The Challenge

VEED saw a gap in the market that was fundamental to their own customers - generalist generative video models were weak at doing the one thing that really counts in marketing. Making people talk. As a result, the team at VEED built a suite of proprietary models focused specifically on the talking head problem - from image-to-video, dubbing & repurposing, subtitles, cleaning audio and removing backgrounds.

VEED builds these proprietary models, and developers building video into their own products need the same capabilities. To reach those developers, VEED needed more than GPUs that scale with other companies’ traffic. Each new model needed documentation, a playground, billing and, above all, an audience: developers already building video into their products, reached on launch day rather than built from scratch.

“Being able to get our own models live into our production environments and scale that quickly is really important.”

Sabba Keynejad, Co-founder and CEO, VEED

The Solution

VEED deploys its models as apps on fal Serverless, and fal Serverless is also the backend for VEED's own product. When someone uses one of these features in the VEED editor, the model runs on fal.

The model that powers a feature in the VEED editor also appears in fal's model gallery as a public endpoint, with a playground, API documentation, client libraries for Python and JavaScript, and usage-based billing that fal processes. VEED sets the price and writes the model page, and fal's developers can call the model the day VEED announces it.

VEED launches each model together with fal. For Clean Audio, fal’s model page and playground was available early on for VEED’s pre-release customers. On launch day, fal’s creative team prepared the launch video and also shared VEED’s announcement on X. In September 2026, VEED co-founder and CEO Sabba Keynejad also presented at fal’s Gen Media Conference.

“The last year has been a huge journey for us at VEED - going from a company who had never built a model before, to one that’s able to serve millions of requests with low infrastructure lift - integrating directly into the majority of the leading Generative Media platforms. fal has been fundamental in making that happen for us.”

Josh Goldman, COO, VEED

Lipsync 2.0 went public on fal on July 14, 2026, and Clean Audio on September 28. Clean Audio is the model behind the Clean Audio toggle in the VEED editor, and any fal developer can now call it for $0.0125 per minute of audio. Developers who already use fal for image or video generation find VEED's models in the same gallery and call them with the same API key.

Twelve VEED endpoints are live on fal today:

• Fabric 1.0: image to video, fast and text to video
• Lipsync: Lipsync 2.0 and the original Lipsync
• Clean Audio: speech enhancement
• Video Background Removal: standard, fast and green screen
• Subtitles: burned-in subtitles
• Avatars: text to video and audio to video

Outcome: More than 10,000 developers and teams, and usage that has multiplied since June

• VEED’s models served more than 5 million generations through fal in 2026.
• More than 10,000 developers and teams called a VEED model on fal in 2026, an audience VEED reached without acquiring each customer directly.