Seedance V1.5 Pro on Novita AI: Text-to-Video vs Image-to-Video, Audio, and Silent Modes

Seedance V1.5 Pro on Novita AI is easiest to choose by starting with the asset you already have: use text-to-video when the scene starts from a prompt, use image-to-video when a first frame or first-and-last-frame setup should guide the result, keep output silent for early visual iteration, add audio only when speech, sound effects, or music are part of acceptance, and use flex service tier for latency-insensitive batch work.

Seedance V1.5 Pro Quick Selection Table

Decision	Start here	Why
You only have a written scene	Text-to-video	The T2V endpoint requires a prompt and can choose an adaptive aspect ratio from the prompt.
You have a product shot, character frame, storyboard frame, or approved visual	Image-to-video	The I2V endpoint requires image plus prompt, so the first frame anchors the generated clip.
You need a controlled start and ending	Image-to-video with last_image	The I2V API supports optional last_image for first-and-last-frame generation.
You need generated speech, sound effects, or background music	Set generate_audio: true	Audio is generated from the prompt and visual content when enabled.
You will add your own audio later	Set generate_audio: false	Silent output avoids paying for generated audio you will discard.
You need a user-facing result sooner	Use service_tier: "default"	The docs describe default as online inference for time-sensitive scenarios.
You can wait and want lower unit cost	Use service_tier: "flex"	The docs describe flex as offline inference at 50% of online-mode price.
You are exploring prompts	Start at 480p silent flex when latency is not important	It keeps early exploration cheaper before regenerating finalists.
You are reviewing near-final clips	Move to 720p, and enable audio only if it is part of acceptance	720p costs more than 480p, and audio doubles the listed per-second rate.

Seedance V1.5 Pro Modes Available on Novita AI

Novita AI exposes Seedance V1.5 Pro through separate asynchronous video APIs:

Mode	Novita API path	Required inputs	Optional inputs and controls
Seedance V1.5 Pro Text to Video	POST /v3/async/seedance-v1.5-pro-t2v	prompt	fps, seed, ratio, duration, watermark, resolution, camera_fixed, service_tier, generate_audio, execution_expires_after
Seedance V1.5 Pro Image to Video	POST /v3/async/seedance-v1.5-pro-i2v	image, prompt	fps, seed, ratio, duration, watermark, last_image, resolution, camera_fixed, service_tier, generate_audio, execution_expires_after

Both APIs return a task ID first, not a finished video. Applications should call the Novita AI Task Result API to check status and retrieve generated media after the task completes.

The shared request controls matter for product design. The docs list duration as a value from 4 to 12 seconds, fps as 24, ratio options including 16:9, 4:3, 1:1, 3:4, 9:16, 21:9, and adaptive, and resolution options of 480p, 720p, and 1080p. The execution_expires_after field defaults to 172800 seconds, with a documented range from 3600 to 259200 seconds.

Seedance V1.5 Pro Text-to-Video vs Image-to-Video

Choose text-to-video when the idea is still verbal. If your workflow begins with a script line, camera direction, subject description, or creative brief, T2V keeps the first pass lightweight. It is the cleaner mode for ideation, prompt exploration, and cases where exact product or character identity is not the main acceptance criterion.

Choose image-to-video when the first visual matters. I2V is better for product stills, design frames, character references, thumbnails, storyboard panels, and brand-controlled visuals. It does not guarantee perfect identity preservation in every frame, but it gives generation a concrete starting point that prompt-only T2V does not have.

The extra I2V-only field is last_image. Use it when a clip needs to move from a known opening frame toward a known ending frame. That is useful for transition tests, short product reveals, simple before-and-after motion, and storyboarded scenes where the final composition is already decided. When the first and last frame aspect ratios differ, the docs say the last frame is automatically cropped to match the first frame, so teams should prepare matching aspect ratios when visual framing is important.

For prompt-only concepts, T2V is usually the first test. For asset-led concepts, I2V is usually the first test. If you already know the starting frame, do not force the model to rediscover it from text.

Seedance V1.5 Pro Audio vs Silent Output

Seedance V1.5 Pro has a direct audio switch through generate_audio. When it is true, the generated video includes synchronized audio such as automatically generated speech, sound effects, and background music based on the prompt and visual content. When it is false, the output is silent.

Audio changes prompt writing. If generated speech is part of the test, write the spoken line explicitly and separate it from visual action and ambient sound. A vague scene prompt may be enough for visual exploration, but it is usually not enough for reliable audio review.

Audio also changes cost. Current Novita pricing page data checked on June 22, 2026 lists audio rows at twice the corresponding silent rows for Seedance V1.5 Pro T2V and I2V at 480P and 720P. If your product will add a voiceover, music bed, or sound design in a separate pipeline, generate silent video for early passes. If synchronized speech or scene audio is part of the actual product requirement, test audio early because it affects both review criteria and budget.

Seedance V1.5 Pro Online vs Flex Batch Processing

The service_tier field is the main processing-tier choice. The Seedance V1.5 Pro docs describe default as online inference with lower RPM and concurrency quotas for time-sensitive scenarios. They describe flex as offline inference with higher TPD quotas at 50% of the online-mode price for latency-insensitive scenarios.

For application design, treat default as the interactive lane and flex as the batch lane. A user-facing creative tool where someone is waiting in the UI should start with default. A nightly prompt sweep, dataset-style creative evaluation, large internal review queue, or non-urgent campaign batch can use flex.

This distinction is separate from the asynchronous API pattern. Both modes still return a task ID first. The difference is the processing tier and cost profile, not whether the API is synchronous.

Seedance V1.5 Pro Resolution: 480p vs 720p vs 1080p

The current Seedance V1.5 Pro docs list 480p, 720p, and 1080p as supported resolution values. The current pricing page data exposes detailed 480P and 720P price rows for Seedance V1.5 Pro T2V and I2V. The same page data did not expose a matching Seedance V1.5 Pro 1080P price row in the data checked for this revision, so production budget plans should verify the live pricing page or console before using 1080p at scale.

For most teams, 480p is the right exploration setting. It is enough to judge prompt direction, rough motion, whether an image reference is helping, and whether audio should remain in scope. Regenerate finalist candidates at 720p when visual review matters more than exploration cost.

Use 1080p only after confirming both availability and billing for the exact job shape you plan to run. The API field supports it, but budgeting from the visible 480P and 720P rows alone would be incomplete.

Seedance V1.5 Pro Pricing by Resolution, Audio, and Flex Batch

Current Novita pricing page data checked on June 22, 2026 lists the following Seedance V1.5 Pro prices for both text-to-video and image-to-video. The same prices appear for first-frame I2V and first-and-last-frame I2V rows.

Mode	Resolution	Output	Online, default	Flex batch, flex
T2V or I2V	480P	Silent	$0.012/s	$0.006/s
T2V or I2V	480P	Audio	$0.024/s	$0.012/s
T2V or I2V	720P	Silent	$0.026/s	$0.013/s
T2V or I2V	720P	Audio	$0.052/s	$0.026/s

The cost multipliers are the useful part:

• Audio costs twice the corresponding silent row.
• Flex batch costs half the corresponding online row.
• 720P costs a little more than twice the corresponding 480P row.
• T2V and I2V have the same listed rate at the same resolution, audio setting, and service tier.

For a 10-second 720P audio clip, the listed online cost is 10 x $0.052, or $0.52. The same clip through flex is 10 x $0.026, or $0.26, when latency is acceptable. For a 10-second 480P silent exploration clip, the listed flex cost is 10 x $0.006, or $0.06.

That makes the early testing path straightforward: reduce cost by turning off audio, using 480P, and using flex when review does not need an immediate result. Raise cost only when the clip is close enough to evaluate at final quality.

Which Seedance V1.5 Pro Mode Should You Test First?

Start with T2V when you are still searching for the scene. Write five to ten compact prompts, keep the duration short enough for review, use 480p, and leave audio off unless speech is the core idea. The goal is to discover whether the concept works before introducing asset constraints.

Start with I2V when the asset is already the concept. If a product frame, character reference, or approved composition is non-negotiable, use that image from the first test. A T2V prompt may create a plausible scene, but it will not know the exact reference image unless you provide it.

Start with I2V plus last_image when the transition matters more than open-ended motion. This is the right first test for a known beginning and ending composition. Keep the first and last images close in aspect ratio and avoid overloading the prompt with too many unrelated actions.

Start silent unless audio is part of acceptance. Once the visual direction is acceptable, enable generate_audio if needed and write the spoken line or sound cue explicitly.

Start with flex for queues and default for live UI. If the generation is going into a background review pipeline, use flex and spend the savings on more variants. If a user is waiting in a product surface, use default and design the UI around task progress.

Seedance V1.5 Pro API Workflow on Novita AI

A production workflow should treat both Seedance V1.5 Pro modes as asynchronous jobs:

1. Choose T2V or I2V based on whether you have an input image.
2. Choose silent or audio with generate_audio.
3. Choose default or flex with service_tier.
4. Choose resolution and duration.
5. Submit the task to the model-specific endpoint.
6. Store the returned task ID.
7. Poll the Task Result API until the task succeeds or fails.
8. Store the returned video URL before the URL expires.

Here is a minimal text-to-video request shape:

curl --location --request POST 'https://api.novita.ai/v3/async/seedance-v1.5-pro-t2v' \
  --header 'Content-Type: application/json' \
  --header "Authorization: Bearer ${NOVITA_API_KEY}" \
  --data-raw '{
    "prompt": "A product demo clip showing a compact smart lamp turning on in a clean studio. Slow push-in camera movement, soft shadows, neutral background.",
    "duration": 5,
    "resolution": "480p",
    "ratio": "16:9",
    "service_tier": "flex",
    "generate_audio": false,
    "watermark": false
  }'

Here is a minimal image-to-video request shape:

curl --location --request POST 'https://api.novita.ai/v3/async/seedance-v1.5-pro-i2v' \
  --header 'Content-Type: application/json' \
  --header "Authorization: Bearer ${NOVITA_API_KEY}" \
  --data-raw '{
    "image": "https://example.com/product-frame.png",
    "prompt": "Animate the lamp with a gentle light pulse while the camera makes a subtle forward move. Keep the product centered and avoid text overlays.",
    "duration": 5,
    "resolution": "480p",
    "ratio": "adaptive",
    "service_tier": "flex",
    "generate_audio": false,
    "watermark": false
  }'

After submission, use the returned task ID with the Task Result API. Build for queued, processing, succeeded, and failed states instead of assuming a video URL will be available immediately.

Final Recommendation for Seedance V1.5 Pro

For most teams, the best first Seedance V1.5 Pro test is 480P, silent, and flex: T2V if you only have a written concept, I2V if you have a source image. Move to 720P after the prompt or reference image is close. Enable audio only when generated speech, sound effects, or music are part of the review. Use default when the result is user-facing and time-sensitive; use flex when the task can wait.

That sequence keeps the first evaluation cheap without changing the core API flow. Once a clip passes visual review, raise the resolution, turn on audio if needed, and run a smaller set of final candidates.

FAQ

Is Seedance V1.5 Pro available for both text-to-video and image-to-video on Novita AI?

Yes. Novita AI documents separate Seedance V1.5 Pro T2V and I2V endpoints: POST /v3/async/seedance-v1.5-pro-t2v and POST /v3/async/seedance-v1.5-pro-i2v.

Does Seedance V1.5 Pro generate audio?

Yes. The generate_audio field controls whether the output includes synchronized audio. Set it to true for generated speech, sound effects, or background music, and false for silent video.

Is image-to-video more expensive than text-to-video?

The current pricing page data checked for this article lists the same Seedance V1.5 Pro rates for T2V and I2V at the same resolution, audio setting, and service tier.

What is the difference between online and flex processing?

The docs describe default as online inference for time-sensitive scenarios and flex as offline inference for latency-insensitive scenarios at 50% of the online-mode price.

Should I test 480P or 720P first?

Use 480P for broad prompt and reference-image exploration. Use 720P when the concept is close enough for more serious visual review. Verify 1080P billing before using it in a large run.

Recommended articles

• Seedance 1.5 Pro on Novita AI: Complete Developer Guide
• Kling V2.5 Turbo API on Novita AI: Text-to-Video and Image-to-Video
• Sora Alternatives for AI Video Generation