Media capabilities

Generate video

Video generation follows the same durable contract as Byta images. Submission returns immediately with HTTP 202, work continues in a background worker, and clients poll the job until a private video file is ready.

Quick 202 · durable background job

Catalog last synchronized: October 3, 2026 (UTC)

Asynchronous by design

Video creation can take longer than a proxy or browser request should remain open. Byta never performs generation or media download in the foreground API request.

Reserve and submit

Credits are reserved and a durable job is created.

Process in background

The worker generates, downloads, and stores the output.

Poll and download

The final file appears in videos[] when completed.

Safe retries

Reuse the same Idempotency-Key after a timeout or transport failure. Byta returns the original job instead of starting or charging for a duplicate generation.

Video models and options

Duration and resolution combinations are model-specific. Query GET /v1/models before presenting options in a dynamic client.

Model	Best for	Durations	Resolutions	References	Runtime
`byta-video-1-5`	Our signature default model for reliable everyday video generation. Safety check is applied.	6s, 10s, 15s	480p, 720p, 1080p	Up to 7	~45s
`bytis-video-1-4`	Fast and stable video generation model, no content moderation.	5s, 10s	480p, 720p	Up to 1	~20s
`bytis-video-1-5`	Fast video generation model with strong prompt understanding and cinematic output. No content moderation.	5s, 8s, 12s	480p, 720p	Up to 1	~150s
`bytis-video-1-5b`	Powerful video model fine-tuned for high-resolution output. No moderation.	5s, 8s, 12s	480p, 720p, 1080p	Up to 1	~120s
`seedance-2-0-mini`	Fast, economical video generation with native audio and optional first-to-last-frame animation. Safety check is applied.	6s, 10s, 15s	480p, 720p	Up to 2	~90s
`minimax-h3`	Open-weights video generation with native stereo audio and optional first-to-last-frame animation. Safety check is applied.	6s, 10s, 15s	480p, 768p	Up to 2	~150s
`minimax-h3-spicy`	MiniMax H3 text-to-video with Spicy image-to-video and optional first-to-last-frame animation. The image route has no upstream safety check; the text route is standard H3.	6s, 10s, 15s	480p, 768p	Up to 2	~150s
`minimax-h3-turbo`	Turbo MiniMax H3 generation optimized for faster inference with optional first-to-last-frame animation. Safety check is applied.	6s, 10s, 15s	480p, 768p	Up to 2	~90s

Which model should I use?

Need	Choose	Why
Current default	`byta-video-1-5`	Our signature default model for reliable everyday video generation. Safety check is applied.
Alternative active model	`bytis-video-1-4`	Fast and stable video generation model, no content moderation.
Alternative active model	`bytis-video-1-5`	Fast video generation model with strong prompt understanding and cinematic output. No content moderation.
Alternative active model	`bytis-video-1-5b`	Powerful video model fine-tuned for high-resolution output. No moderation.
Alternative active model	`seedance-2-0-mini`	Fast, economical video generation with native audio and optional first-to-last-frame animation. Safety check is applied.
Alternative active model	`minimax-h3`	Open-weights video generation with native stereo audio and optional first-to-last-frame animation. Safety check is applied.
Alternative active model	`minimax-h3-spicy`	MiniMax H3 text-to-video with Spicy image-to-video and optional first-to-last-frame animation. The image route has no upstream safety check; the text route is standard H3.
Alternative active model	`minimax-h3-turbo`	Turbo MiniMax H3 generation optimized for faster inference with optional first-to-last-frame animation. Safety check is applied.

Pricing

Credits are reserved when the job is accepted. The exact row is returned by GET /v1/models as videoOptions.

Model	Duration	Resolution	Modes	Credits
`byta-video-1-5`	6 seconds	480p	T2V · I2V	45
`byta-video-1-5`	6 seconds	720p	T2V · I2V	90
`byta-video-1-5`	10 seconds	480p	T2V · I2V	68
`byta-video-1-5`	10 seconds	720p	T2V · I2V	135
`byta-video-1-5`	15 seconds	480p	T2V	113
`byta-video-1-5`	15 seconds	720p	T2V	225
`byta-video-1-5`	15 seconds	480p	I2V	113
`byta-video-1-5`	15 seconds	720p	I2V	225
`byta-video-1-5`	6 seconds	1080p	T2V · I2V	180
`byta-video-1-5`	10 seconds	1080p	T2V · I2V	270
`byta-video-1-5`	15 seconds	1080p	T2V · I2V	450
`bytis-video-1-4`	5 seconds	480p	T2V · I2V	120
`bytis-video-1-4`	5 seconds	720p	T2V · I2V	225
`bytis-video-1-4`	10 seconds	480p	T2V · I2V	225
`bytis-video-1-4`	10 seconds	720p	T2V · I2V	450
`bytis-video-1-5`	5 seconds	480p	T2V	125
`bytis-video-1-5`	5 seconds	720p	T2V · I2V	250
`bytis-video-1-5`	8 seconds	480p	T2V	200
`bytis-video-1-5`	8 seconds	720p	T2V · I2V	400
`bytis-video-1-5`	12 seconds	720p	I2V	550
`bytis-video-1-5b`	5 seconds	480p	T2V	300
`bytis-video-1-5b`	5 seconds	480p	I2V	250
`bytis-video-1-5b`	5 seconds	720p	T2V	225
`bytis-video-1-5b`	8 seconds	480p	I2V	400
`bytis-video-1-5b`	5 seconds	720p	I2V	500
`bytis-video-1-5b`	8 seconds	480p	T2V	400
`bytis-video-1-5b`	8 seconds	720p	T2V	360
`bytis-video-1-5b`	8 seconds	720p	I2V	800
`bytis-video-1-5b`	5 seconds	1080p	I2V	750
`bytis-video-1-5b`	8 seconds	1080p	I2V	1200
`bytis-video-1-5b`	12 seconds	720p	T2V	540
`seedance-2-0-mini`	6 seconds	480p	T2V · I2V	165
`seedance-2-0-mini`	6 seconds	720p	T2V · I2V	165
`seedance-2-0-mini`	10 seconds	480p	T2V · I2V	275
`seedance-2-0-mini`	10 seconds	720p	T2V · I2V	275
`seedance-2-0-mini`	15 seconds	480p	T2V · I2V	413
`seedance-2-0-mini`	15 seconds	720p	T2V · I2V	413
`minimax-h3`	6 seconds	480p	T2V · I2V	300
`minimax-h3`	6 seconds	768p	T2V · I2V	600
`minimax-h3`	10 seconds	480p	T2V · I2V	500
`minimax-h3`	10 seconds	768p	T2V · I2V	1000
`minimax-h3`	15 seconds	480p	T2V · I2V	750
`minimax-h3`	15 seconds	768p	T2V · I2V	1500
`minimax-h3-spicy`	6 seconds	480p	T2V · I2V	300
`minimax-h3-spicy`	6 seconds	768p	T2V · I2V	600
`minimax-h3-spicy`	10 seconds	480p	T2V · I2V	500
`minimax-h3-spicy`	10 seconds	768p	T2V · I2V	1000
`minimax-h3-spicy`	15 seconds	480p	T2V · I2V	750
`minimax-h3-spicy`	15 seconds	768p	T2V · I2V	1500
`minimax-h3-turbo`	6 seconds	480p	T2V · I2V	360
`minimax-h3-turbo`	6 seconds	768p	T2V · I2V	360
`minimax-h3-turbo`	10 seconds	480p	T2V · I2V	600
`minimax-h3-turbo`	10 seconds	768p	T2V · I2V	600
`minimax-h3-turbo`	15 seconds	480p	T2V · I2V	900
`minimax-h3-turbo`	15 seconds	768p	T2V · I2V	900

Configuration

Generation mode, duration, resolution, and aspect ratio are validated against the selected model before credits are reserved.

Aspect ratios

These are the complete active aspect-ratio choices returned for each model.

Model	Accepted values	Meaning
`byta-video-1-5`	`auto · 16:9 · 9:16 · 3:2 · 2:3 · 1:1`	Use a value returned for this model.
`bytis-video-1-4`	`auto · 16:9 · 9:16 · 1:1`	Use a value returned for this model.
`bytis-video-1-5`	`auto · 16:9 · 9:16`	Use a value returned for this model.
`bytis-video-1-5b`	`auto · 21:9 · 16:9 · 4:3 · 1:1 · 3:4 · 9:16`	Use a value returned for this model.
`seedance-2-0-mini`	`auto · 16:9 · 9:16 · 1:1 · 4:3 · 3:4 · 21:9`	Use a value returned for this model.
`minimax-h3`	`auto · 16:9 · 9:16 · 1:1 · 4:3 · 3:4 · 21:9`	Use a value returned for this model.
`minimax-h3-spicy`	`auto · 16:9 · 9:16 · 1:1 · 4:3 · 3:4 · 21:9`	Use a value returned for this model.
`minimax-h3-turbo`	`auto · 16:9 · 9:16 · 1:1 · 4:3 · 3:4 · 21:9`	Use a value returned for this model.

OpenAI-style sizes and resolutions

size sets both resolution and the closest model-supported aspect ratio. You can instead send resolution and aspectRatio separately.

size	Resolution	Shape
`854x480`	480p	Landscape / 16:9
`480x854`	480p	Portrait / 9:16
`480x480`	480p	Square / 1:1
`1280x720`	720p	Landscape / 16:9
`720x1280`	720p	Portrait / 9:16
`720x720`	720p	Square / 1:1

Duration defaults

When public API duration or resolution is omitted, Byta uses the first activevideoOptions row that supports the request mode. The t2vEnabled andi2vEnabled fields describe mode availability. Explicit values are clearer and recommended for reproducible billing.

Create a video job

Send JSON using the same Bearer API key and idempotency pattern as image generation.

POST/v1/videos/generations

Returns 202 Accepted

Text-to-video

import os
import requests

response = requests.post(
    "https://api.byta-ai.com/v1/videos/generations",
    headers={
        "Authorization": f"Bearer {os.environ['BYTA_API_KEY']}",
        "Idempotency-Key": "launch-film-001",
    },
    json={
        "model": "byta-video-1-5",
        "prompt": "A slow cinematic push through silver grass at sunrise",
        "seconds": 6,
        "size": "1280x720",
    },
)
response.raise_for_status()
job = response.json()
print(job["statusUrl"])

202 Accepted · JSON

{
  "id": "6aa0dca3ec6b40f89cbcf006a8f63ca0",
  "jobId": "6aa0dca3ec6b40f89cbcf006a8f63ca0",
  "object": "video_generation",
  "mode": "video",
  "status": "running",
  "statusUrl": "https://api.byta-ai.com/v1/videos/generations/6aa0dca3ec6b40f89cbcf006a8f63ca0",
  "durationSeconds": 6,
  "resolution": "480p",
  "creditsReserved": 45,
  "modelId": "byta-video-1-5",
  "modelName": "Byta-Video-1.5"
}

Poll job status

Poll every 3–5 seconds. running is non-terminal; completed and failed are terminal.

GET/v1/videos/generations/{jobId}

cURL · job status

curl https://api.byta-ai.com/v1/videos/generations/6aa0dca3ec6b40f89cbcf006a8f63ca0 \
  -H "Authorization: Bearer $BYTA_API_KEY"

200 OK · completed

{
  "id": "6aa0dca3ec6b40f89cbcf006a8f63ca0",
  "object": "video_generation",
  "status": "completed",
  "durationSeconds": 6,
  "resolution": "720p",
  "creditsCharged": 150,
  "videos": [
    {
      "id": "8d7f620318204948a35b3dfab6f1ee42",
      "mimeType": "video/mp4",
      "sizeBytes": 12840232,
      "url": "https://api.byta-ai.com/v1/videos/generations/6aa0.../files/8d7f..."
    }
  ]
}

Download the completed video

Output URLs are private. Send the same Bearer key used to create and poll the job. Range requests are supported for streaming clients.

GET/v1/videos/generations/{jobId}/files/{fileId}

Add ?download=1 for attachment disposition

cURL · MP4 download

curl "https://api.byta-ai.com/v1/videos/generations/6aa0.../files/8d7f...?download=1" \
  -H "Authorization: Bearer $BYTA_API_KEY" \
  --output generated-video.mp4

Image-to-video

Provide an owned Studio file, a public image URL, or an inline base64 image. The model uses it as visual guidance for the generated clip.

cURL · image-to-video

curl https://api.byta-ai.com/v1/videos/generations \
  -H "Authorization: Bearer $BYTA_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: image-to-video-001" \
  -d '{
    "model": "byta-video-1-5",
    "prompt": "The camera moves slowly forward while the fabric moves in a light breeze.",
    "seconds": 5,
    "size": "854x480",
    "input_reference": {"file_id": "owned_image_file_id"}
  }'

JSON · accepted reference shapes

// Existing image uploaded to Byta Studio
{"input_reference": {"file_id": "owned_image_file_id"}}

// Publicly reachable image
{"input_reference": {"image_url": "https://cdn.example.com/reference.jpg"}}

// Inline base64 image
{"input_reference": {"image_url": "data:image/jpeg;base64,/9j/4AAQ..."}}

Import and ownership

file_idmust identify an existing Byta Studio image owned by the API-key account. Public URLs and base64 data URIs are verified, imported into that account's Studio storage, and count toward its upload quota. JPEG, PNG, and WebP inputs up to 20 MiB are accepted.

API reference

promptrequired

string

Description of subject, scene, motion, camera, lighting, and mood.

modeloptional

string

An active video model ID from GET /v1/models. modelId is also accepted.

secondsoptional

integer

A duration supported by the selected model. durationSeconds is also accepted.

sizeoptional

string

Width x height. resolution and aspectRatio are also accepted separately.

input_referenceoptional

object

Exactly one file_id or image_url. image_url accepts a public HTTP(S) URL or base64 data URI.

Errors

Fix request-validation failures before resubmitting. Preserve the same Idempotency-Key only for a transport retry of the same logical request.

Code	Meaning	What to do
`unsupported_size`	The size is not one of the six public aliases.	Choose a documented size or send resolution and aspectRatio.
`invalid_input_reference`	input_reference is malformed or ambiguous.	Send exactly one file_id or image_url.
`unsafe_reference_url`	URL resolves to a non-public address.	Use a public image host.
`invalid_reference_image`	Reference bytes are not a valid supported image.	Send JPEG, PNG, or WebP up to 20 MiB.
`insufficient_credits`	The fixed job price cannot be reserved.	Add credits or select a lower-cost option.
Model validation	Duration, resolution, aspect ratio, or reference count is unsupported.	Read availableOptions or availableAspectRatios and choose a supported row.