Clint, can you start by giving us a brief outline of your background in the media industry?
I’ve spent 23 years in broadcast television—editing, producing, and supervising master control. I’ve worked across the whole pipeline: cutting content, managing live operations, and the part most viewers never think about, keeping broadcasts on air and out of trouble from the MCR. Currently, I work casual MCR shifts alongside full-time QC: reviewing content against broadcast specs before it goes to air. That background is really where Long Play comes from. The encoding choices, the instinct to protect the master, the expectation that quality is non-negotiable—none of that came from a startup playbook. It came from years of sitting in master control knowing exactly what goes wrong when you cut corners.
What led you to create Long Play?
Frustration, honestly. YouTube has pushed further into algorithmic recommendation, to the point where the platform feels less like a place to watch things and more like a machine optimising for engagement. Then Vimeo, which had been the obvious home for serious independent work, started drifting after the Bending Spoons acquisition, cutting the free tier and leaving creators wondering what the platform actually stood for. I kept looking for something that just respected the work: no algorithm deciding what gets seen, no platform advertising, no monetising creator data, encoding built to broadcast expectations as a baseline. Nothing fit. Given my background, I knew exactly what doing it properly would look like technically, so I built it myself. Watching what’s happened to Vimeo since has only reinforced that platforms with clear values aren’t a niche—they’re a need.
What is Long Play?
An algorithm-free video hosting platform built for filmmakers and independent creators. No recommendation engine, no platform advertising. Creators retain their masters—stored on Backblaze B2, byte-exact retrieval guaranteed. Encoding is built to broadcast expectations by default. We never train AI on uploaded content. Analytics are minimal: date, time, nothing invasive. Monetisation tools take no cut from the platform. The line I keep coming back to is: ‘The algorithm was never on your side. This is.’ It’s for people who want their work watched and judged on its own terms, not because a feed decided to surface it.
How does it differ from other video hosting platforms?
Most platforms are built around a feed and an algorithm deciding what rises to the top – that’s the entire business model, and everything else follows from it. Long Play has none of that architecture: no algorithmic feed, no follower graph, no recommendation engine, no ad model. Because there’s nothing to optimise attention for, there’s no incentive to harvest behavioural data, and that changes everything downstream.
The practical difference shows up in what we’ve built in place of that infrastructure. Broadcast-informed correction tools built directly into the platform—deinterlace, audio sync correction, HDR-to-SDR tone mapping, film restoration—applied non-destructively, with a 720p preview before the full re-encode commits. No consumer hosting platform comes close to that. And the commitment not to train AI on uploaded content isn’t a policy that might quietly change when it becomes commercially convenient. It’s a permanent design decision.
How do users get their content on the site, and how do viewers watch?
Upload is straightforward. Bring your file or import your existing Vimeo library directly with the built-in importer. Long Play handles encoding: H.264 built to broadcast expectations, optional AV1 mastering pass, HLS adaptive streaming. On the viewer side: scrub preview with sprite-sheet thumbnails, chapter markers, director’s notes, and caption support. Viewers can build a personal library and organise saved content into shelves. Creators can group their work into collections, and Creator plan subscribers can point a custom domain at their portfolio page—the platform effectively disappears behind their own branding if they want that.
Please explain Long Play’s infrastructure in detail and explain why you chose those workflows.
The stack is deliberately unglamorous: Python/Flask, SQLite, Gunicorn on DigitalOcean, with ephemeral encode workers spun up via the DigitalOcean API and torn down when they’re done. Those burst workers handle H.264; AV1 runs on the main server instead—AV1 encodes are slow enough that spin-up overhead on a short-lived droplet doesn’t pay off. Storage is Backblaze B2, with Cloudflare handling CDN, DNS, and bot protection. The Bandwidth Alliance means B2-to-Cloudflare egress is free, which matters when you’re storing byte-exact masters permanently rather than cycling them to cold archive.
Encoding runs through FFmpeg: CRF 18 quality-based H.264 with a 12 Mbps ceiling, AV1 at CRF 24 constrained quality, 192k stereo AAC at 48kHz—with EBU R128 loudness normalisation and automatic BT.709 colour space normalisation applied at encode time. Broadcast timecode, start TC, frame rate, drop-frame flag, is captured at ingest via FFprobe and stored with the metadata.
On top of that, a growing suite of non-destructive post-upload correction tools: crop, rotate & flip, trim, section removal, deinterlacing via yadif, A/V sync offset, VFR-to-CFR conversion, noise reduction, colour correction, speed change, HDR-to-SDR tone mapping, stabilisation, and film restoration. Every one is previewed at 720p before committing to a full re-encode, so the original master is never touched regardless of what corrections get applied downstream.
I built it this way because it’s the stack a broadcast engineer would reach for: predictable, inspectable, and built so quality control doesn’t quietly erode when there’s only one person maintaining it.
Does the platform employ AI in any way?
There’s an AI-generated content flag that creators set at upload, which works the same way a mature content warning does, a disclosure rather than a prohibition. Viewers can filter it out entirely in their settings if they’d prefer. The thinking was: don’t make a value judgement about AI-generated work, just give people enough information to decide for themselves what they want to watch.
Where AI does appear is in optional auto-captioning via Whisper, which runs locally on our own infrastructure when a creator requests it—audio is processed and discarded, never sent to a third party. What we don’t do is use uploaded content to train AI models, and that’s a permanent design decision. The encoding pipeline otherwise runs on deterministic tools, FFmpeg, FFprobe, not models making opaque decisions about anyone’s work.
What are your hopes for Long Play – if we were talking in 12 months, where would Long Play be?
The preservation infrastructure is already running. SHA-256 fingerprinting at ingest, C2PA manifest detection when the source carries one, cryptographic timestamp anchoring via OpenTimestamps to create a tamper-evident record. In 12 months, I’d like that to be visible to creators rather than something quietly happening in the background: a downloadable Preservation Certificate you could keep as a chain-of-custody record if you ever needed to prove a file’s provenance or demonstrate it hasn’t been touched since upload.
I’d also like the Founding Creator cohort filled: the first 100 people on the paid Creator plan with a public video live get a permanent badge, no expiry, no conditions after the fact. More plan tiers are coming, and the free plan isn’t going anywhere. And keep building the way it’s been built: one change at a time, properly tested, by someone who uses the platform themselves and takes ideas directly from the people who do—some of the best changes on the roadmap started as a creator suggestion.
What haven’t I asked you that you think our readers should know about?
Probably that this has all been built by one person, alongside a full-time broadcast career—not a funded startup with a team. Every decision reflects someone who has actually sat in master control worrying about loudness compliance and drop-frame timecode, not a product team optimising for growth metrics.
The free plan exists permanently and will only ever gain features, not lose them. The Creator plan is priced at roughly a cup of coffee a month, available in eight currencies, and that price isn’t moving. We’re not on the trajectory that took Vimeo and YouTube away from the people who cared most about their work. That’s not an accident. It’s the point.
