D-ID audio talks: 15 MB, 5-10 minutes vs Sume's 4-60 s script

D-ID's Create a talk page lists text scripts up to 40,000 characters (10,000 excluding SSML tags) and audio files up to 15 MB, with audio length limited to 5 minutes for clips and 10 minutes for talks. Sume's Avatar 1.0 takes a text script or multi-scene plan whose estimated duration is 4 to 60 seconds, in English, from an avatar you created once. If you start from a recording, Sume's separate still-plus-audio routes apply instead.

D-ID's figures are from Create a talk and Create a Video, read 2026-10-10. Sume's are from Generate avatar video and Models.

Inputs side by side

The two products are shaped differently, so the limits do not line up.

Where your audio goes on Sume

Sume does not take an audio URL on POST /v1/avatar-1.0/talking-video. For a recorded voice and a face, the models page lists VEED Fabric 1.0 at POST /v1/veed/fabric-1.0, which takes a still plus audio_url and a measured duration_seconds, and MiniMax H3 Max Lip Sync at POST /v1/minimax/h3-max/lip-sync, which takes the same body with audio of 5 to 14.8 seconds. The repo's pricing fixture describes Fabric as up to 300 seconds of audio, Sume-hosted and under 10 MB, at $0.1875 per audio second in 720p.

So a 5-minute recording that D-ID can take as clip audio fits Fabric's 300-second ceiling exactly: 300 seconds at $0.1875 is $56.25. That price is for the Fabric route, not the Avatar 1.0 talking video, whose tiers are $0.184, $0.245 and $0.55 per second (standard, plus, max) before the product-image surcharge.

Choosing by what you have

• You have a script and want a presenter who is reusable across videos: Sume Avatar 1.0, within 4 to 60 seconds. A 60-second standard clip costs 60 x $0.184 = $11.04.
• You have a studio recording of 5 to 10 minutes: D-ID accepts audio of that length; on Sume, split it, or use Fabric up to 300 seconds.
• You have text longer than a minute: D-ID accepts far more characters, while Sume asks you to split the script into several jobs or scenes, each within the window.
• You need languages beyond English: D-ID lists several TTS providers with language-specific voices; Sume Avatar 1.0 will not do it today.

Planning a long script on Sume

Sume estimates duration from the script, so word count is the lever. If an estimate lands above 60 seconds, the request is rejected, so cut it or split it before you submit, and keep the same avatar handle across the pieces so the presenter stays consistent. Each piece is its own job with its own idempotency key. When a script is long, a quick dry run with the avatar-video preview stage shows first frames before any final render is paid for.

Cost check before you split

Splitting a long script is cheap on Sume when you do it deliberately. Two 55-second standard clips are 110 seconds at $0.184, which is $20.24, while the same words as one request would be rejected for exceeding 60 seconds. Cut at a sentence break, keep the same avatar_handle, and give each piece its own Idempotency-Key, so a retry never bills a piece twice.

The D-ID figures here are limits on what a request may contain, not prices, and I did not read D-ID's price list for this post. Compare cost only after you have fetched the current plan page yourself.