Creative Workflow Roundup: Runway’s Act-One Character Animation Tool, Midjourney Updates, Claude 3.5’s Computer Use, Stability AI’s New Models, Eleven Labs’ New Features and More

In This Week’s Roundup: Runway’s Act-One debuts with a powerful character animation tool, Anthropic unveils an upgraded Claude 3.5 and an innovative computer-use feature, and Midjourney launches fresh image editing features. We’ll also cover Stability AI’s latest model variants, Eleven Labs introducing new voice AI features, Timbaland’s remix contest with Suno, and a strong statement on AI ethics signed by 11,500 creatives.

Runway’s Act-One: A New Creative Tool for Character Animation

The News: Runway has introduced Act-One, a tool integrated into its Gen-3 Alpha platform that focuses on generating expressive character performances. This tool allows users to animate digital characters by capturing an actor's video performance with minimal equipment, such as a simple camera setup. Act-One leverages generative AI to faithfully translate eye-lines, micro-expressions, and delivery pacing from a live actor into a generated character. It can adapt performances into different character designs while preserving realistic expressions. This technology aims to streamline the traditionally complex facial animation workflows, which often require motion capture equipment and manual face rigging. The tool will be available to users through a phased rollout and features a suite of safety precautions, including content moderation and rights verification measures.

Lab Notes: Act-One brings what looks like professional-grade character animation to producers in a fraction of the time it would typically take to get similar outputs. What once took weeks can now be achieved in an afternoon. But it’s not just about having the right tech—you still need solid acting skills to bring these characters to life. The actor’s performance is at the core of the final output, proving that while AI can handle the technical execution, human creativity remains essential in driving expressive storytelling. It’s an exciting step forward for indie producers and small teams aiming to produce high-caliber animated content.

Midjourney Expands Features with New Image Editor and Re-Texturing Capabilities

The News: Midjourney has announced the launch of two new features: an updated image editor that supports uploaded images and a new re-texturing tool. The re-texturing tool allows users explore various materials, surface effects, and lighting options while keeping the structure of the initial image intact. The rollout is limited to users with yearly memberships, those who have subscribed for the past 12 months, or users with at least 10,000 generated images. Midjourney emphasized a cautious release approach, involving human moderators and AI-based moderation systems to manage the new capabilities smoothly.

Lab Notes: Midjourney’s new image editor and re-texturing features are expanding the platform’s creative possibilities. These features add versatility for those looking to make detailed adjustments to their images within the Midjourney web interface. Adobe Firefly has previously launched similar features, especially with its “Composition Reference” option which enables users to generate images with new styles while matching the outlines and depth of a reference image. I think Midjourney delivers higher-quality outputs (arguably, excluding Flux) verses alternative models, which could make their new re-texturing tool and image editor particularly appealing.

Anthropic Expands Claude’s Capabilities: Sonnet Upgrade, New Haiku Release, and Groundbreaking Computer Use Feature

The News: Anthropic has announced an upgraded version of its Claude AI model, Claude 3.5 Sonnet, alongside a new variant, Claude 3.5 Haiku. The upgraded Claude 3.5 Sonnet comes with broad improvements, especially in coding tasks, where it surpasses its previous version and even leading models on several benchmarks. In addition to this upgrade, Anthropic unveiled Claude 3.5 Haiku, a faster and more cost-efficient model that maintains the same pricing and speed as its predecessor while delivering significant improvements in capabilities.

The even more intriguing release is the new "computer use" feature, currently in public beta. This capability allows Claude to interact with computers—by viewing screens, moving a cursor, typing text, and more—enabling AI to execute tasks that involve standard user interfaces. Although in its early stages and prone to occasional errors, this new feature has the potential to streamline and automate complex workflows.

Lab Notes: Claude 3.5 Sonnet's new “computer use” feature feels like a pivotal step in making AI a more active participant in creative and technical workflows. The idea of an AI navigating computer interfaces isn’t new—it reminds me of Open Interpreter’s efforts—but the advancements Anthropic has made are noteworthy.

If refined effectively, these capabilities could transform creative workflows. I see potential in using these tools to interact with complex creative software via voice commands, automate repetitive tasks in video editing, script management, or project planning, ultimately letting creatives focus more on high-level decision-making. However, with both Anthropic’s computer use (and Open Interpreter’s) software still in development, there’s a risk of errors, so it’s wise to approach with patience and caution.

Stable Diffusion 3.5 Released with Customizable Model Variants

The News: Stability AI has launched Stable Diffusion 3.5, offering new model variants: Large, Large Turbo, and the upcoming Medium model (releasing on October 29th). The Large model has 8 billion parameters for better image quality and accuracy, while Large Turbo is a faster version that maintains high quality with fewer processing steps. The Medium model is designed to work efficiently on standard hardware and balances quality with ease of customization. These models are free to use under the Stability AI Community License for non-commercial and smaller commercial projects.

Lab Notes: Stable Diffusion 3.5 adds more options to an already crowded field of image generation models. With its focus on customizability, speed, and accessible licensing, it’s clear Stability AI wants to attract a broad user base, from researchers to startups. But with so many new models like Flux, Adobe Firefly, and Midjourney also evolving quickly, the real differences between them are getting harder to distinguish. I still rely on Midjourney and Firefly for most of my work, as they consistently deliver quality and have familiar, refined workflows. That said, the improvements in Stable Diffusion 3.5 are worth noting, especially for those who need flexibility in fine-tuning or are looking for cost-effective solutions.

ElevenLabs Launches New Voice Design Tool

The News: ElevenLabs has updated their Voice Design feature, enabling users to create custom voices using only text prompts. The tool allows for the description of various attributes such as age, accent, gender, tone, and pitch, making it possible to generate unique voices in seconds. Additionally, users can experiment with fictional characters like elves, ogres, and aliens, expanding creative possibilities. The goal of this update is to offer virtually unlimited voice options for both real and imaginary characters.

Lab Notes: The updated Voice Design feature by ElevenLabs is a big step forward from its earlier, more limited version. Previously, users had only a handful of adjustments available before generating a voice. Now, the use of text prompts makes the entire process much simpler and more flexible, allowing for a broader range of customized voices. This change offers a more streamlined approach to creating specific and unique voices, which could be particularly useful in storytelling and content creation.