Generate media

Every generation command follows the same pattern: create queues the job and polls until done, list shows past generations (--page <n>), status <id> checks one job, and delete <id> removes it. image, music, thumbnail, and clip also have get <id>. Add --save to get or status to download the result.

Images

unsora image create --prompt "isometric 3d icon of a rocket" \
  --model nano-banana-2 --ratio 1:1

# With reference images
unsora image create -p "same character, waving" --ref https://.../char.png

nano-banana-2 is the default. Run unsora models --category image for every image model, then pass its key with --model. --resolution (1k, 2k, 4k) and --quality work on models that offer them.

Video

unsora video create --prompt "cinematic drone shot over mountains at sunset"

# Pick another model, with sound, at 1080p
unsora video create -p "a chef plating pasta, close-up" \
  --model veo-3.1-fast --sound --resolution 1080p --ratio 9:16

# Image-to-video with start and end frames
unsora video create -p "slow push-in" \
  --start-image https://.../first.png --last-image https://.../last.png

seedance-2.0 is the default. Run unsora models --category video for every video model (Seedance, Veo, Kling, Sora, Wan, and more) and the options each one accepts.

Flag	Effect
`-m, --model <model>`	Video model (default `seedance-2.0`)
`-d, --duration <seconds>`	Length in seconds (default 5)
`-r, --ratio <ratio>`	Aspect ratio (default 16:9; valid values depend on the model)
`--resolution <res>`	Output resolution, such as 720p or 1080p
`--sound` / `--no-sound`	Generate with or without audio, on models that offer it
`--negative-prompt <text>`	What to avoid (not supported by Sora or Seedance)
`--start-image <url>` / `--last-image <url>`	Start and end frames
`--ref <url>`	Reference image (repeatable)
`--ref-video <url>` / `--ref-audio <url>`	Seedance omni reference video or audio (repeatable, up to 3 each)

unsora video list -m <model> filters past videos by model.

Any model with the catalog

unsora catalog works with every video, image, and motion-control model, using the input names from the live catalog:

unsora catalog list --category video                 # every model in a category
unsora models veo-3.1-fast                           # modes and inputs for one model
unsora catalog quote -m veo-3.1-fast -p "a red fox in snow"   # price it, free
unsora catalog generate -m veo-3.1-fast -p "a red fox in snow" \
  --expected-credits 120

Pass any other input with -i <key>=<value>, using the keys unsora models <model> lists. With --expected-credits, nothing is charged if the live price is higher.

Flag	Effect
`--mode <mode>`	Mode key (default: picked from the attached media)
`-i, --input <key=value>`	One model input by its catalog key (repeatable; repeat a key for a list)
`--inputs <json>`	All inputs as one JSON object
`--category <category>`	Reject a model from another category

Music

# Song (lyrics required for song models)
unsora music create --prompt "r&b, slow, passionate, male vocal" \
  --lyrics "[Verse] ..." --model mureka-9

# Instrumental BGM
unsora music create --prompt "calm lo-fi study beat" --model mureka-7.5

The default model is auto. --format sets the output to mp3 (default), wav, or flac.

Voiceovers

# Browse the 20 Eleven v3 voices (with preview URLs)
unsora voiceover voices

# Generate speech (6 credits per started 1,000 characters)
unsora voiceover create --text "Welcome to Unsora." --voice Brian

Use <#1.5#> between words in the text to pause for 1.5 seconds. --stability (0–1) makes delivery more consistent, and --similarity (0–1) sets how closely it matches the base voice.

Thumbnails & influencer portraits

# YouTube-style 16:9 thumbnails, multiple variations
unsora thumbnail create --prompt "shocked face, red arrow, MrBeast style" \
  --variations 3

# AI influencer portraits
unsora influencer create --prompt "young woman, coffee shop" \
  --style golden_hour --angle portrait --count 2

Thumbnail options:

Flag	Effect
`--ref <url>`	Reference image, such as the creator’s face (repeatable, up to 16)
`--template <url>`	Layout template image (1)
`--context <url>`	A YouTube video (title and transcript) or a public PDF to read
`--context-type <type>`	`youtube` or `document` (default: detected from the URL)
`--expression <name>`	`auto`, `neutral`, `soft-smile`, `big-smile`, `surprised`, `confused`, `worried`, `angry`, `sad`, `disgusted`, or `determined`
`--variations <n>`	Number of variations (1–20)

Influencer options: -r, --ratio (1:1, 16:9, 9:16, 4:3, 3:4), -c, --count (1–10), --angle, --style, and --age (18–70). Thumbnail and influencer jobs can return several files, so they don’t take -o; results download to the current directory.

Clipping

# Auto-detect the most viral clips
unsora clip create --url "https://www.youtube.com/watch?v=..."

# Five vertical 30–60s clips with captions, and wait for the result
unsora clip create -u "https://www.youtube.com/watch?v=..." \
  --limit 5 --ratio 9:16 --duration 30-60 \
  --captions --caption-style bold-white --wait

# Find specific moments
unsora clip create -u "https://..." --query "every time we talk about pricing"

# Translate the captions
unsora clip create -u "https://..." --captions --target-lang es

# Check progress / fetch results later
unsora clip list
unsora clip get <clipping-id>
unsora clip status <clipping-id>
unsora clip delete <clipping-id>
unsora clip delete-clip <clipping-id> <clip-id>

--source-lang sets the spoken language if detection needs a hint, and --target-lang translates the captions. Clipping can take several minutes, so clip create returns right away unless you add --wait. It accepts YouTube links or direct video file URLs. See the clipping API reference for every option.

API documentation

Guides

CLI

Integrations

User

Accounts

Posts

Analytics

Models and pricing

Generation status

Video generator

Image generator

Music generator

Voiceover

Voice clones

Voice changer

AI influencer

Thumbnail generator

Clipping

Image upscaler

Video upscaler

Subtitle removal

Motion control

Talking avatar

Movie materials

Uploads

Generate media

Images

Video

Any model with the catalog

Music

Voiceovers

Thumbnails & influencer portraits

Clipping