Seedance 2.0 Fast Text-to-Video API

Overview

Seedance 2.0 Fast Text-to-Video is the speed-first Seedance endpoint for turning a written scene brief into a short audio-video clip. It is designed for creative work where the next revision needs to arrive quickly: changing a camera move, tightening the pacing, testing another line of dialogue, or comparing two action directions.

Fast occupies a different place from the other Seedance 2.0 tiers. Mini is the cost-first option for larger batches, while the mainline endpoint is the quality-first option for selected final outputs. Fast sits between them, prioritizing shorter iteration cycles without moving to the mainline model's 1080P or 4K delivery tiers.

The iCreat workflow is prompt-only. One text description defines the scene, action, camera, pacing, and optional sound. Output is available at 480P or 720P for 4–15 seconds, with fixed or automatic duration and fixed or adaptive aspect ratio.

Use the model through:

bytedance/seedance-2-0-fast/text-to-video

Key Features

Revisions begin with the Prompt, not asset preparation. Fast does not require a starting image or reference pack on this endpoint, so a new version can be tested by changing the written direction and submitting again.

Camera ideas can be compared quickly. Keep the subject and action unchanged while testing a tracking shot, handheld view, low-angle follow, slow push-in, or locked camera.

Pacing is easy to isolate. A fixed duration and one clear action make it possible to compare slower buildup, faster payoff, alternate transitions, or different endings without changing the entire concept.

Audio can develop with the visual. Set generate_audio to true to test dialogue, ambience, sound effects, or music in the same generation rather than evaluating the picture alone.

Fixed settings support controlled A/B tests. Keep ratio, resolution, duration, and audio settings unchanged when the goal is to judge one Prompt revision against another.

Automatic settings support early exploration. Use duration: -1 or adaptive when the scene can determine its own length or frame shape before delivery requirements are fixed.

480P and 720P match the review workflow. Use 480P for quick structural checks and 720P when motion, facial behavior, sound timing, or camera detail needs closer review.

Web-assisted context is optional. The iCreat web_search tool can add external context to a request, but it does not guarantee factual or visual accuracy.

Model Comparison

Seedance 2.0 Mini vs Fast vs Mainline Text-to-Video

Seedance 2.0 Fast vs Wan 2.7 SP vs HappyHorse 1.1 Spicy

Inputs

The required content array contains one text item. For useful iteration, describe the stable parts of the scene first, then identify the single element being tested, such as camera movement, action timing, dialogue, or sound treatment.

Chinese Prompts can contain up to 2,000 characters, and English Prompts can contain up to 2,000 words. Long descriptions can create competing priorities, so Fast works best when each revision changes one clear direction while the rest of the brief remains stable.

Parameters

Pricing

Total Cost = Unit Price × Output Video Duration

Quick Start

This request creates one continuous shot for rapid camera and timing review. Keep the scene description unchanged and replace only the camera instruction when comparing variants.

curl --fail-with-body --connect-timeout 10 --max-time 60 \
  -X POST https://api.icreat.ai/v1/task/submit/bytedance/seedance-2-0-fast/text-to-video \
  -H "Authorization: Bearer ${ICREAT_API_KEY}" \
  -H "Content-Type: application/json" \
  -d '{
    "content": [
      {
        "type": "text",
        "text": "Create an 8-second single-shot scene inside a small bakery at dawn. Begin close to steam rising from a tray of fresh bread. The camera makes a smooth low tracking move beside the tray as the baker slides it onto a wooden counter, brushes flour from one hand, and says, \"First batch is ready.\" Keep the movement continuous, the warm window light stable, and the action physically natural. Add quiet room ambience, the sound of the metal tray touching wood, and restrained acoustic music."
      }
    ],
    "generate_audio": true,
    "ratio": "16:9",
    "resolution": "720p",
    "duration": 8,
    "watermark": false
  }'

The request returns a task_id.

Submit request → Receive task_id → Poll status → Retrieve result

Query the task status until it reaches SUCCEEDED, then use the same task_id to retrieve the result.

Output

Seedance 2.0 Fast Text-to-Video uses an asynchronous task flow.

The submission request returns a task_id. Task status is checked through the query-status endpoint, and the completed result is retrieved through the get-result endpoint after the task succeeds.

Detailed result fields are not listed until they are confirmed from an actual iCreat response.

Use Cases

Rehearsing camera language before a final render. Test several camera paths around the same action, then move the preferred direction to the mainline endpoint only when higher-resolution delivery is needed.

Responding to time-sensitive creative feedback. Revise a shot after a stakeholder requests a faster opening, calmer ending, different reaction, or clearer product moment without rebuilding source assets.

Testing dialogue and sound timing. Compare alternate lines, pauses, sound cues, or music intensity while keeping the visual scene and duration stable.

Producing fast social-cut options. Explore different hooks, pacing patterns, camera energy, and endings for a short-form concept before choosing the versions worth further production.

Previsualizing difficult movement. Use a lower-resolution pass to judge whether an action, interaction, or camera move is understandable before committing to a more expensive final generation.

Limitations

Fast supports 480P and 720P only. It is intended for review and iteration, not for the 1080P or 4K delivery available from the mainline Seedance 2.0 endpoint.

The speed tier prioritizes turnaround and cost over maximum final quality. Demanding interactions, subtle facial detail, fine textures, and complex physical action may benefit from a final pass on the mainline model.

Because this endpoint starts from text alone, it cannot lock an exact face, product design, costume, or composition. Visual identity and small details may change across revisions even when the Prompt remains similar.

Changing several instructions at once makes comparisons less useful. A new action, camera move, duration, ratio, and sound treatment in the same revision can make it difficult to identify why one result improved or failed.

Generated dialogue may be unclear, audio can distort, and sound timing may vary. Readable text, logos, labels, and interface elements are also not guaranteed.

The provided iCreat schema does not expose a seed. The same Prompt and parameters can produce different movement, framing, timing, and fine details across runs.

FAQ

When should Fast be chosen instead of Mini?

Choose Fast when shorter revision cycles matter more than reaching the lowest possible cost per batch. Choose Mini when the main goal is producing a larger number of variations as economically as possible.

When should Fast be chosen instead of the mainline Seedance 2.0 endpoint?

Use Fast while testing scene structure, camera direction, pacing, dialogue, and sound. Move the selected concept to the mainline endpoint when it needs 1080P, 4K, or more demanding final-scene quality.

Should prompt tests use 480P or 720P?

Use 480P to check whether the scene structure, action, and camera idea work. Use 720P when facial behavior, motion detail, audio timing, or presentation quality must be judged more carefully.

How can two Prompt revisions be compared fairly?

Keep ratio, resolution, duration, audio settings, subject, and setting unchanged. Revise one instruction at a time, such as the camera move, action speed, spoken line, or ending.

Can the same Prompt reproduce the same video?

No exact match should be expected. The endpoint does not expose a seed, so repeated requests can vary in motion, framing, timing, sound, and fine detail.