Synthesia dubbing at 80 credits a minute vs a Sume lip-sync chain

Synthesia's October 1 plans page lists dubbing with lip-sync at 80 credits per minute, so a 10-minute video is 800 credits. On Sume there is no single dub switch: you chain steps, and the part that moves lips is the H3 Max lip-sync route, which costs $0.10 per second at 768p for segments of 5 to 14.8 seconds. Ten minutes of re-lipped speech would therefore bill $60 at 768p, before the cheap steps around it.

The numbers side by side

Credits are Synthesia's unit and I do not convert them to dollars here, because the page I read gives credit costs, not a dollar rate for every plan. The Sume rows are dollars from the repo rate card.

The chain on Sume

The 41 segments come from 600 seconds divided by 14.8, rounded up. In practice, cut at sentence ends, so expect more.

• Detach the source audio with /v1/audio-detach. Output is capped at 900 seconds, so a 10-minute file fits.
• Transcribe and translate the speech with your own tools; Sume's TTS then voices the translated lines.
• Synthesize each segment of audio first, then lip-sync it. Never re-cut audio after the lip-sync job, or the mouth will drift.
• Submit each segment to /v1/minimax/h3-max/lip-sync with the same still or avatar and a duration_seconds equal to the audio length.
• Join the clips and the voice track with /v1/timeline-1.0/render.

Where the money goes

Nearly all of the Sume cost is the lip-sync step: $60 of roughly $61 at 768p. The detach job is a cent, and the timeline render for ten minutes is a dollar. That makes resolution the lever to pull. The same 600 billed seconds cost $37.50 at 480p and $120 at 1080p, from the rates of $0.0625 and $0.20 per second. A sensible plan is to run the first two segments at 480p, check that the face and timing hold, and only then run the full set at the resolution you ship. The TTS line is not in the total because it depends on your character count and the voice you choose.

What Sume does not do

Sume does not take an existing video and re-lip its original speaker. The lip-sync route starts from a still image and audio, so it makes a new talking-head clip rather than editing the mouth in your footage. Pieces that fit one person speaking to camera work well; footage with cuts, several speakers or moving cameras does not map onto it.

So the comparison is partly about scope. Synthesia's line item covers a dub inside its own editor. The Sume chain is cheaper in the fixed steps and more expensive per second of lip-sync, and it asks you to do the segmenting.