Video Generation

Three video models are served today: veo-3.1-fast, kling-3-pro and kling-avatar-v2-pro. The other video ids that used to be listed (kling-2.5-pro, kling-3-pro-audio, seedance-2-pro, seedance-2-fast, veo-3-fast, veo-3, the hailuo-02-* family) are still paused and return 404 model_not_found. Check GET /v1/models for what is live. Billing is per second of generated footage.

curl -X POST https://kymaapi.com/v1/videos/generations \
  -H "Authorization: Bearer $KYMA_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "veo-3.1-fast",
    "prompt": "a street vendor in Hanoi greets the camera and says: chào buổi sáng, ambient traffic",
    "duration_s": 6
  }'

The endpoint is asynchronous. POST returns 202 with a job_id immediately; poll GET /v1/jobs/{job_id} until status is succeeded.

Pick a model

Model	Best for	Cost / sec	Durations	Audio	Input
`veo-3.1-fast`	Clips with speech and sound, talking heads	$0.135	4, 6 or 8s	native	text + image
`kling-3-pro`	Premium cinematic, hero brand video	$0.1512	3 to 15s	—	text + image
`kling-avatar-v2-pro`	Talking or singing clip from a photo and your audio	$0.15525	as long as your audio, 5 to 60s	your own track	image + PCM WAV audio

veo-3.1-fast and kling-3-pro accept image_url for image-to-video; kling-avatar-v2-pro requires it. Live canonical source: GET /v1/pricing.

veo-3.1-fast

Google Veo 3.1 Fast. 720p video with native audio: dialogue, ambient sound and lip-sync, including spoken Vietnamese. Use it when the clip needs to talk.

{
  "model": "veo-3.1-fast",
  "prompt": "a barista hands over a cup and says 'your flat white', cafe murmur in the background",
  "duration_s": 8,
  "aspect_ratio": "9:16"
}

Cost: $0.135 / sec, times the clip length
Durations: 4, 6 or 8 seconds (default 4; other values snap down to the nearest of these)
Resolution: 720p, aspect_ratio 16:9 (default) or 9:16
Audio: native, always on
Modes: text-to-video, image-to-video (image_url becomes the first frame)
Best for: talking heads, social clips with dialogue, scenes with diegetic sound

kling-3-pro

Flagship Kling. Photoreal humans, smooth motion. Use it for hero shots and premium brand video where the quality needs to stand up at full screen. Clips are silent.

{
  "model": "kling-3-pro",
  "prompt": "an architectural fly-through of a glass-and-steel tower at sunset, cinematic anamorphic look",
  "duration_s": 5,
  "aspect_ratio": "16:9"
}

Cost: $0.1512 / sec ($ 0.7560 for 5s)
Length: any whole number of seconds from 3 to 15; default 3
Aspect ratio: 16:9, 9:16 or 1:1 for text-to-video; image-to-video follows the image
Modes: text-to-video, image-to-video (pass image_url)
extra: only cfg_scale (a number from 0 to 1, how closely the clip follows the prompt) is forwarded. Every other extra field is ignored, so a job always runs the shape it was priced and checked for.
Best for: hero brand video, character and face shots, premium cinematic

kling-avatar-v2-pro

Photo + audio → lip-synced clip. Send one portrait and one speech or song track. The face lip-syncs to the audio, the audio plays through unchanged, and the clip runs exactly as long as the audio.

curl -X POST https://kymaapi.com/v1/videos/generations \
  -H "Authorization: Bearer $KYMA_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "kling-avatar-v2-pro",
    "prompt": "a friendly presenter, natural head movement, looking at the camera",
    "image_url": "https://example.com/portrait.jpg",
    "audio_url": "https://example.com/voice.wav",
    "duration_s": 13
  }'

Field	Required	What it is
`image_url`	yes	HTTPS URL of the face to animate
`audio_url`	yes	HTTPS URL of the speech or song as an uncompressed PCM WAV file (MP3, M4A, AAC and OGG are not accepted yet), at least 5 seconds
`duration_s`	yes	Your audio’s length in seconds, rounded up (5 to 60). It caps the charge; audio longer than it is cut at `duration_s`
`prompt`	yes	Style and motion hints; the audio drives the mouth

Cost: $0.15525 per second of the finished clip
Billed on the real clip length: the hold is duration_s + 1 second at that rate; on success you are charged the clip’s actual length and the rest of the hold comes back
Refused before any charge: audio that is not PCM WAV, is shorter than 5 s, or whose WAV header does not match its bytes; the job fails and the hold is refunded in full
Longer audio is cut: anything past duration_s (plus 0.25 s for encoder padding) is dropped before generation
extra is ignored by this model
Likeness and voice: you must have the right to use the likeness and voice of the person in the image and the audio (see the Acceptable Use Policy)

Image-to-video (I2V)

veo-3.1-fast and kling-3-pro accept an image_url. When present, the image becomes the first frame and the prompt drives the motion.

{
  "model": "veo-3.1-fast",
  "prompt": "the camera slowly pushes in, the subject blinks once",
  "image_url": "https://example.com/portrait.jpg",
  "duration_s": 4
}

Billing flow

POST creates a job and places a hold for estimated_cost (per-second list rate × duration).
On succeeded, the hold is finalized as a usage transaction at the actual cost.
On failed or expired, the hold is fully refunded — you only pay for clips you receive.

You can verify the charge on GET /v1/jobs/{id}: charged_amount is the final billed amount, estimated_cost is what was held up front.

Idempotency

Pass idempotency_key to make POST safe to retry. The same (api_key, idempotency_key) pair always returns the same job — no duplicate charges, no duplicate generations.

{
  "model": "veo-3.1-fast",
  "prompt": "...",
  "duration_s": 6,
  "idempotency_key": "campaign-spring-asset-7"
}

Models

Pick a model

veo-3.1-fast

kling-3-pro

kling-avatar-v2-pro

Image-to-video (I2V)

Billing flow

Idempotency

See also

​Pick a model

​veo-3.1-fast

​kling-3-pro

​kling-avatar-v2-pro

​Image-to-video (I2V)

​Billing flow

​Idempotency

​See also

Pick a model

veo-3.1-fast

kling-3-pro

kling-avatar-v2-pro

Image-to-video (I2V)

Billing flow

Idempotency

See also