menu. Controlled interface example; no report was sent.](https://3654894688-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2FDWDEEFntaLPMsch2rMsg%2Frelease048-c1-011-selected-run-menu.jpg?alt=media)

<figure><img src="https://3654894688-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2FztYOIkZQ0tKUabkgdxpj%2Fimage.png?alt=media&amp;token=ddbb03d7-2091-4964-b5e1-ce5fc1cc1574" alt="" width="320"><figcaption></figcaption></figure>

## Animate an image with Mischief-1

In Text + image to video → Simple, choose TBC :: Mischief-1 and supply a first image. Generate a silent landscape clip at 3 or 5 seconds in 480p (832×480) or 720p (1280×720), at 16 fps. Image Fit offers Center crop or Stretch; Prompt Processing offers Original or image-aware Enhanced rewriting. Defaults are 5 seconds, 720p, Center crop, Enhanced and a randomized seed; prompts accept up to 2,000 characters and images up to 25 MiB/100 megapixels. Enhanced can alter precise instructions; Original keeps the prompt as written.

<figure><img src="https://3654894688-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2FcRmv63brjBzqzZIBgZpg%2Fimage.png?alt=media&amp;token=2319f529-23c1-42b0-8b91-a405d9a09393" alt=""><figcaption></figcaption></figure>

<figure><img src="https://3654894688-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2F3KPLQC5GLZ6qdkPE7at2%2Foutput%20-%202026-11-04T160532.656.png?alt=media&amp;token=ed319d51-c6f7-4ac1-9365-5eb0cfc86e5c" alt=""><figcaption><p>Input Image</p></figcaption></figure>

{% embed url="<https://files.gitbook.com/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2FIfcKNHKcQPf2makcaELi%2Fmischief%20(1).mp4?alt=media&token=de5ed5c5-c383-4b97-ad85-93b338da4eab>" %}
Output Video
{% endembed %}

## Dub speech with VEED Standard or Pro in Edit video

Choose VEED :: Dubbing Standard or VEED :: Dubbing Pro, provide a video with readable spoken audio up to 10 minutes, and select Dub Into and Spoken Language. Language choices are stored separately for each tier; Pro is not a strict language superset. Pro adds Voice Cloning Strength from 0–10 (default 7). Optional Keyterms use square-bracket groups, such as \`\[Atlas] \[Clockwork Knight]\`. VEED renders a new dubbed video with altered lips/faces; check every speaker because some lines can remain untranslated.

<figure><img src="https://3654894688-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2FtRQSVUdUVcKiT3M8qS47%2Fimage.png?alt=media&amp;token=69756fdf-b8bf-4c71-a7fb-29f835874de8" alt=""><figcaption></figcaption></figure>

{% embed url="<https://files.gitbook.com/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2F6fSRFbDlkNhGNJnK9heq%2Fgenerated_video%20(13).mp4?alt=media&token=7bad50e2-8378-4494-938e-b25443db1bbf>" %}
Original English Video
{% endembed %}

{% embed url="<https://files.gitbook.com/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2FC1LRjcFyYi8XpCzFflca%2Fedited_video%20(3).mp4?alt=media&token=3db42a0b-2924-4656-9d48-450fe039ead4>" %}
Dubbed Spanish Video
{% endembed %}

## Choose Recraft V4.1 image engines

Text to image adds Recraft V4.1, V4.1 Pro, Utility, Utility Pro and Flash. Standard/Utility/Flash produce 1K output; Pro variants produce 2K. Utility variants target simpler flat-lit mockups and product scenes. The four non-Flash engines offer 14 aspect ratios; Flash offers five. V4.1 accepts prompts up to 10,000 characters and does not support Style ID or Seed.

![Recraft flash result.](https://3654894688-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2FMZ1SsbxmhRu8bYvNBCl0%2Frelease048-c1-015-c1-015-recraft-flash-result.jpg?alt=media)

![Recraft pro utility utility pro raster graph.](https://3654894688-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2F0F8FNywMkOhM5vFirBTG%2Frelease048-c1-015-c1-015-recraft-pro-utility-utility-pro-raster-graph.jpg?alt=media)

## Choose H3 Max Turbo in Text + image to video

The existing MiniMax H3 Max backend gains a Model choice for H3 Max or H3 Max Turbo across Text, Image and Reference modes. Requests support 0.92–15 seconds at 480P, 768P or 1080P; generated output may run about 0.7 seconds longer. Turbo Reference accepts up to nine images, three videos and three audio clips, with at most 12 files combined. Reference video/audio clips are 2–15 seconds each and at most 15 seconds combined per modality. Turbo Reference uses reference arrays; first/middle/last frame controls are unavailable there. Turbo Image retains optional first/last frames.

<figure><img src="https://3654894688-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2FP4cR2ucJIsrD2yvsb4dI%2Frelease048-c1-016-feature-minimax-h3-turbo-reference.png?alt=media" alt="H3 Max Turbo reference-image mode and generated result"><figcaption><p>H3 Max Turbo Reference mode uses the supplied reference image. This example shows the full Text + image to video node.</p></figcaption></figure>

## See waiting queues on supported 3D and mesh nodes

Supported 3D and mesh nodes show waiting-queue counts or a full-queue badge, with details explaining what each queue measures. Queue unknown keeps unavailable status explicit, and Hunyuan Rapid distinguishes its primary queue from configured backup availability. Status is advisory and does not guarantee an immediate start. Existing rigging status uses the shared presentation.

![Controlled interface example showing ready, primary-queue and full-queue indicators on supported 3D and mesh nodes. These values illustrate advisory states.](https://3654894688-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2Fcyq0d14RiFqMWdu2PsJI%2Frelease048-c1-018-queue-graph-ready-primary-backup-mesh.png?alt=media)

## Generate and edit with Krea 3

In Text + image to image or Text + image list to image, choose Krea 3 and Berry or Jam to generate from a prompt or edit and combine references. The list form accepts up to eight references; the single form has four image connectors. Use 1K/2K presets, custom dimensions, automatic framing, or rough/exact matching to an input image. Berry adds a negative prompt. Defaults are Berry, medium reasoning and automatic 2K; this slice offers low and medium reasoning.

<figure><img src="https://3654894688-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2FssD15tKZaLB7NT53584e%2Fimage.png?alt=media&amp;token=8b924671-0c87-48e8-a5f2-5ac53988e738" alt=""><figcaption></figcaption></figure>

<figure><img src="https://3654894688-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2FZ8ZNC7IGA7hElv2UT3zi%2Fimage.png?alt=media&amp;token=77c05ad5-0990-4a34-907d-9273cb41a88e" alt=""><figcaption></figcaption></figure>

## LTX-2.5 Pro and Fast video generation

Text + image to video → Simple now includes LTX-2.5 Pro and Fast. Generate from a prompt or first frame, add an optional end frame, and choose duration, resolution, frame rate, camera motion and generated audio. Pro offers 6/8/10-second clips at 720p or 1080p. Fast adds even durations through 20 seconds and output up to 4K; clips over 10 seconds require 720p/1080p at 24/25 FPS.

Open Text + image to video → Simple and select Lightricks :: LTX-2.5 Pro or Lightricks :: LTX-2.5 Fast in Engine. Use Duration, Resolution, Frame Rate, Aspect Ratio, Camera Motion and Disable Audio. End Image needs a first frame and a fixed duration. Auto aspect ratio needs a first frame.

<figure><img src="https://3654894688-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2Fnao4gCYbPZ84l9ZFhw80%2Fimage.png?alt=media&amp;token=33519f94-c257-4cd5-bccc-3fa61896a131" alt=""><figcaption></figcaption></figure>

<figure><img src="https://3654894688-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2FZIUg2G6UoJxoLruclS9t%2Fimage.png?alt=media&amp;token=69597e50-c63e-4b2e-a755-ab5d9ca25d5c" alt=""><figcaption></figcaption></figure>

{% embed url="<https://files.gitbook.com/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2FYLgkjPFNRntZasQWdlqR%2Fltx25_video.mp4?alt=media&token=a074847a-bf96-4f28-bbe0-78b8e5e43e88>" %}
Dolly in
{% endembed %}

{% embed url="<https://files.gitbook.com/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2FlIXcQY3yDRISkIrXRTzP%2Fltx25_video%20(1).mp4?alt=media&token=f828e343-4942-440e-83f3-3f9e7e2e710d>" %}
Dolly Right
{% endembed %}

## Create a talking video with OmniHuman V2 in Lipsync

Lipsync adds ByteDance OmniHuman V2 for generating a talking video from one person image and an audio file. An optional Prompt guides expression, gestures, movement and camera. Supply 4–300 seconds of audio and an image 300–6000 pixels per side, with a width-to-height ratio from 0.4 to 2.5.

In Lipsync, select ByteDance :: OmniHuman V2 in Engine, connect Input Image and Audio File, and optionally add directions in Prompt \[OmniHuman V2 only].

The existing default engine remains PixVerse. Selecting OmniHuman V2 leaves Prompt empty; image and audio must be supplied.

<figure><img src="https://3654894688-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2FLXwcMfqqDXB4Ajp2W4Mz%2Fimage.png?alt=media&amp;token=709ef277-1696-400b-8762-ca1376d4668e" alt=""><figcaption></figcaption></figure>

<figure><img src="https://3654894688-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2FdcOw3oicdIiVaC1uPql0%2Foutput%20-%202026-11-04T164243.096.png?alt=media&amp;token=be9c9e6c-6ef1-4ee8-b148-6dcc450c31ea" alt="" width="563"><figcaption><p>Input image</p></figcaption></figure>

{% embed url="<https://files.gitbook.com/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2FMLWDOKNQoYqtOH7irzIY%2Flipsync_video%20(11).mp4?alt=media&token=d1a43c93-ebd5-4594-8dc6-45ce7166c8dc>" %}
Lipsync Video
{% endembed %}

## Create videos with PixVerse V6 Max

Create a clip from a prompt or animate a first image with PixVerse V6 Max in Text + image to video → Simple. Choose 5, 8, 10, or 15 seconds at 360p–1080p, keep generated audio or turn it off, and use Multi-Clip for prompts describing several shots.

Open Text + image to video → Simple, choose Engine → PixVerse V6 Max, enter a prompt, and run. For image-to-video, attach Reference / First Frame and select Aspect Ratio → Auto to keep its shape. Multi-Clip and Negative Prompt are available in advanced settings; Seed is also available.

The node defaults to Seedance; choose PixVerse V6 Max explicitly. Its new settings default to 5s, 720p and 16:9, with audio on, Multi-Clip off, an empty Negative Prompt and a randomized Seed. With a first image and fixed framing, Outpainting Resize is enabled by default and can add cost; Auto keeps the image shape.

<figure><img src="https://3654894688-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2FWGvI1KTWwjKBTHGoNJIW%2Fimage.png?alt=media&amp;token=9dc6153a-ffc6-4e61-9af5-909b5ad5fdb0" alt=""><figcaption></figcaption></figure>

{% embed url="<https://files.gitbook.com/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2FBl26VNnX8iZE8YT0mVqq%2Fpixverse_v6_max_video.mp4?alt=media&token=6884bdf9-eb75-4ace-9da9-d3b270d56dbb>" %}

## MiniMax H3 Lite and Highspeed

Choose MiniMax H3 Lite or Highspeed in Text + image to video for 4–15-second clips from text, start/end frames, or media references. Lite offers 480P and 768P; Highspeed uses 768P.

Open Text + image to video, choose the MiniMax :: H3 Lite / Highspeed engine, set Generation Mode to Text to Video, and enter a prompt. For Image to Video, supply a start or end frame. Reference to Video needs a reference image or video.

Defaults: H3 Model: Lite; Duration: 5 seconds; Resolution: 768P for Lite; Highspeed has no Resolution control; Generation Mode: Reference to Video; Aspect Ratio: 16:9 for text and Lite references; adaptive for Highspeed references.

<figure><img src="https://3654894688-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2FDccVPz2mbZIomB8kw1mG%2Fimage.png?alt=media&amp;token=2f897199-313e-4e4f-8bec-f996e05fee78" alt=""><figcaption></figcaption></figure>

## Reframe videos with MiniMax H3 Max

Expand an existing clip to a new aspect ratio in Edit video. MiniMax H3 Max Reframe fits the source inside your chosen canvas and generates the surrounding area.

Add Edit video and choose MiniMax :: H3 Max Reframe.

Connect a source clip lasting 0.5–300 seconds, choose Reframe Aspect Ratio and Reframe Resolution, and optionally describe the new surroundings.

<figure><img src="https://3654894688-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2F4cYBdrHiQeEhqidGkqWR%2Frelease048-c1-024-reframe-3.jpg?alt=media" alt="Historical H3 Max Reframe input and retained 16:9/768p result, with the original settings visible. Controls were inspected later during a source-parity rollout; this is not a fresh generation."><figcaption><p>H3 Max Reframe with the retained portrait input, 16:9 canvas and 768p result.</p></figcaption></figure>

## Try a free Lab example, then make your own

On your phone, replay a Lab’s saved example without spending credits, then add your own sentence or photo to make a new result. Labs with a draft step pause at a concept picture so you can review it before choosing whether to finish the remaining steps. A finished result can offer a priced next step, with a before-and-after slider when it produces the same kind of result.

Open a phone Lab that offers Try the free example, then choose Now make your own.

Add your sentence or photo. If the Lab offers Make a draft, review the picture and choose Finish it, Change something or Keep the draft.

On a finished result, choose an offered next step. If a comparison appears, drag the divider or use its arrow keys, then choose which result to keep.

The free example replays prepared pictures and can be skipped to the result. Making your own picture draft or taking a next step runs the Lab and can spend credits. Controls depend on the Lab; Draft ready waits for Finish it. Any displayed \~credit figure is an estimate, and a price may be absent.

<figure><img src="https://3654894688-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2Fc3pXfC8XgYV9fYMPE5Ma%2Frelease048-c1-025-phone-lab-free-example.png?alt=media" alt="A saved Lab example with a free replay button. Controlled interface example; making your own result uses credits." width="375"><figcaption><p>A saved Lab example with a free replay button. Illustrative interface with prepared artwork; displayed credit and time estimates are not guarantees.</p></figcaption></figure>

<figure><img src="https://3654894688-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2FonmjCXk7PTCgOyFZvNrl%2Frelease048-c1-025-six-helper-08.jpg?alt=media" alt="Review a picture draft, find it in Runs, and reopen it before choosing Finish it. Illustrated with retained results."><figcaption><p>Review a picture draft, find it in Runs, and reopen it before choosing Finish it. Illustrated with retained results.</p></figcaption></figure>

<figure><img src="https://3654894688-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2FT6CuO4tkYzpJ7lGBpJ3y%2Frelease048-c1-025-six-helper-09.jpg?alt=media" alt="Compare the original treasure-chest mesh with its low-poly result, then choose what to keep. Illustrated with retained results."><figcaption><p>Compare the original treasure-chest mesh with its low-poly result, then choose what to keep. Illustrated with retained results.</p></figcaption></figure>

## Create images with Qwen Image 2.1 Pro and Turbo

Text to image now offers Alibaba :: Qwen Image 2.1 Pro and Alibaba :: Qwen Image 2.1 Turbo. Choose a 1K or 2K output option and automatic or fixed aspect ratio.

<figure><img src="https://3654894688-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FR7boiMixMhR4q36Ns33Y%2Fuploads%2F31nTAAv0zocxEYmxtnKt%2Fimage.png?alt=media&amp;token=c16d1902-25a9-45c5-9c08-e5f8aec51077" alt="" width="540"><figcaption></figcaption></figure>


---

# 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-0.48.0.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.
