ByteDance's newest video model. Generate up to 30 seconds in one pass, direct it with 50 image, video, and audio references, and get dialogue, music, and effects locked to the timeline.
What is Seedance 2.5?
Seedance 2.5 is ByteDance's newest video model, now in HM.AI's AI tool. It generates native 30-second clips in a single pass. Each generation accepts up to 50 references: 30 images, 10 video clips, and 10 audio files. Audio is created with the visuals, so sound arrives in sync. Included on HM.AI.
Why creators choose Seedance 2.5
Longer clips, deeper reference control, and sound that arrives finished. Here is what changes when you generate scenes with more control.
Create a beginning, middle, and end Thirty seconds means you can craft a narrative that actually tells a story. Your character holds across the scene, keeps the same face, the light holds, and nothing drifts at the seams. Your own material does the directing With 30 images, 10 clips, and 10 audio files in one generation, you can label each in the prompt, and the model reads it as a brief: this face, that palette, this camera move, that voice. Sound that lands finished Dialogue, music, and effects are written alongside the picture rather than laid over it afterwards. Footsteps hit on the footstep. That takes a whole pass out of your edit, not just a step out of your workflow. |
How to create with Seedance 2.5?
Seedance 2.5 runs inside HM.AI's AI Tool. Four steps from an empty prompt to a finished clip with sound.
Open the AI Tool Head to HM.AI's AI Tool and pick Seedance 2.5 from the video model list. Add your references and prompt Upload images, clips, or audio, then use the prompt to say what each one is for: subject, style, camera move, dialogue, or soundtrack. Set duration and framing Choose a length between 4 and 30 seconds, or leave it to Smart Duration. Pick an aspect ratio, or let the model read one from your prompt. Generate and refine Your clip comes back with audio already in sync. Extend it, edit a scoped region, or swap a reference and run it again. |
What Seedance 2.5 can do
Beyond duration and references, Seedance 2.5 handles multi-shot sequences, targeted edits within a defined time range, and clip extension, all within one generation.
Multi-shot generation Describe several shots in one prompt, and the model cuts between them, holding style and continuity across each change. Reference-driven editing Replace a subject, add or remove an element, or redraw a region. Limit the change to a time window, for example 0:02 to 0:05, and the rest of the clip stays untouched. Extension in both directions Continue a clip from its final frame, or generate the moment before it using the first frame as the endpoint. Style, rhythm, and camera logic carry across. |
Start creating with Seedance 2.5
| Generate 30-second clips from up to 50 references, alongside Kling, Veo, Sora, and every other leading AI video model in Artlist's AI Tool. One subscription, one commercial license, and every model in the same browser tab. |
Frequently asked questions
What is Seedance 2.5? Seedance 2.5 by ByteDance is an AI video generator that creates native 30-second clips from text, image, video, and audio references. Audio is generated in the same pass as the visuals, so it comes back in sync. It is available in HM.AI's AI Tool and included in Unlimited plans. |
How is Seedance 2.5 different from Seedance 2.0? Seedance 2.5 generates clips up to 30 seconds against 15, and accepts 50 multimodal references against 12. It adds audio-only references, adaptive aspect ratio, and Smart Duration. Resolution starts lower, so Seedance 2.0 remains the pick when 4K output matters more than duration. |
How long are Seedance 2.5 videos? Each clip runs 4 to 30 seconds, generated natively in one pass rather than stitched together. Set the duration manually, or use Smart Duration and the model selects a natural length in the 4 to 15 second range. Duration affects credit usage. Longer clips consume more credits per generation. See Artlist's pricing page for the full breakdown. |
| How does Seedance 2.5 keep audio in sync? Seedance 2.5 generates dialogue, music, and sound effects in the same model pass as the visuals, so timing is locked during creation rather than aligned afterward. There is no post-production sync step and no second pass. Write what you want to hear in your prompt, or upload an audio reference, and the model incorporates it during generation. |
| What reference inputs does Seedance 2.5 support? Up to 30 images, 10 video clips, and 10 audio files per generation, 50 in total. Audio-only references are new in 2.5. Assign each reference a role through the prompt: subject, style, camera, motion, dialogue, music, or SFX. |
What resolution does Seedance 2.5 support? 480p and 720p at 24 frames per second (FPS) at launch. For maximum output resolution today, use Seedance 2.0 in the HM.AI AI Tool, which delivers up to 4K. Both models sit under the same subscription, so you can pick per shot. |
Is Seedance 2.5 included in Unlimited? No. Seedance 2.5 isn't part of the Unlimited plan — but you can still use it with a paid AI plan, where you generate using credits. Just keep in mind that longer clips and heavier reference stacks use more credits than shorter ones. |
What languages can Seedance 2.5 speak? Eleven: English, Chinese, Spanish, Portuguese, Japanese, Korean, Arabic, Vietnamese, Thai, Indonesian, and Malay. Feed a reference audio clip to steer accent and delivery, or describe the voice you want directly in the prompt. Dialogue is generated alongside the visuals rather than dubbed on afterwards. |
