Fivelabs, on one page.
For investors, accelerators, and grant, credit and inception programs. Everything here describes software that exists and numbers we have measured.
Summary
Fivelabs is an early-stage company building Fivelabs Studio, an AI video studio that turns an idea or a script into a finished, narrated video. It makes vertical Shorts of 30–60 seconds and long-form videos of 1–20 minutes, in Hindi, English, Spanish, Portuguese, French, German and Italian.
We render with the open-weights MiniMax H3 video model on our own GPU render stack, using GPUs rented from Nosana’s decentralised network. Measured, a 6-second clip costs about $0.015 in GPU time; at the hosted API’s price of about $0.08 per second, the same clip is $0.48.
We are building for creators, educators and media teams, starting with Hindi-language content.
What is built
- A four-step creation flow: idea or script and visual direction; narrator and length; style (15 image styles) and upgrades; review, generate and export in 1080p.
- An AI writer (Claude) with an enforced story structure from hook to payoff, writer’s notes with sources, and an owner review before anything renders.
- A render stack for MiniMax H3 with keyframe-first portrait rendering, automatic orientation review, retakes and automatic retries.
- Long-form tools: chapters, AI images with Ken Burns motion, moving images, 19 motion-graphics templates, cinematic intros, avatars, music, captions and film-look grading.
- An editor with a timeline, trim, replace, reorder, crossfades, music upload, caption styles, undo and redo, and four thumbnail concepts per video.
- 33 curated narrators, ElevenLabs natural voices and custom voices.
- A REST API, an MCP server with 26 tools and OAuth 2.1, and publishing to YouTube and Instagram.
- Credits that meter real GPU cost, with budgets. An independent security audit and more than 90 automated tests.
Cost structure
| Item | Measured or quoted |
|---|---|
| RTX A6000 on Nosana | about $0.36 per hour |
| One 6-second, 480×864 clip | about 2.5 GPU minutes, about $0.015 |
| Per second of generated video | about $0.0025 (GPU time) |
| Visuals for a 5-minute long-form video | about $0.20 (GPU time) |
| Hosted MiniMax H3 API | about $0.08 per second of video |
GPU time only: narration, writing, storage, idle GPU time and engineering are excluded. Resolution and settings differ from the hosted API, so treat the comparison as an order of magnitude, about 32× per second of video.
Why it is hard to copy
- Cost. Running open weights on rented, decentralised GPUs means our cost per second of video follows GPU minutes, not a per-second API price.
- Craft built into the pipeline. Story structure, portrait-orientation handling, retakes and assembly are encoded in the product, not left to the user.
- Language. Hindi narration with English or Hinglish captions from day one, plus six more languages.
- Distribution through AI assistants. The MCP server lets people make videos from inside Claude, ChatGPT and other MCP clients.
Stage and what we need
Fivelabs is early stage and actively building. We are raising, and applying to accelerator, grant and inception programs.
- Investment to grow the team and take the studio to more creators.
- Accelerator places and mentorship.
- GPU and cloud credit programs, which go straight into rendering.
- Distribution and content partners, especially in education and Indian-language media.
Founder details, a live demo and further material are available on request.
Contact
Write to hello@fivelabs.xyz with “Investor” or the program’s name in the subject.