How to Get Started With Seedance 2.5 on Dreamina, Runway, and fal
A beginner’s guide to Seedance 2.5: where to use it on Dreamina, Runway, and fal, how its modes and 50-reference limit work, prompting tips, and what it costs.
Seedance 2.5 is ByteDance’s latest AI video model, released July 31. It generates up to 30 seconds of video with audio in a single pass and accepts up to 50 reference files, and it can follow a shot-by-shot timeline written into your prompt.
It’s available in several places, and each one works a little differently. Here’s how to get started on Dreamina, Runway, and fal, what the main modes do, and how to write prompts that get better results.
What’s New in Seedance 2.5
The biggest change is length. Seedance 2.5 generates a single continuous clip of up to 30 seconds, double the single-pass length of earlier versions.
It also generates audio, including dialogue and vocals in more than ten languages with matching lip sync. For background on how that works, see What Is Native Audio in AI Video?.
Prompts can be written as timed blocks that set shot size, camera, and transitions, and the model follows that schedule, including hard cuts. It can also edit existing footage from plain-language instructions and extend clips forward or backward.
Where to Use Seedance 2.5
Dreamina is ByteDance’s own creative platform and the most direct way to try Seedance 2.5. It offers a free starting tier, and paid plans add more credits. Exact pricing is available on Dreamina’s website.
Runway offers Seedance 2.5 in its Tools, Workflows, and Agent features, alongside Runway’s own models. It’s a good option if you already have a Runway plan and want to use your credits across models.
fal offers Seedance 2.5 through its API, which suits developers and anyone building automated workflows. For more on the platform, see What Is fal.ai?.
Seedance 2.5 is also available in ComfyUI through Partner Nodes, with ready-made reference-to-video and text-to-video templates.
The Main Modes
Runway’s help documentation breaks Seedance 2.5 into four modes, which map to how the model works on most platforms.
- Reference: Combine images, video, and audio, and give each one a role in the shot.
- Keyframe: Set a first and last frame and describe the motion between them.
- Edit: Change an existing video with a prompt, or by sketching on a frame to point at what should change.
- Extend: Continue a video forward or backward, producing the new footage as a separate clip.
For more on start and end frames, see What Are Keyframes in AI Video?.
Using References
Seedance 2.5 accepts up to 50 references per generation: 30 images, 10 videos, and 10 audio clips. Reference them by name in your prompt, like “@Image 1,” and say what each one is for.
For example, you might use @Image 1 for the character, @Image 2 for the location, and an audio clip for the dialogue. Clear roles help the model keep characters and settings consistent.
References are one of the strongest tools for keeping characters the same across shots, covered in How to Keep Characters Consistent in AI Video.
Your First Generation, Step by Step
- Start in Reference mode. Add one or two images for your character or setting.
- Write a short prompt. Describe the action, camera, and mood, and refer to your references by name.
- Set a short duration. Start with 5 to 8 seconds while you test.
- Generate a draft. On Runway, Draft mode creates a fast 480p preview you can enhance to 1080p later.
- Adjust and regenerate. Refine one thing at a time, like the camera move or the character’s action.
- Go longer once it works. Expand to 15 or 30 seconds and add timed shot blocks.
Prompting Tips
Write prompts that describe what you want to see, not what you want to avoid. Runway’s guidance recommends positive, outcome-focused prompts.
For longer clips, structure the prompt as timed blocks, such as “0–5s: wide shot of the street” and “5–10s: close-up on her face.” Runway notes that timestamps guide pacing but aren’t frame-accurate, so plan to fine-tune timing in your edit.
Camera language helps too. For a list of terms that work well in AI video, see Camera Movement Prompts for AI Filmmaking.
What It Costs
On Runway, Seedance 2.5 costs 68 credits per second at 1080p, 30 at 720p, and 20 at 480p or Draft. Using video references adds credits based on the length of the input video.
For scale, Runway’s $95 Max plan includes 9,500 credits per month, which covers about 2 minutes and 20 seconds of 1080p Seedance 2.5 output.
On fal, pricing is based on resolution and length, at roughly $0.47 per second at 720p and $0.22 per second at 480p. A 10-second 720p clip costs about $4.73.
Competitive Context
Seedance 2.5’s 30-second single-pass generation is longer than most competing models, which typically generate shorter clips that have to be extended or stitched. Its 50-reference limit is also among the highest available.
The main alternatives include Kling, Google’s Veo, Runway’s own Gen-4.5, and MiniMax H3. Each has different strengths in motion, realism, and audio, and many filmmakers use more than one.
ByteDance is also building a creator program around its models through Dreamina, covered in ByteDance’s Dreamina Launches “Originals”.
The Signal in the Noise
For beginners, Dreamina is the simplest starting point, Runway makes sense if you already use its tools, and fal is best for developers. Starting with short drafts keeps costs down while you learn how the model responds.
The 30-second length and timeline prompts make Seedance 2.5 especially useful for sequences with multiple shots. Expect to fine-tune timing and pick the best takes in your editor.
The Details
- Released: July 31, 2026
- Max length: 30 seconds per generation
- References: Up to 50 (30 images, 10 videos, 10 audio)
- Audio: Native dialogue and vocals with lip sync
- Modes: Reference, Keyframe, Edit, Extend
- Runway: 68 credits/sec (1080p), 30 (720p), 20 (480p/Draft)
- fal: ~$0.47/sec (720p), ~$0.22/sec (480p)
- Dreamina: Free tier available; exact pricing on Dreamina’s website
- Also on: ComfyUI Partner Nodes
Have you tried Seedance 2.5 yet? Share where you’re using it and your best prompting tip in the comments.
