Monday, June 17, 2024
- Advertisement -

    Latest Posts

    Stability AI Introduces ‘Stable Audio Open’ Amid Copyright Controversies

    Stability AI has launched ‘Stable Audio Open’, an open-source AI model for generating up to 47 seconds of audio samples and sound effects based on text prompts. According to the company blog post, users can fine-tune the model using their own custom data.

    Stable Audio Open can be used to create drum beats, instrument riffs, ambient sounds, foley recordings, and other audio samples for music production the company claims. Unlike Stability AI’s Stable Audio—the company’s commercial service for creating full-length music tracks—Stable Audio Open cannot generate full songs or melodies.  Further, the blog post informed that the new model was trained on audio data from Freesound and the Free Music Archive.

    Stability AI is facing a lawsuit by Getty Images for alleged copyright infringement. Getty Images has alleged that the Stable Diffusion creator copied over 12 million images from their database without Getty Images’ permission or compensation “as part of its efforts to build a competing business.” In another copyright infringement matter last year, three artists, Sarah Andersen, Kelly McKernan, and Karla Ortiz, filed a class action lawsuit against Stability AI, Deviant Art, and Midjourney, alleging that their AI tools have infringed upon the works of thousands of artists.

    More about the training datasets used to create Stable Audio Open:

    Amidst ongoing copyright complaints, Stability AI launched Stable Audio in September 2023 raising concerns about the data used to train its new audio tool. In releasing the current Stable Audio Open tool, the company has made the model weights available on Hugging Face, an open-source AI community and platform.

    AI model weights are numerical parameters that define a machine learning model’s internal structure and decision-making logic, allowing an AI system to learn patterns from data and make predictions or decisions.

    According to the details available, Stable Audio Open is trained on a dataset consisting of 486492 audio recordings, where 472618 are from Freesound and 13874 are from the Free Music Archive (FMA). Stability AI claims that all audio files are licensed under Creative Commons Zero (CC0), Creative Commons Attribution (CC BY), or Creative Commons Sampling Plus (CC Sampling+).

    Secondly, to prevent copyright-associated risks, the company claims it has conducted “an in-depth analysis” to filter out copyrighted content from the datasets before using it for training the model. The company has also cautioned against the use of the model for downstream applications without further evaluation and risk mitigation.

    However, whether the measures employed to prevent copyright infringement are foolproof remains unanswered. In 2023, Google also launched a text-to-audio generator called MusicLM in February. Back then MediaNama spoke to copyright lawyers to understand the myriad ways in which the audio works produced by AI generators infringe on copyright. You can read that story here.

    Also Read

    The post Stability AI Introduces ‘Stable Audio Open’ Amid Copyright Controversies appeared first on MEDIANAMA.

    Latest Posts

    - Advertisement -

    Don't Miss

    Stay in touch

    To be updated with all the latest news, offers and special announcements.