What are TikTok’s new AI avatar features for creators and brands?
Cover Photo Major News from TikTok, Runway, DeepSeek, Google's DeepMind and Autify

TikTok Introduces AI Avatars and Dubbing Tool for Ads

TikTok has unveiled a groundbreaking update, introducing generative AI avatars of creators and stock actors for branded content and ads. The new “Custom Avatars” allow creators and brands to create multilingual representations, expanding global reach and collaborations. Additionally, “Stock Avatars” offer pre-built options with diverse actors for a human touch in content. Despite concerns from Hollywood actors over AI usage, creators retain control over their likeness and usage terms.

Moreover, TikTok’s new AI Dubbing tool enables translation into 10 languages, facilitating global communication for creators and brands. This launch coincides with TikTok’s success in driving purchases through the platform, with 61% of users making purchases directly or after viewing ads. These features are part of “TikTok Symphony,” the platform’s suite of ad solutions powered by generative AI, supporting marketers in scriptwriting, video production, and asset enhancement. 

Runway Unveils Gen-3 Alpha: Enhanced Video-Generating AI

Runway introduced Gen-3 Alpha, a new AI model for generating video clips from text descriptions and images. Gen-3 offers improved speed, fidelity, and control over video creation, focusing on expressive human characters and diverse actions. Despite limitations like 10-second clips, Runway plans to release more models in the future. The training data sources are undisclosed, aligning with industry practices. Runway collaborates with artists to address copyright concerns and implements safeguards like moderation systems and provenance authentication. Partnerships with entertainment organizations aim to customize Gen-3 for specific needs. Challenges remain in achieving consistent video output aligned with creators’ intentions. Runway’s investments in generative AI technology, supported by major investors like Google and Nvidia, highlight its dedication to the creative industry. 

DeepSeek Coder V2: Open-Source Coding Model Outperforms GPT-4 Turbo

Chinese AI startup DeepSeek introduces DeepSeek Coder V2, an open-source MoE code language model surpassing GPT-4 Turbo and other closed-source models. With support for 300+ programming languages, it excels in coding, math tasks, and general reasoning. DeepSeek’s mission to unlock AGI mysteries led to the development of this advanced model, expanding language support to 338 and context window to 128K for complex coding tasks. Achieving high scores in benchmarks like MBPP+, HumanEval, and Aider, DeepSeek Coder V2 offers strong performance in language understanding and reasoning tasks, challenging closed-source models like GPT-4o and Claude 3 Opus. Available under an MIT license, users can access the models via Hugging Face or API for research and commercial use. Interact with DeepSeek Coder V2 via chatbot to explore its capabilities.

DeepMind Introduces V2A: AI Soundtrack and Dialogue Generator for Videos

DeepMind, Google’s AI research lab, unveils V2A technology to create soundtracks for videos, enhancing AI-generated media. The diffusion model behind V2A is trained on sounds, dialogue transcripts, and video clips to synchronize audio with visual scenes. DeepMind’s SynthID technology watermark ensures authenticity. While V2A shows promise in bringing movies to life, it has limitations in handling artifacts and generating high-quality audio. DeepMind prioritizes safety and collaboration with creators before public release. The technology targets archivists and filmmakers, posing potential disruptions to the entertainment industry.

Autify Launches Zenes: AI Agent Revolutionizing Software Quality Assurance

Autify, a San Francisco- and Tokyo-based startup, introduces Zenes, an AI agent for software quality assurance, aiming to enhance productivity and streamline workflows for software engineers. The $13 million Series B funding, led by Global Capital Partners and LG Technology Ventures, brings Autify’s total raised funds to $30 million. Co-founded in 2016 by Ryo Chikazawa and Sam Yamashita, Autify addresses software testing challenges with its platform Autify NoCode, catering to users in Japan and South Korea. Zenes, tailored for U.S. customers, automates test case generation, reducing creation time by 55%. With plans to expand AI capabilities and enter the Korean market through a partnership with LG CNS, Autify envisions generative AI enhancing human creativity in software development and quality assurance processes. The startup serves customers in 16 countries, including DeNA, NEC, NTT Smart Communication, Yahoo, and ZoZo, with a staff increase from 30 to 100.

Frequently asked questions

TikTok has introduced two new AI avatar tools: Custom Avatars and Stock Avatars. Custom Avatars allow creators to make multilingual versions of themselves for branded content, while Stock Avatars provide pre-built diverse actor options. The platform also launched an AI Dubbing tool that can translate content into 10 different languages. These features are part of TikTok Symphony, helping marketers create more engaging content while giving creators control over their likeness usage.
Runway’s Gen-3 Alpha is an advanced AI model that generates video clips from text descriptions and images with improved speed, fidelity, and control. It specializes in creating expressive human characters and diverse actions, though currently limited to 10-second clips. The model includes safeguards like moderation systems and provenance authentication, making it particularly valuable for entertainment industry professionals seeking customized video content solutions.
DeepSeek Coder V2 is an open-source MoE code language model that outperforms GPT-4 Turbo and other closed-source models. It supports over 300 programming languages and excels in coding, math tasks, and general reasoning. With a 128K context window and high scores in benchmarks like MBPP+ and HumanEval, it offers superior performance for complex coding tasks and is freely available under an MIT license.
V2A is DeepMind’s new AI technology that generates soundtracks and dialogue for videos. It uses a diffusion model trained on sounds, dialogue transcripts, and video clips to create synchronized audio with visual scenes. The technology includes SynthID watermarking for authenticity verification and is primarily targeted at archivists and filmmakers, though it currently has some limitations in audio quality generation.
TikTok’s AI Dubbing tool is transforming content creation by enabling automatic translation into 10 different languages, making it easier for creators and brands to reach global audiences. This feature has contributed to TikTok’s success in driving purchases, with 61% of users making purchases directly or after viewing ads on the platform. It’s particularly valuable for international marketing campaigns and cross-cultural content distribution.
Zenes is an AI agent developed by Autify that revolutionizes software quality assurance by automating test case generation. It reduces test creation time by 55% and streamlines workflows for software engineers. The tool is particularly designed for U.S. customers and represents a significant advancement in automated testing solutions, helping companies improve their software development processes more efficiently.
Both Runway and TikTok have implemented various safeguards to address AI content creation concerns. TikTok ensures creators maintain control over their likeness and usage terms, while Runway collaborates with artists to address copyright issues and implements moderation systems and provenance authentication. Both platforms focus on responsible AI development while providing innovative tools for content creation.
Picture of Gor Gasparyan

Gor Gasparyan

Optimizing digital experiences for growth-stage & enterprise brands through research-driven design, automation, and AI