Hedra Character 3 vs Sume Avatar 1.0: per-second cost, 30 and 60 s

On list price alone Hedra Character 3 is cheaper per second: its model page lists 2.5 cents at 540p, 5 cents at 720p and 6.25 cents at 1080p, against Sume Avatar 1.0 at $0.184, $0.245 and $0.55 per second at 720p. The two are not the same product, though: Hedra's model is audio-driven and needs a portrait plus an audio file, while Sume writes and voices the script for you from a reusable avatar.

Hedra's numbers are from Hedra Character 3, read 2026-10-10. Sume's are from Generate avatar video and Create new avatar.

Totals for 30 and 60 seconds

Multiplying per-second rates by length, with no product image on the Sume side. Hedra bills by generated video length, per its page, and an audio file has to exist before you start.

What the Hedra number leaves out

Character 3 takes audio. Hedra's page lists audio length up to 10 minutes and an optional start-frame image, and its API reference also describes server-side audio generation, but I did not read a price for that step, so the per-second figure should be treated as the video render only.

Sume's tier price is for the whole result: the script is spoken in an English voice, the face is animated from your avatar, and one job returns an MP4. The avatar itself costs $0.95 once, so a campaign of ten 30-second standard clips is $0.95 + 10 x $5.52 = $56.15, before any product image surcharge.

Choosing

• You already have recorded or generated audio and want a face on it cheaply: Hedra's audio-driven model is priced for that.
• You have only a script and want one repeatable presenter, in English: Sume Avatar 1.0, from 4 to 60 seconds.
• You need 1080p: Hedra lists it; Sume Avatar 1.0 is 720p.
• You need a Sume route for existing audio: Fabric and H3 Max lip-sync are the audio-driven options on Sume, with their own prices on the models page.

Check before you commit

Price lists change. Run one clip on each route, compare the face, voice and timing against the brief, and only then multiply by your volume. A cheaper second that needs a retake costs more than a pricier one that lands first time.

A worked ten-clip campaign

Ten 30-second clips on Sume at standard cost 10 x $5.52 = $55.20, plus $0.95 for the avatar, so $56.15 in total. At plus it is 10 x $7.35 = $73.50 plus $0.95, or $74.45. At max it is 10 x $16.50 = $165.00 plus $0.95, or $165.95.

On Hedra at 720p the same 300 seconds of video is 300 x $0.05 = $15.00, and at 1080p it is 300 x $0.0625 = $18.75. Those totals exclude whatever it costs to produce the ten audio files, so the real gap depends on where your audio comes from.

If your audio already exists, the fair comparison on Sume is the Fabric lip-sync route rather than Avatar 1.0, because both take audio and a still. The models page lists its price per audio second; price that route for a like-for-like check.

Reading the 720p row fairly

Both products list 720p, so that row is the closest like-for-like on resolution: $0.05 per second at Hedra against $0.184 at Sume standard, which is 3.68 times as much. The extra money buys a script-to-video path with a stored avatar, an English voice and optional product-image placement, none of which an audio-driven model does for you.

If you are comparing for a high-volume channel, the multiplier matters more than the sticker: at 1,000 thirty-second clips the 720p difference is $1,500 against $5,520 on video render alone. Whether that gap is worth it depends entirely on how much work your pipeline spends making the audio and the portrait.