
Spektral
Bredgade 75, 4. sal, Copenhagen, Hovedstaden, 1260, Denmark
Overview
Spektral is a Danish visual-effects startup building real-time green-screen technology that uses machine learning and spectral graph theory to separate people and objects from backgrounds. The team originally worked on still-frame cutouts under the name CloudCutout and is now moving into live video. Spektral has invested heavily in research and development and is accumulating patents rather than shipping a traditional product today. The company aims to enable uses in production, advertising and potential integrations with platforms like Snapchat or Messenger. Future work the team has flagged includes more complex video editing and object manipulation. To accelerate development and credibility, Spektral added Danny Lange, formerly head of machine learning at Unity, to its board. CloudCutout is a Copenhagen-based startup applying machine learning to automate image background removal, initially targeting school photography and product shots. Its algorithm, developed by co-founder Toke Jansen (machine-learning PhD), was trained on more than 100,000 professional cutouts and uses spectral graph theory and neural networks to compute alpha channels. The company charges $0.25 per image with a 50,000-images-per-year minimum, positioning itself to undercut the standard bulk cost (about $0.50). CloudCutout is launching a school cutouts service initially limited to green-screen knockouts and plans to expand to other backgrounds. Later this year it intends to release a Photoshop plug-in (likely freemium with subscription tiers) so designers can manually tweak results. To scale processing, CloudCutout runs neural computations on IBM SoftLayer (NVIDIA Tesla K80 GPUs) while hosting remaining infrastructure on AWS.
- Total raised
- $3M
- Funding rounds
- 2
- Latest round
- Equity
- Latest activity
- Jun 2017
Industries
- Artificial Intelligence (AI)
- Digital Media
- Machine Learning
- Video Editing
Recent funding
Equity
Jun 2017
$3M
Seed
Aug 2015
$1M
Team
No current team members are available.