AI video photo requirements: one person per photo
Use a front-facing, well-lit portrait with visible eyes and some room around the head and shoulders. Avoid group photos, strong beauty filters, masks and deep shadows. Supported files are JPEG, PNG and WebP, up to 2 MB each. Only use photos you have permission to use.
Choose the performance before the settings
Compare the available performances in the theme library. Check the number of people, photo requirements and duration shown for your selected theme before preparing the inputs.
Set the format for where you will share
Choose the format for where you will share: a vertical phone clip, a landscape frame or a square post. The settings show supported aspect ratios and durations; some themes use a fixed length.
Review the model and credit quote
Seedance 2.0 Mini and Fast offer 480p and 720p. Seedance 2.0 and 2.5 also offer 1080p. Start with the settings you actually need: the exact credit quote updates with model, resolution and duration. Native audio is included. Review the final quote before you confirm generation.
If the result misses your idea
Change the input that controls the problem. Adding more instructions to the audio topic will not fix a blurry face or change the selected aspect ratio.
- The face does not look enough like the person
- Try a sharper portrait with even light, visible eyes and fewer filters. Keep one person in each image and avoid a face partly hidden by a hand or accessory. Better input helps, but likeness is not guaranteed.
- The words or language are unexpected
- Put the main subject first, state the language and remove competing requests. The topic directs original audio rather than exact word-for-word lyrics. Listen to the full result before sharing.
- The frame cuts off something important
- Choose the ratio in Video settings before generating. Use a source portrait with room around the head and shoulders. Writing 'vertical' in the audio field does not change the video ratio.
- The job is slow, failed or uncertain
- Check its saved status in My Videos. A slow queue is different from a confirmed failure. Avoid submitting duplicates while the outcome is uncertain. Confirmed failure or rejected output receives a full credit refund.
Before you confirm generation
Check the people and theme, read the audio topic, confirm the ratio and resolution and review the displayed credits. Uploading photos alone does not start a video job. Once generation is confirmed, follow the saved task before submitting another request.