Sora is so easy to use. Early this morning, the internal beta version of OpenAI's video generation tool Sora was leaked because some early testers (artists) were dissatisfied with OpenAI. The artist used the "most extreme" method to obtain the shortest moment of universal benefit. An open letter exposed by netizens shows that OpenAI invited hundreds of artists to participate in bug testing, feedback and experimental work. Although these artists contributed a large number of works free of charge, only a few of their Sora works were selected for screening through competition.

Website: https://huggingface.co/spaces/PR-Puppets/PR-Puppet-Sora And, compared to the huge public relations and marketing value that OpenAI gets from this, the compensation they give to artists is minimal . They felt like they were caught in an “art whitewashing” campaign, in which each work had to be reviewed and approved by the OpenAI team before being shared, ignoring their demands for creative expression and criticism. What artists want is creative freedom, but OpenAI gives them conditional exhibition opportunities, and even becomes Sora's endorsement tool. Outraged, some artists leaked Sora's API on Hugging Face, the world's largest open source community platform, causing the tool to be briefly made available to the public for several hours.

The artists also wrote in their open letter: We are releasing this tool to give everyone the opportunity to experience what approximately 300 artists have received: free, unlimited access. We are not opposed to using AI technology as an artistic tool (if we were, we wouldn’t be invited to participate in the program). What we disagree with is the lack of artist-friendly support in the way this artist program is being implemented, and the direction in which the tools are being developed to favor corporate interests ahead of a possible public release. We hope that by sharing this information with the world, OpenAI can move beyond PR stunts and become more open. Judging from the current leaked information, Sora supports the generation of videos in three resolutions: 1080p, 720p, and 360p, with a duration of 5-10s. And the model is codenamed Turbo. Foreign media The Information previously reported that there are multiple versions of Sora models, so it is not ruled out that Sora will have other higher/lower versions when it is officially released.

After the incident, OpenAI took quick action, shutting down all invitation access and banning its Discord channel. Although it has only been open for a few hours, users are suspected of generating a large number of videos through Sora. Let’s take a look at these works.

Do you still remember that classic demo walking on the streets of Tokyo? Some netizens used Sora to re-createOne version, the presentation effect is stunning. However, if you look carefully, you can find that there are six finger flaws in one frame of the picture.

In the background of ice and snow, a cute little bear guards its Coca-Cola.

Pay attention, there is a unique logo from OpenAI in the lower right corner, moving from Tokyo to Paris. The character in the picture is wearing a yellow raincoat and standing on a crowded street.

The classic noodle-eating test also failed.

The puppy flutters in the snow, and the accuracy of dynamically capturing the moment and animal behavior is good.

Cat and mouse, this time it’s really not Tom and Jerry.

Two faces of the robot appeared. One second they were threatening humans, but the next second they were vulnerable.

Why does the monkey still feel reluctant when playing with toys?

Coincidentally, just two months ago, foreign media The Information reported that OpenAI was training a new version of Sora with the goal of generating higher quality and longer video clips. To do this, they needed to collect millions of hours of high-resolution video data covering a variety of styles and topics. Earlier, Sora was exposed that the actual effect was not as good as expected.

The video generation speed was slow. It initially took more than 10 minutes to generate a short film of about 1 minute;

It was difficult to use and the movie Producers need to generate hundreds of clips before finding one that is usable;

It is difficult to maintain a consistent style, and objects and characters are difficult to keep consistent in different clips;

There are errors in physics and anatomy . Judging from the information currently exposed, the performance of the current Sora version is indeed remarkable, but there is still room for improvement in aspects such as image quality. Moreover, Sora's similarly generated videos are still the "mime" that is often criticized, and it seems that it is still far away from being promoted to the altar of "reality no longer exists". We look forward to the subsequent official version being released to bring us more surprises.

Just now, according to foreign media Techcrunch, OpenAI issued a formal response to the leak. An OpenAI spokesperson noted that Sora is still in beta and the company is working to find a balance between creativity and security to prepare for widespread use in the future. The spokesperson also mentioned that the hundreds of artists participating in Alpha testing have made important contributions to the development of Sora, and their feedback has helped prioritize new feature development and security measures. They specifically noted that the participation of these artists was entirely voluntary, with no mandatory feedback or usage requirements. OpenAI said they are pleased to provide free access to these artists and will continue to support them through grants, events, and other programs. The company firmly believes that AI can be an important tool for creative work and is committed to ensuring that Sora is both functional and safe. In addition, the spokesperson also emphasized that artists are usingWhen using Sora, you need to follow the "responsible use" principle and strictly keep confidential information during development.

However, OpenAI does not clearly define what "responsible use" is and what information is confidential. A paradoxical reality is that in order to fight for openness, OpenAI artists have to use leaks to fight against closure. Looking back on this turmoil, considering OpenAI’s past performance, it is even difficult for us to judge whether this so-called “artist leak incident” was a premeditated public relations and marketing method. If this is indeed an unexpected leak, the unexpected publicity may make OpenAI reconsider the release date. Now that it has been exposed, let’s take advantage of the situation and make Sora’s official debut in advance. Sit online to see if Sora is released this week.

Attached is a version of the open letter We were given access to Sora and committed to participating as early testers, red teamers, and creative partners. However, we believe this is actually being lured into engaging in so-called "artistic whitewashing" to claim to the world that Sora is a useful tool for artists. Artists are not your free R&D resources! We are not yours: free bug testers, PR puppets, training data or validation tokens. Hundreds of artists provide pro bono labor to a $150 billion company through bug testing, feedback, and experimental work. And among these free contributions, only a handful of artists were selected through competitions to showcase films created with Sora. The compensation for these artists pales in comparison to the massive PR and marketing value OpenAI derives from it. Refuse to normalize the exploitation of artists! Don’t allow billion dollar brands to exploit artists for free R&D and PR! Additionally, each output requires approval from the OpenAI team before it can be shared. This early access program seems more about PR and advertising than actually supporting creative expression and criticism. We're releasing this tool to give everyone a chance to experience what approximately 300 artists have received: free, unlimited access. We are not opposed to using AI technology as an artistic tool (if we were, we wouldn’t be invited to participate in the program).

What we do not agree with is the lack of artist-friendly support in the way this artist program is being implemented, and the direction in which the tools are being developed to favor corporate interests ahead of a possible public release. We hope that by sharing this information with the world, OpenAI can move beyond PR stunts and become more open. We call on artists to use non-proprietary tools: open source video generation tools allow artists to experiment on the cutting edge without serving commercial interests or becoming corporate PR tools. We also invite artists to train their own models using their own datasets. Some open source video tools available include:

CogVideoX

Mochi 1

LTx Video

Pyramid Flow However, we also realize that not everyone has this resource. Therefore, we call for fair compensation of artists for their expression. Happy creating, some of the Sora Alpha project artists.