28-second voice-over: two H3 Max halves $2.81 or one Fabric job $5.25

For a 28-second voice-over on one still, two MiniMax H3 Max Lip Sync takes of 14 seconds each cost 28 x $0.10 = $2.80 at 768p, plus $0.01 for the audio split, so $2.81. One VEED Fabric job costs 28 x $0.1875 = $5.25 at 720p, or 28 x $0.10 = $2.80 at 480p.

The numbers side by side

H3 Max Lip Sync accepts audio of 5 to 14.8 seconds, so 28 seconds does not fit in one call. Fabric accepts audio up to 300 seconds. Both need Sume-hosted audio under 10 MB.

Splitting the audio

Sume's Timeline audio route splits a Sume-hosted file into ranges in one job at a flat $0.01, and returns durable media.sume.com URLs. Ranges can overlap, so a cut at a pause is easy. Pick the cut at a natural pause, since the two clips will be joined and a mid-word cut will show. Two 14-second slices are inside the 5 to 14.8 second window with room to spare.

Which one

If the shot is one continuous face and you care about the join, Fabric in a single job avoids a seam. If the cost matters and a cut is acceptable, H3 Max at 480p costs 28 x $0.0625 = $1.75 plus $0.01 for the split, the cheapest figure in the table. Join the two finished clips with the timeline route, which charges $0.10 per output minute, minimum one minute.

Checks before you run either plan

Measure the 28-second file and confirm it is under 10 MB; a 28-second stereo 44.1 kHz 16-bit WAV is about 4.94 MB (28 x 176,400 bytes), so it fits. Upload it to Sume so it has a Sume-hosted URL. For the split plan, create two ranges such as 0 to 14 and 14 to 28 seconds, and confirm each slice is between 5 and 14.8 seconds.

Then send the same still to both H3 Max calls with the same resolution so the faces match. Poll each job, then join the clips on the timeline. If the join point falls on a pause, a cut is hard to see; if it falls mid-word, re-split at the nearest pause.

If any job fails, the reservation is refunded before capture, so the cost of a failed half is only your time.