AI Video & Image Model Profiles (2026): segment length, references, audio and list prices
Specs and list prices of 14 models — Seedance, MiniMax H3, Wan 3.0, Kling, Veo, Sora, Runway Gen-4.5, Vidu, PixVerse V6, SeedDream, Qwen-Image — and which platforms expose them.
Segment length and price
Video models: seconds per generation
11 of 11 video models publish a segment length · longer segments allow more cuts per call
051015202530
Published per-second prices (API list + platform prices)
Only prices verifiable on official pages; undisclosed ones are named in the caption
00.250.50.751
Video model spec table
| Seedance 2.0 / 2.5 | Seedance 2.5 | MiniMax Hailuo H3 | Wan 3.0 / 3.0 Prime | Kling 3.0 / Omni | Veo 3.1 | Sora | Runway Gen-4.5 / Aleph | Vidu | PixVerse V6 | LTX-2.5 | |
|---|---|---|---|---|---|---|---|---|---|---|---|
| Segment length | 4–15 s (2.0); 4–30 s (2.5), with multi-round extension on 2.5 | 4–30 s per generation, with multi-round extension | 4–15 s (H3); 5–15 s (H3-Max) | 2–30 s per generation (or smart duration); video extension supported | 3–15 s per generation (VIDEO 3.0 / 3.0 Omni); video extension available | 4, 6 or 8 s per generation (8 s at 1080p/4K or with reference images); extension adds 7 s up to 20 times (Veo 3.1 / Fast, 720p) | The API reference lists 4, 8 or 12 s; OpenAI's video guide also lists 16 and 20 s. Extensions add up to 20 s each, 120 s in total | 2–10 s per generation (API, Gen-4.5) | 1–16 s (Q3 text/image-to-video); 3–16 s (Q3 reference-to-video) | 1–15 s per generation (extendable) | 6–20 s on the API: Fast up to 20 s at 1080p and 10 s at 1440p/4K; Pro up to 10 s; Extend and Retake on Pro |
| Resolution | 480p / 720p / 1080p / 4K (2.0); 480p / 720p / 1080p (2.5); 2.0 Fast and Mini 480p / 720p | 480p / 720p / 1080p | 768p / 2K (H3); 480p / 768p (H3-Max) | 480p / 720p / 1080p | 720p / 1080p; 4K on the API price list | 720p / 1080p / 4K (no 4K on Lite) | 720p (sora-2); up to 1080p (sora-2-pro) | 720p output (API ratios 1280:720 / 720:1280); up to 4K via upscale | 540p / 720p / 1080p | 360p / 540p / 720p / 1080p | 720p / 1080p / 1440p / 4K, 25 or 50 fps |
| Audio | Native audio (generate_audio, on by default) | Native audio (generate_audio, on by default) | Native stereo speech + ambience; ≤3 reference audio clips, ≤15 s total | Native audio; dialogue delimited by quotes, 'no dialogue' as control phrase | Native audio with dialogue in Chinese, English, Japanese, Korean and Spanish; voices can be bound to elements; lip-sync | Native audio and dialogue (always on) | With audio | Audio; custom voices (Pro+) | Native speech, sound effects and music on Q3 (on by default in the API; speech-only mode available) | Native audio switch; lip-sync | Joint audio-video generation; audio-to-video |
| References | First/last frame, or references: up to 9 images, 3 videos, 3 audio clips (2.0); up to 30 images, 10 videos, 10 audio clips (2.5) | Up to 30 reference images, 10 reference videos and 10 audio clips | ≤9 reference images ($0.04 each beyond 5), ≤3 reference videos, 12 files in total; first/last frames and references are mutually exclusive; H3-Max takes no references | Up to 10 reference images, 5 reference videos and 5 reference audio clips (videos and audio ≤15 s each in total); first/last frames cannot be combined with references | Elements built from 2–4 images or a short video; 3.0 Omni takes up to 7 images/elements (4 with a reference video) | Gemini API: up to 3 reference images (Veo 3.1 / Fast; not Lite); first and last frames. Flow Ingredients: Lite and Fast at 8 s, not Quality | First-frame image (input_reference; input images with human faces are rejected); reusable non-human characters, up to two per video (human likeness blocked by default) | Character reference, image-to-video | Subjects reference: up to 7 images or text entries per request (Q3) | Reference images (Fusion reference-to-video) | Casting references; Elements in LTX Studio |
| Intra-segment cuts | Multi-shot prompts within one generation; 2.5 improves shot transitions (vendor claim) | Multi-shot prompts within one generation; 2.5 improves shot transitions (vendor claim) | Native multi-shot modeling (vendor claim); cuts are written in the prompt | Documented multi-shot storyboard format with timestamps, 4–6 s per shot | Multi-Shot / Custom Multi-Shot: up to 6 shots per generation | Timestamp prompting: several shots inside one 8 s generation (Google's Veo 3.1 prompting guide) | The app's storyboard tool closed with the app; the API documents no multi-shot control | Single-shot | Q3: 'intelligent camera switching' within a clip (vendor claim) | Multi-clip switch on text/image-to-video; chained on Canvas | Native multi-shot in one generation (vendor claim); LTX Studio's storyboard layer on top |
| List price | Token-billed (BytePlus, no video input): 2.0 $7.0 per 1M tokens at 480p/720p and $7.7 at 1080p; 2.5 $10.7 and $11.7 (+53% at 480p/720p) | Token-billed (BytePlus, no video input): $10.7 per 1M tokens at 480p/720p and $11.7 at 1080p | API list (international): H3 $0.13/s (2K), $0.08/s (768p); H3-Max $0.05/s (480p), $0.08/s (768p) | International API (Singapore) $0.20/s (1080p), $0.10 (720p), $0.05 (480p); $0.165 / $0.083 / $0.041 in Tokyo, Frankfurt, Virginia and Hong Kong | API list: $0.084/s (720p) and $0.112/s (1080p) without native audio, $0.126 / $0.168 with it, $0.42 at 4K (unit $0.14; packages from $700). App: 6–12 credits per second | Gemini API per second: Lite $0.05 (720p) / $0.08 (1080p); Fast $0.10 / $0.12 / $0.30 (4K); Standard $0.40 (720p–1080p) / $0.60 (4K); blocked videos are not charged. Flow: 10 / 20 / 100 credits per Lite / Fast / Quality generation | API: sora-2 $0.10/s (720p); sora-2-pro $0.30 / $0.50 / $0.70 per s (720p / 1024p / 1080p); Batch half price | App: 60 credits per 5 s ≈ 720/min; Standard 625 credits/mo at $15. API: 12 credits/s at $0.01 per credit ($0.12/s) | API ($0.005 per credit): Q3 reference-to-video $0.035 / $0.06 / $0.075 per s (540p / 720p / 1080p); Q3-pro text/image-to-video $0.045–0.12; off-peak half price. Consumer plans not captured | API: 9 / 12 credits per s at 720p and 18 / 23 at 1080p (without / with audio); $0.09–0.23 per s at the $0.01 pack rate, about a third less on the $100 Essential plan | API per second: Fast $0.09 (720p), $0.13 (1080p), $0.19 (1440p), $0.30 (4K); Pro $0.12 / $0.17 / $0.25 / $0.39. Inside LTX Studio, subscription credits |
| Where to use | BytePlus ModelArk API, Dreamina, and platforms integrating it (SceneMixer, PixVerse, Runway, Luma, invideo, Morph Studio, LTX Studio) | BytePlus ModelArk API, Dreamina, and platforms integrating it (SceneMixer, PixVerse, Runway, Luma, invideo, Morph Studio, Scenes, Pixmax, onsolo, Frameo, Melies) | MiniMax API (platform.minimax.io), Hailuo app, PixVerse, Runway, Luma, Morph Studio, SceneMixer | Model Studio API (workspace endpoint), wan.video, Runway, Morph Studio, SceneMixer | kling.ai app and API, Runway, Luma, LTX Studio, Morph Studio, invideo | Flow, Gemini API, Vertex AI, Runway, Luma, LTX Studio, invideo | OpenAI API until 2026-09-24; Sora app and web discontinued 2026-04-26 | Runway web, iOS / Android apps, API | vidu.com, API (platform.vidu.com) | PixVerse web/iOS/Android/API/CLI | LTX Studio, LTX API, open weights |
All models
ByteDance (BytePlus ModelArk, Dreamina) · Video model
Seedance 2.0 / 2.5
ByteDance's video model, available via BytePlus ModelArk and Dreamina; Seedance 2.5 launched on 2026-07-31 with up to 30 s per generation.
Token-billed (BytePlus, no video input): 2.0 $7.0 per 1M tokens Profile →
ByteDance (BytePlus ModelArk, Dreamina) · Video model
Seedance 2.5
ByteDance's Seedance 2.5 video model: 4–30 s per generation with multi-round extension, multimodal references and native audio; launched 2026-07-31.
Token-billed (BytePlus, no video input): $10.7 per 1M tokens at Profile →
MiniMax · Video model
MiniMax Hailuo H3
MiniMax's video model: multiple references, reference audio, native speech and intra-segment cuts — a common drama-segment engine.
API list (international): H3 $0.13/s (2K), $0.08/s (768p); H3-MaProfile →
Alibaba Cloud (Model Studio) · Video model
Wan 3.0 / 3.0 Prime
Alibaba's Wan 3.0 video model: up to 30 s per segment, native audio, reference images, videos and audio with no separate charge on the price list; Prime is the high-speed tier.
International API (Singapore) $0.20/s (1080p), $0.10 (720p), $0.Profile →
Kuaishou · Video model
Kling 3.0 / Omni
Kuaishou's Kling video model, known for motion and physics; VIDEO 3.0 adds 3–15 s clips, native audio in five languages and Multi-Shot. Also offered inside Runway, Luma and LTX Studio.
API list: $0.084/s (720p) and $0.112/s (1080p) without native auProfile →
Google DeepMind · Video model
Veo 3.1
Google's video model with native audio and dialogue and strong realism, in Quality, Fast and Lite tiers; via Flow, Gemini API/Vertex, Runway, Luma, LTX Studio and invideo.
Gemini API per second: Lite $0.05 (720p) / $0.08 (1080p); Fast $Profile →
OpenAI · Video model
Sora
OpenAI's video model (sora-2 / sora-2-pro) with audio. The Sora app and web closed on 2026-04-26; OpenAI removes the models and the Videos API on 2026-09-24, with no replacement named.
API: sora-2 $0.10/s (720p); sora-2-pro $0.30 / $0.50 / $0.70 perProfile →
Runway · Video model
Runway Gen-4.5 / Aleph
Runway's in-house generation (Gen-4.5) and editing (Aleph) models.
App: 60 credits per 5 s ≈ 720/min; Standard 625 credits/mo at $1Profile →
ShengShu Technology · Video model
Vidu
Shengshu's video model, built around reference-to-video and multi-subject consistency; Vidu Q3 adds native speech and music and clips of up to 16 s.
API ($0.005 per credit): Q3 reference-to-video $0.035 / $0.06 / Profile →
AIsphere · Video model
PixVerse V6
PixVerse's flagship in-house model; the platform also offers Seedance 2.5 and MiniMax H3.
API: 9 / 12 credits per s at 720p and 18 / 23 at 1080p (without Profile →
Lightricks · Video model
LTX-2.5
Lightricks' in-house video model family powering LTX Studio; the current release is LTX-2.5, with open weights and an API.
API per second: Fast $0.09 (720p), $0.13 (1080p), $0.19 (1440p),Profile →
ByteDance (BytePlus ModelArk) · Image model
SeedDream 5.0
ByteDance's image model (Pro / Lite) with 2K output and strong understanding of Chinese costume terms — a common choice for character sheets and scene plates.
Per image (BytePlus): Pro 2K $0.09 ($0.045 at ≤1.5K), Lite $0.03Profile →
Alibaba Cloud (Model Studio) · Image model
Qwen-Image 3.0 / 3.0 Pro
Alibaba's Qwen-Image 3.0 family: text-to-image and image editing in one model, pixel area up to 2048×2048.
International: 3.0 $0.03/image; Pro 1K $0.04, 2K $0.075; input iProfile →
Google · Image model
Gemini 3 Pro Image / 3.1 Flash Image
Google's image models (Nano Banana family), strong on English prompts and modern settings.
Pro ≈ $0.134/image; Flash ≈ $0.067 (1K)Profile →