Ollanode: hosted video and CDN with the depth of Mux at a fraction of the price.

#The setup: video is too complicated, too fragmented, and too expensive

Every product team that ships video runs into the same three problems within the first month of trying to do it well. Video is too complicated — the pipeline from raw upload to playback touches at least seven distinct concerns (ingest, transcoding, storage, delivery, transcripts, subtitles, the player itself) and there is no single open-source toolchain that handles them coherently. Video is too fragmented — the established hosted vendors each own a slice of the surface, none of them own all of it, and the team ends up running three vendors in parallel with three integrations and three monthly bills. Video is too expensive — at scale, the per-minute encoding fees and per-gigabyte egress fees that define the category make the line item one of the largest in the COGS, often second only to the cloud compute itself.

The conventional answers solve one problem at a time. Mux solves "video is too complicated" by shipping a polished, deep, well-engineered pipeline behind a clean API — and charges premium per-minute rates for the privilege. Cloudflare Stream solves "video is too expensive" by riding Cloudflare's own bandwidth pricing — and ships a feature surface that is meaningfully shallower than Mux on the dimensions teams care about most (transcripts, server-side clipping, player flexibility). Component-level vendors — AWS MediaConvert, CloudFront, Bunny Stream, Cloudinary — solve narrow slices and force the team to handle integration as their own problem.

The third problem — fragmentation — is the one nobody in the category solves, because solving it requires shipping a single product across the entire video-and-CDN surface. Mux owns the video pipeline but not the CDN at any depth. Cloudflare Stream owns the CDN at depth but ships the pipeline as a thinner add-on. Bunny ships both but at the price tier that fits its volume-economics position. Cloudinary ships images first and video second. None of them include edge compute as a first-class layer, which is the thing the modern web player increasingly needs sitting between itself and the cache.

Ollanode exists because the founders watched their own portfolio companies — and a growing crowd of streaming-heavy product teams across India and beyond — run into all three problems at once and end up assembling the same uncomfortable three-vendor stack to solve them. The bet was simple: ship a single hosted platform that does the entire video, CDN, storage and edge-compute surface at depth, with the polish and developer-grade UX of a top-tier product, at a price structure that does not punish success.

#What Ollanode actually is, in one paragraph and then in detail

Ollanode is a fully hosted, cloud-operated platform for video, CDN, storage and edge compute. The mental model is closest to NextDNS for this category: a polished hosted product that you sign up for, configure, point your application at, and use. There is nothing to install. There is no cloud account of yours that the platform requires access to. There is no infrastructure decision to make. You sign up at the dashboard, you create a project, you get an API key, you start uploading video and serving it through the CDN. The platform handles the rest.

Inside the platform there are eight microservices behind sixty-plus REST endpoints, organised into four composable layers: video, delivery, edge compute, and control plane. The layers compose so that a typical workload — an uploaded video, transcoded, transcribed, streamed through a CDN, with an edge function handling auth — touches all four layers through one API key and one set of webhook subscriptions. The pipeline is async by default, fast in the common case, and observable end-to-end in the dashboard.

The hosting model is the same hosting model the rest of the Ollasoftware portfolio uses: cloud-operated by the team, running on Ollasoftware's own infrastructure at the regional pools that serve global traffic, with a transparent pricing surface, a self-serve free tier, and the operational maturity that comes from a team that has been running production AI infrastructure at billion-page-a-month scale for several years. The customer's job is to ship product; the platform's job is to be invisibly reliable underneath it.

Competitively, the platform sits in a specific place. It is not trying to undercut Bunny on the rock-bottom-cost tier (Bunny serves that market well and is the right answer for teams whose only requirement is the lowest bandwidth bill). It is not trying to win the largest enterprise contracts away from Mux (Mux owns that market through depth of relationships and SLAs the team has spent years cultivating). It is trying to be the right answer for the very large middle — series-A to mid-market product teams who want depth, UX and operational maturity comparable to Mux at a price comparable to Stream, with edge compute included rather than bolted on.

#The four primitives, and what each one does

The platform exposes four composable layers: video, delivery, edge compute, and control plane. Each is a family of capabilities behind a coherent API; the layers were designed together rather than acquired or assembled, which is what makes the composition clean.

The video layer is the full pipeline from raw upload to playback. Ingest accepts direct uploads for small files, resumable multipart uploads for large files, and pull-from-URL when the source is hosted elsewhere. Transcoding produces an adaptive HLS ladder with configurable quality presets, H.264 + AAC defaults, loudness normalisation, and MP4 renditions for the use cases that need them. Thumbnails and storyboards generate poster frames and scrubbing sprite sheets with a `storyboard.vtt` modern players consume directly. Transcripts run on a Whisper-powered backend and produce VTT, SRT, and timed-segment outputs. Subtitles and chapters handle the metadata surface with multi-language WebVTT tracks. Server-side clipping cuts sub-clips into brand-new assets in one API call. Playback is signed or public, with short-lived rotating tokens. A drop-in iframe hosted player handles the player surface for teams that do not want to wire their own.

The delivery layer is a multi-tenant CDN with pull zones and storage zones as the two primary objects. Pull zones map any hostname to either an arbitrary upstream or a managed storage zone, with configurable TTLs and cache-bypass prefixes for dynamic paths. Purge is instant and granular — whole-zone in milliseconds, path-level in the same time. Hotlink protection uses HMAC-signed URLs with rotatable secrets. CORS is handled at the edge so the origin never sees the preflight. Storage zones are S3-style buckets with presigned uploads, CDN-fronted by default. Edge rules let you set headers, redirect or block based on ordered conditions.

The edge compute layer runs JavaScript in V8 isolates next to the cache. The handler shape is standard `Deno.serve(handler)`. There are no containers, no cold starts, no build step beyond the JavaScript itself. Push code, invoke at `/__fn/` within seconds, or wire the handler into the pull-zone request lifecycle for auth, A/B and geo-routing. Per-project isolation is structural; metering is per-invocation.

The control plane layer is the operational substrate that the rest inherits. API keys at read, write, admin tiers; rotatable, scoped per project. Webhooks HMAC-signed and retried, with structured events. Analytics for storage, views, status breakdowns, per-day rollups, top videos. Team and roles for scoped multi-user access. Audit log for every mutation. Self-serve auth covering signup, email verification, password reset.

#The video pipeline in depth

The video pipeline is the layer where the platform gets compared most directly to Mux, and it is the layer the team has invested the most engineering in over the last two years. The capability surface matches Mux closely on the features production customers actually use, and diverges where the team has had a deliberate reason to choose differently.

Adaptive HLS is the default output. The ladder is configurable per project — by default the platform produces a sensible four-rung ladder for the median customer (240p, 480p, 720p, 1080p, H.264 video, AAC audio), and individual projects can extend it upward (4K, HEVC for the projects that opt in) or downward (180p for the low-bandwidth markets). Loudness normalisation runs by default so that audio levels across the catalogue stay within a couple of LUFS of each other — a small but operationally important detail for streaming products where audio inconsistency is a CSAT-killer.

Transcripts are first-class. Every uploaded video gets a Whisper-powered transcript by default, available as VTT, SRT, and a timed-segment JSON shape. The transcript completes asynchronously and triggers a `transcript.ready` webhook the moment it is available. The transcript backend is configurable — customers who need a specific ASR model (compliance reasons, a language Whisper covers poorly, a fine-tuned vertical model) can wire a custom backend behind the same API surface. The default is good for most teams; the escape hatch is there for the teams that need it.

Server-side clipping is one of the most underrated features in the pipeline. The clip endpoint takes a source video, a start time, an end time, and returns a brand-new asset that goes through the same pipeline as a fresh upload — its own HLS ladder, its own transcript, its own playback URL. For teams shipping highlight reels, social cuts, or multimodal-AI-driven extraction workflows (an emerging use case as multimodal models are used to identify "the interesting bit" of a longer recording), the clip endpoint turns a custom ffmpeg pipeline into a one-call workflow.

Playback security uses short-lived signed tokens. Public playback works the way you would expect for non-restricted content; signed playback issues a JWT-shaped token with configurable TTL, rotating secrets, and a revoke surface for compromised tokens. The hosted player handles the refresh dance transparently; teams using their own player get the token primitives and write the refresh themselves. The model is the one production paid-content workflows have converged on.

#The CDN: pull zones, storage zones, and granular purge

The delivery layer is what separates Ollanode from the pipeline-only vendors in the category. Mux ships a CDN, but the depth of the CDN surface (purge, signed URLs, edge rules, analytics) is shallower than what teams running production streaming at scale actually need. Cloudflare Stream rides Cloudflare's own CDN at depth but is opinionated about how content is served and limits configurability. Ollanode's CDN is built as a first-class layer, configurable, observable, and tightly integrated with the pipeline above it and the edge compute below it.

Pull zones are the primary CDN object. A pull zone maps a hostname — your `cdn.you.com` — to either an arbitrary upstream URL or a managed storage zone. The TTL is configurable per zone, the cache-bypass prefix list lets you exempt dynamic paths from caching without inventing a separate non-cached hostname, and CORS is handled at the edge so the origin never sees the preflight. The pull zone is the unit a team configures, monitors, and bills against.

Purge is where the team has invested most. Instant whole-zone purge runs in milliseconds across the platform's edge fleet. Path-level purge takes an explicit list and clears just those paths, also in milliseconds. The granularity matters because teams shipping personalised video or signed paid content cannot afford to purge the whole cache every time a single asset changes — the cache-miss storm against the origin would be catastrophic. The granular purge is the difference between a CDN you can operate at scale and one you cannot.

Hotlink protection uses HMAC-signed URLs with rotatable secrets. For paid-content workflows, the player receives a short-lived URL that the cache will serve and that other origins cannot replay. For non-paid workflows, the same primitive serves as a hotlink-prevention measure against scrapers. Signing is deterministic, verification is at the edge, and the origin only sees requests for content that has already been authorised.

Storage zones are S3-style buckets with presigned uploads, CDN-fronted by default. The canonical pattern is to put a storage zone behind a pull zone: the customer uploads via presigned PUT, the storage zone holds the bytes, the pull zone serves the cached version at the edge. Live edge analytics expose requests, egress and cache-hit ratio per zone in the dashboard, real-time, with drill-downs by status code and origin response time.

“ Granular path-level purge in milliseconds is the difference between a CDN you can operate at scale and one you cannot.

#Edge compute included, not bolted on

The thing the platform ships that the obvious alternatives do not is a first-class edge-compute layer integrated with the CDN. Edge compute in the video-and-CDN category is usually an add-on, usually with a separate billing surface, usually with cold-start latency that makes the obvious use cases — auth wrapping, A/B routing — feel painful. Ollanode ships edge compute as a first-class layer with V8 isolates as the execution model, which sidesteps the cold-start problem because there are no containers to start.

The handler shape is standard `Deno.serve(handler)`. A handler is a JavaScript function. Push the code and the platform deploys it within seconds, invokable at `/__fn/` from anywhere on the internet. The same handler can be wired into the pull-zone request lifecycle as a request-rewrite hook, sitting between the inbound client request and the cache lookup, which is where the most operationally-valuable use cases live.

Authentication is the canonical first use case. A handler can inspect the inbound request, verify a customer-issued JWT, look up entitlements in the customer's identity provider (the handler can call out to any HTTP endpoint), and either let the request through to the cache or return a 401 — all at the edge, with no origin round-trip. The latency overhead is in the low single-digit milliseconds because the isolate is already warm.

A/B routing is the second canonical use case. The handler reads a customer identifier from a cookie or a header, hashes it into a treatment bucket, and rewrites the cache key to serve different content per bucket. The cache hits are still cache hits — they just hit different cached objects depending on bucket. Geo-routing for regulatory content rules — serving a different cut of a video in certain jurisdictions, or blocking access in others — is the same pattern with a different routing function. Header rewriting handles the compatibility-shim use cases.

Per-project isolation is structural. Code from one project cannot read state, cache or storage belonging to another. The isolates are sandboxed at the platform level. Metering is per-project and per-invocation, exposed in the dashboard alongside the other usage analytics. The control plane handles the deploy, the versioning, the rollback, and the audit-logging of every code change.

#API-first developer experience

The platform is API-first, not dashboard-first. The dashboard exists, it is well-designed, and it covers the operational surfaces a developer needs to monitor the platform without leaving the browser. But every feature on the dashboard is reachable through a documented REST endpoint, scoped per project, with consistent request and response shapes across the 60+ endpoints in the surface. A workflow you can do through the dashboard is a workflow you can automate through the API.

The API conventions are conventional in a deliberate way. Bearer-token authentication. Project-scoped resources under `/v1/`. POST to create, GET to read, PATCH to update, DELETE to remove. Webhook payloads are JSON with a shared envelope shape, HMAC-signed for verification, and retried with exponential backoff on customer-side failure. The conventions are boring on purpose: every team that has written a REST integration before can integrate with the platform without learning a vendor-specific paradigm.

The published API reference covers all 60+ endpoints with example requests in curl, Python, JavaScript, and several other languages. The OpenAPI specification is available for download and is the source of truth for the SDK clients the team ships and for the OpenAPI-derived clients customers can generate themselves. The Postman collection is published. The platform behaves the way an API-first product from a developer-tools-savvy team should behave, because that is what the parent team builds for a living across the rest of the portfolio.

For teams that want infrastructure-as-code, the most-used resources — projects, zones, storage buckets, edge functions — are exposed through a published Terraform provider. The provider is the canonical path for teams whose deployment discipline includes "the platform is provisioned by the same Terraform run that provisions the rest of the cloud account."

#Pricing: predictable, published, no per-minute surprises

The pricing surface is one of the places the platform differentiates most cleanly from the established hosted alternatives in the category. The principle is straightforward: the cost model the customer signed up for at the prototyping stage should still describe the bill at the production stage. Mux's per-minute-encoding plus per-minute-streaming plus per-gigabyte-egress model has the structural property that the bill grows in three dimensions at once and surprises every team that scales successfully. The platform's model is structured around sensible monthly tiers with bundled inclusive volumes — encoding minutes, storage gigabytes, streaming bandwidth, edge-function invocations — that cover the median workload, with overage rates that are published, predictable, and on the lower end of the category.

The free tier is generous enough to evaluate the platform on real workloads. New signups get a free monthly allowance of encoding minutes, storage, and bandwidth that is sized to let a team upload several hours of video, transcode them through the full pipeline, generate transcripts, run them through the CDN, and exercise the edge-compute layer — all without entering a credit card. The principle is consistent with the rest of the Ollasoftware portfolio: any platform whose value is only legible after a contract negotiation is not as valuable as it claims.

The paid tiers move up in roughly the cost-per-unit structure customers expect: more bundled volume per dollar at higher tiers, lower overage rates at higher tiers, with the operational primitives — RBAC, audit logs, the SLA — added at the enterprise tier rather than at the early tiers where they are unnecessary overhead. The honest framing is that the platform is meaningfully cheaper than Mux at every comparable volume, broadly competitive with Cloudflare Stream on raw bandwidth and meaningfully deeper than Stream on feature surface, and priced above the rock-bottom-cost specialist vendors at a level that reflects the depth of the platform.

For enterprise tiers — the customers who need SCIM, SAML, dedicated regions, fine-grained RBAC, audit logs streaming to a SIEM, and a contractual SLA — the pricing shifts to a volume contract with predictable monthly minimums. The same data-handling guarantees that apply to the free tier apply to enterprise: the dashboard and the audit log are the canonical source of truth, the platform meters what it bills, and the customer can reconcile the invoice line by line.

#Who Ollanode is for

The customer profiles the platform fits well are concrete enough to enumerate. Streaming product teams — D2C video products, niche subscription streaming services, video-heavy community products — who want the depth of Mux without the bill of Mux. Learning platforms shipping course content at scale where the per-student streaming volume is high and the unit economics need to work. Social and UGC products that handle user-uploaded video and need the full pipeline running at production reliability.

Multimodal-AI product teams that are starting to use video as a first-class input — clip extraction, frame-level inference, transcript-driven retrieval. The clip endpoint and the configurable transcript backend fit these workflows naturally; the alternative is gluing together ffmpeg, an ASR vendor, and a CDN by hand, which is a multi-month engineering investment most teams cannot justify.

Internal-use teams shipping product demos, onboarding videos, support recordings, training videos, customer-uploaded support clips — workloads where the volume is real but the case for a premium-priced hosted vendor was never strong. The cost saving here is often not the headline reason; the operational depth is.

Indian and India-adjacent product teams shipping video to a domestic audience, where the cost discipline that defines the broader Ollasoftware posture matters more than the specific feature gap on a marquee American-frontier vendor. The platform is built and priced with this customer in mind; the broader global customer base benefits from the same cost discipline as a side effect.

The customer profile the platform fits less well is the very-large enterprise that has standardised on Mux through a multi-year contract and a deep set of integrations the team would have to rebuild. Migration in that case is a real cost. The platform is the right answer for teams that have not yet made the Mux commitment, or for teams whose Mux contract is up for renewal and looks like it will scale uncomfortably for the next year of growth. For the rest, the conversation is more nuanced and starts with the team's actual workload rather than with the marketing pitch.

#How Ollanode compares to the established vendors

The hosted video category has a small number of credible vendors and it is worth being direct about where the platform sits.

Mux is the polished, well-engineered, well-supported, expensive option. The feature set the platform matches is the working subset of Mux that production customers actually use — the HLS ladder, the transcripts, the storyboards, the signed playback, the hosted player, the API-first developer experience. Where Mux extends past the platform is the breadth of integrations with adjacent SaaS, the depth of the analytics surface, and the maturity of the SLA and support tier for very large customers. Where the platform extends past Mux is the integrated edge-compute layer, the more honest pricing structure, and a feature-for-feature offering that is meaningfully cheaper at every comparable volume. The honest framing is that Mux is the right choice when operational simplicity dominates and budget is not a constraint; Ollanode is the right choice everywhere else.

Cloudflare Stream is the high-scale alternative. Stream is cheap at the upper volume curve, runs on Cloudflare's own network, and is operationally simple. It is also tightly bound to the rest of the Cloudflare ecosystem, opinionated about how content is served, and offers less feature depth in the pipeline than either Mux or the platform — transcripts arrived recently, server-side clipping is limited, the player surface is narrower, and the configurability of the CDN layer is constrained by Cloudflare's own product priorities. For teams already deep in Cloudflare, Stream is reasonable; for teams who want more pipeline depth, more CDN configurability, or both, the platform is the alternative that compares favourably.

Bunny Stream is the smaller-vendor comparison. Bunny has very competitive unit economics at the volume tier, a reasonable feature surface, and an honest approach to pricing. The platform's extension over Bunny is in the depth of the pipeline (Whisper transcripts as a default rather than an add-on, server-side clipping, the included edge-compute layer) and in the integrated operational surface (RBAC, audit logs, edge analytics). For teams who chose Bunny mostly for cost, the platform delivers a competitive cost posture with a meaningfully deeper feature set.

Cloudinary, AWS MediaConvert + CloudFront, Wowza, Bitmovin, and the various component-level vendors are different shapes of bet. They are appropriate when the team genuinely wants to assemble the pipeline from parts. The platform's value over those is the integration work the team has done so the customer does not have to: the same engineering hours that would be spent wiring MediaConvert pipelines and CloudFront configs are spent shipping product instead.

Across all of these, the question is not "which platform is cheapest per minute or per gigabyte." It is "for the video workload my team is actually building, what is the total cost of ownership — including the engineering cost of integration, the operational cost of running on a fragmented stack, and the unit cost of the bill itself — compared to a single hosted platform that handles the whole surface." For most product teams in the series-A-to-mid-market range, the answer points clearly at the platform.

#The team and the operational heritage

Ollanode is built and operated by Ollasoftware, the AI software development company headquartered in Bengaluru that has shipped more than forty AI brands in production over the last four years. The infrastructure heritage of the team is dense for a video product. The engineers behind the CDN layer came from the team that built the high-volume crawler infrastructure powering Crawlcrawl — billions of pages a month, hundreds of millions of cache hits a day — and bring the bandwidth-engineering muscle that lives there. The control-plane discipline — webhooks, audit logs, scoped API keys, team-and-roles — is shared with the team that ships Aeoniti's enterprise-tier surface. The edge-compute layer is built by the engineers who run the V8-based gateway underneath Switchllm.

The video-pipeline expertise came from a different and unusual source. The parent group, Networkers Home, has been running a course-content production and delivery operation at meaningful scale since 2007 — currently delivering training video to a population of forty-five-thousand-plus alumni and an active student body across cybersecurity and networking certifications. The pipeline-engineering instincts behind the video layer are anchored on real operational experience at running a streaming service whose costs the founders care about, on a content catalogue whose value the founders understand, for customers whose experience the founders have to honour. The design choices in the platform reflect what the team has had to ship for its own learning operation, not what looks good in a vendor comparison.

That operational heritage is the answer to the implicit question any team evaluating a new hosted platform asks: who is going to be running this in five years. The platform is built by a team that already runs production AI and content infrastructure at scale, inside a company that has been operational for the better part of two decades, with a customer base inside the parent group that gives the team a permanent stake in the platform behaving well. The platform is not a venture-funded sprint to acquisition; it is a product the parent team intends to ship indefinitely.

#What is on the roadmap

The team publishes a roadmap on the docs site and ships a changelog against it on a regular cadence. The visible near-term threads are: expanded regional pool coverage so latency is competitive globally rather than only on the routes the platform already serves well, deeper integration with the analytics-and-business-intelligence stack streaming product teams already run (Segment, Mixpanel, Amplitude, Snowflake) so view-level data flows automatically rather than through customer-built ETL, a richer hosted player for teams that use it as their default (chapter UI, transcript-driven search, configurable controls, brandable theming), and an expanded webhook event catalogue for the operational events teams currently have to poll for.

Underneath those visible features is steady investment in the codec frontier. The current default is H.264 + AAC for compatibility reasons; the platform supports HEVC and AV1 for projects that opt in, and the trajectory is toward HEVC and AV1 being viable defaults as device support compounds. The team has been explicit that the transition will be opt-in for some time — the operational rule "the default codec is the one that plays everywhere" is the right rule for most teams and the platform will not break it without explicit customer choice.

On the edge-compute layer, the roadmap thread is around expanded standard-library APIs in the isolate runtime — particularly the Web Crypto APIs, the WebSocket APIs, and the streaming-response APIs that production edge workloads increasingly need. The current surface is sufficient for the canonical use cases (auth wrapping, A/B routing, request rewriting); the expansion targets the workloads emerging at the intersection of video and AI inference at the edge.

Pricing during the current phase is the published tier model with the generous free signup allowance. The team has signalled that the unit economics will not get more expensive over time — they will get cheaper as the underlying infrastructure cost compounds down and the savings are passed through. The principle is consistent with the rest of the portfolio: a platform whose pricing only goes up loses customers as fast as it adds them.

#How to start

If you ship video at any meaningful volume and your current invoice from Mux, Cloudflare Stream or one of the equivalent vendors has become a line item you would rather not be paying, the right next move is to evaluate the platform on a real workload rather than on a synthetic benchmark. Sign up at ollanode.com, create a project, upload a representative video, run a representative streaming session against the resulting playback URL, and look at how the platform handles your specific shape of workload.

The signup flow is sixty seconds. No credit card. The free tier covers enough encoding minutes, storage and bandwidth to run the kind of evaluation that gives you real information — not just "the upload worked," but "the transcription quality is acceptable, the streaming latency is what we need, the edge function does what we need it to do, the dashboard tells us what we need to know."

Run the platform on real customer traffic for a few days before drawing any conclusion. The numbers that matter — first-byte latency at the edge, cache hit ratio, transcript quality, overall unit cost — only become legible after the platform has had enough traffic to calibrate. The trajectory matters more than the snapshot, and the trajectory on the platform tends to compound favourably as the regional pools warm up to your traffic pattern.

If you would like the team to walk you through a specific workload — high-volume streaming evaluation, edge-compute use case design, an enterprise procurement conversation, or a migration plan from an existing hosted vendor — the Ollasoftware contact page reaches the engineers who ship the platform. Response times are unusually fast for the B2B-video category, which the team treats as one of the small operational signals that the product takes itself seriously.

And if you are not yet sure the platform is the right answer — if you are still triangulating between vendors, or still deciding whether a single-platform consolidation is the right priority for your team this quarter — the published documentation, the live dashboard demo, the OpenAPI specification, and the changelog are all open. The platform's case is most legible when measured against your own workload; the docs and the free signup are the team's way of inviting that measurement directly.