42-second talking still: image, voice, lip sync, $4.33 to $8.01

A 42-second talking still costs $4.3304 on Sume with MiniMax H3 Max Lip Sync at 768p, or $8.0054 with VEED Fabric 1.0 at 720p. Both include a $0.10 Nano Banana 2.1 still at 1K and the voiceover: 640 characters of text to speech at $0.0475 per 1,000, which is $0.0304. The lip-sync line is at least 96 percent of the total, so choose that model first.

Per-audio-second pricing

Both lip-sync models bill per second of audio, not per second of picture. MiniMax H3 Max Lip Sync is $0.10 per audio second at 768p and $0.0625 at 480p; VEED Fabric 1.0 is $0.1875 at 720p and $0.10 at 480p. I assume the script is 640 characters and that you measured it at 42 seconds of speech.

Reading the table

At matching 480p, H3 Max ($2.625) is cheaper than Fabric ($4.20) by $1.575. At the sizes each model is built for, the gap is $3.675 (7.875 - 4.20). The still and voice together are $0.1304 in every row, so they never change the ranking.

Shorter audio is the biggest lever. A 30-second version on H3 Max at 768p costs 30 x 0.10 = $3.00 for lip sync, or $3.1304 in total. Each second you cut from the script saves $0.10 on H3 Max or $0.1875 on Fabric.

Script length scenarios

The script length drives almost all of the cost. On H3 Max at 768p the lip-sync line is $1.50 for 15 seconds, $3.00 for 30, $4.20 for 42 and $6.00 for 60. On Fabric at 720p the same lengths cost $2.8125, $5.625, $7.875 and $11.25.

For a series of 10 talking stills of 30 seconds each, the lip-sync line is 10 x $3.00 = $30.00 on H3 Max or 10 x $5.625 = $56.25 on Fabric, before the $0.1304-ish per-clip still and voice (about $0.13 each, so $1.30 more for the set).

If the picture changes only a little between episodes, reuse the still; each reuse removes $0.10 from the cost of a clip and keeps the character consistent.

What the price does not tell you

I did not compare quality, and the catalog does not rank these models. Run a 5-second sample on each (5 x 0.10 = $0.50 on H3 Max, 5 x 0.1875 = $0.9375 on Fabric) before you commit a 42-second render. The still itself is a separate generation, so a retake of the picture adds $0.10, not a new lip-sync fee.

Before you commit

Three checks cost almost nothing compared with a 42-second render.

• Cut the script to the shortest version that still says the thing. Each second is $0.10 on H3 Max.
• Confirm the audio length from the TTS result before submitting lip sync, because lip sync bills audio seconds, not characters.
• Run the 5-second sample on both models with the same still and the same audio, then look at the mouth shapes and the jaw.