The studio video stack

All three tools here have a genuine free tier, so a working configuration costs nothing while you evaluate it. The 03 COSTS table above breaks down what each vendor publishes. Three meters climb with use: Synthesia's credit-based system, which converts to video minutes per plan; ElevenLabs' shared credit pool, which dubbing drains faster than narration; and Runway's per-second credit burn on every take. The sticker is not the bill; budget the procurement cycles alongside the spend, and plan volume, since minutes do not roll over.

Rarely. Synthesia is the core: governed presenter video for training and communications is the job this stack exists for, and many teams start there alone. The numbered steps double as the adoption order. Add ElevenLabs when audio consistency across languages and formats becomes a brand question rather than a production detail. Add Runway last, and only when generative cinematic work is genuinely on the brief.

Then you already own the presenter. Your Synthesia subscription covers avatar video for training and comms, and you can defer the rest until the work demands it. This stack graduates that base: Synthesia onto the enterprise governance L&D scale needs, SCORM, SSO and unlimited minutes; ElevenLabs for one consistent, licensed voice across narration, dubbing and product audio; and Runway for the cinematic footage cannot cover. Keep Synthesia as the spine; add voice and generation around it.

Because the three do different jobs, and a general generator does only one of them. A model like Runway makes cinematic clips from a prompt, but it cannot run governed, translated training video at organisational scale: that is Synthesia's presenter library, consent controls and SCORM. Nor does it hold one licensed voice across every course and walkthrough: that is ElevenLabs. Generation is the occasional layer, not the spine; a studio needs the presenter and the voice under governance first.

Often, and this tier says so plainly. The studio stack repays its enterprise contracts and studio process only at real organisational scale: many languages, training under compliance, a brand voice held identical across hundreds of assets. A team producing polished video occasionally, without the L&D governance load or the translation reach, is paying for process it will not run. That team wants the competitive tier, the professional video stack.

Synthesia states it does not use customer inputs or outputs to train its AI models on any plan, with fine-tuning only by written customer instruction; treat ElevenLabs' consumer tiers as semi-public until its enterprise/DPA tier, and put confidential internal policy, scripts and audio on the governed contracts, which also add tenancy and the admin controls enterprise L&D needs. Two consent obligations hold throughout: no one becomes a Synthesia avatar, and no voice is cloned in ElevenLabs, without documented permission. Human-review every generated script for accuracy before release.