D-ID

D-ID built its name on one trick done at scale: a single still photograph becomes a talking, expressive face, and text input produces presenter-style video with synchronized speech in many languages. The current generation is branded V4 Expressive Avatars, sold both as a studio for non-editors and as an API for products that want a face on the other end. It is genuinely good at volume talking heads — training modules, localized explainers, agent interfaces — where the alternative is a camera, a presenter and a reshoot whenever the script changes.

Metering is the part to read before committing. Usage bills as minutes: each render deducts its duration from the plan balance, durations round up into 15-second intervals, and unused minutes go void monthly rather than rolling over. API calls draw on the same pool as the web studio, so an integration competes with your own production work. Trial output carries a full-screen watermark, which the vendor frames as transparency about synthetic media rather than as a paywall. Plan prices and resolution ceilings were not on the page read here — they sit on separate studio and API sheets — and neither were the consent terms for cloning a real person's face or voice, so get those in writing before animating anyone. Where it loses: the talking head is the entire repertoire. Multi-shot scenes, camera work and photoreal bodies belong to Veo, Runway, Kling or Pixverse, and a weak source photo still yields the category's stock uncanny result.