01
Match the job to the input
| Job | Better start | Why |
|---|---|---|
| Keep a real product or face | Photo | Likeness lives in the still |
| Invent a scene from words | Text prompt | No usable still yet |
| Need social structure fast | Template | Timing and captions pre-built |
Photo paths and text paths are not interchangeable. A soft crowded still often burns a credit on a distorted face; a vague prompt burns another on a clip that never needed that photo.
Secondary phrases such as image to video and image to video ai generator support the same decision without stuffing. “how to image to video” and “photo to video converter” usually map to photo-first or template-first.
02
Decision rules that save credits
- Recognizable still → photo first, short motion line.
- Only a sentence → one text-to-video with one subject.
- Need hook-body-end today → template, swap media, Studio polish.
03
When the first result misses
Do not auto-regenerate. Ask: wrong input, wrong motion, or only needs Studio trim/captions? Path changes beat blind retries.
04
How we tested this in Soranova
We ran three short jobs from one brief (product still, text-only, template) and compared credit use and likeness. Photo kept packaging readable; text invented a backdrop; template shipped fastest with captions. Results still vary by model and device. Credits apply to generations and some exports. Read the meter before you retry.
When the path is clear, open Soranova on Android, run one focused job, and export only after the gallery tile appears.
05
Related guides
