2.7-RC1

[hacksider]

✨ New Features

Live Modes

• Decart Live — Cloud-powered realtime diffusion. Select decart from the Model dropdown, enter your API key and prompt, and stream your webcam through Decart's cloud GPUs. Supports both Lucy 2.0 and Lucy 2.1 variants.
• FLUX Live (prompt-edit) — Real-time prompt-driven face editing at 30fps, powered by FluxRT. Includes a Quality dropdown for performance tuning. Available on RTX 5090 and RTX 6000-class GPUs only.
• ReSwapper — New face-swap engine, selectable from the Model dropdown alongside Inswapper and Hyperswap.

UI / UX

• Classic / Modern View Toggle — One-click switch between two complete UI layouts, accessible from the status bar.
• Language Dropdown — Full UI localization with 5 supported languages:
  -- English
  -- 中文 (Chinese)
  -- 日本語 (Japanese)
  -- Français (French)
  العربية (Arabic -
• RTX Upscaler — New toggle in the settings panel that uses RTX hardware to upscale output beyond source resolution. Surfaced automatically on compatible GPUs.

🔧 Improvements

• Window Projection — Now automatically resizes during live sessions.
• Window Projection — Added a dedicated close button on the overlay.
• Live Mode Cleanup — Floating player overlay is now hidden during live streaming for a cleaner viewport.

🌍 Localization

• Added language files for English, Chinese, Japanese, French, and Arabic. Switch languages from Other Settings → Language. Restart required to apply.

📋 Notes

• Decart Live requires an active Decart API key and is live-only mode.
• FLUX Live is hardware-gated to RTX 5090 / RTX 6000-class GPUs for performance reasons.
• Language changes require an application restart to take effect.
• All existing features from 2.7-build (Inswapper, Hyperswap, Face Enhancer, Lip Mask, Poisson Blend, RIFE Interpolation, LUT, Virtual Camera, etc.) remain fully supported.

Available Exclusively as a Quickstart on our Website

---

This discussion was created from the release 2.7-RC1.

[Najeebullah111]

Great

[bartoon60]

Disappointed: No DirectML / AMD support in 2.7-RC1?tle: Disappointed: No DirectML / AMD support in 2.7-RC1?

Good job on the new features, but why did you secretly drop DirectML support? --execution-provider dml is completely gone from the arguments, and the app crashes on AMD cards (Sapphire).

Your release notes say "All existing features remain fully supported", but you forgot that a huge part of the community runs on AMD. Not everyone has an RTX 5090. Please bring DirectML back for the stable release.

Thanks

[hacksider]

You don't need that on RC's as it automatically detects your gpu. Also, there's a ticket support on the site. As you can see, you can use the installer on both AMD and NVIDIA gpu.

[hacksider]

Also, you can use the support on the website. You will just fragment your issues and it will not be included on the RC3 if you post it here. This is for the github version.

[bartoon60]

Your official .bat fails because your requirements.txt forces the installation of onnxruntime-gpu (Nvidia CUDA) by default, which makes Python completely blind to AMD cards under Windows.

For Sapphire/AMD owners, your run-directml.bat file is totally useless unless we manually run pip uninstall onnxruntime-gpu followed by pip install onnxruntime-directml inside the venv. Here is a free fix for your RC3.

[hacksider]

Where did you get the RC2? Are you sure you're using the QuickStart? The bat is just a single click installation.

[safolabi953-gif]

can this work on gpu rtx 4050, 32gb ram, 512gb ssd? 6gb graphics card

[hacksider]

yes

[safolabi953-gif]

i just installed it and its showing version 2.1.5.

[safolabi953-gif]

i dont understand if i did it wrongly but its showing me version 2.1.5 and no addictional features

[hacksider]

Did you read the whole article? You can get the QuickStart as written on the bottom.

[David19-glitch]

I will like to know, with the Decart live....Do i get the watermark effect when i buy your own ?

[hacksider]

We don't have control with the API itself. Consider our app as a container of what your API is using.

[nanaleowan]

is there any way to help with the submitting the request because any reason you give they reject after using your id and face verifications too

[hacksider]

I am not affiliated with decart in any way. And decart is a third party service.

[Wspjason]

Which mode do you suggest using for real tome faceswap when video chatting

[hacksider]

Inswapper + gfpgan 256 usually will suffice.

[leyndoza-web]

Which version is the deeplivecam- lite version?

[hacksider]

We haven't posted anything on github as of this moment but it is included with the QuickStart

[leyndoza-web]

[hacksider]

The lite version doesn't need a gpu but requires decart api which transforms your whole body

[leyndoza-web]

Yes I understand. But I still don't have money to buy the lite version. Instead which version I can use for now for my laptop? To stop from lagging. Thankyou

…

On Fri, Jul 3, 2026 at 1:49 PM Kenneth Estanislao ***@***.***> wrote: The lite version doesn't need a gpu but requires decart api which transforms your whole body — Reply to this email directly, view it on GitHub <#1837?email_source=notifications&email_token=CA63DGL645M2VUR25T24DCL5C56RNA5CNFSNUABIM5UWIORPF5TWS5BNNB2WEL2ENFZWG5LTONUW63SDN5WW2ZLOOQXTCNZVGIYTQMBYUZZGKYLTN5XKOY3PNVWWK3TUUVSXMZLOOSWGM33PORSXEX3DNRUWG2Y#discussioncomment-17521808>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/CA63DGK7X4LMPMYF224CTQT5C56RNAVCNFSNUABIKJSXA33TNF2G64TZHM3DSNJYGY2DKMJVHNCGS43DOVZXG2LPNY5TCMBRGAZDGMRVUF3AE> . Triage notifications, keep track of coding agent tasks and review pull requests on the go with GitHub Mobile for iOS <https://github.com/notifications/mobile/ios/CA63DGPHCIZE2ODON7Z6R4D5C56RNA5CNFSNUABIM5UWIORPF5TWS5BNNB2WEL2ENFZWG5LTONUW63SDN5WW2ZLOOQXTCNZVGIYTQMBYUZZGKYLTN5XKOY3PNVWWK3TUUVSXMZLOOSVGM33PORSXEX3JN5ZQ> and Android <https://github.com/notifications/mobile/android/CA63DGNYX3UWNVHEHDNOEGD5C56RNA5CNFSNUABIM5UWIORPF5TWS5BNNB2WEL2ENFZWG5LTONUW63SDN5WW2ZLOOQXTCNZVGIYTQMBYUZZGKYLTN5XKOY3PNVWWK3TUUVSXMZLOOSXGM33PORSXEX3BNZSHE33JMQ>. Download it today! You are receiving this because you commented.Message ID: ***@***.*** com>

--

[hacksider]

All the offline versions require a good gpu to function properly so I can't recommend anything aside from getting a capable laptop to run the GitHub version.

[ENHANCEITSOLUTIONS]

Hi, for the next version you are planning to release, can we expect improvements that reduce flickering or artifacts around the chin, ears, and face edges during face swaps? what will be the best setting for RTX5090 with 96 Gb ram to get maximum fps and stable video output for live cam.

Also, are there any planned improvements around output quality, consistency, and responsible AI safeguards, such as clearer detection handling, watermarking options, or compatibility with content authenticity standards? Some current outputs are still being flagged by deepfake detection tools, so it would be helpful to understand your roadmap for making results more stable, natural, and compliant with platform policies.

[hacksider]

We are always improving the quality and speed every iteration.