VisionStory’s native character model turns one image plus text or audio into an identity-consistent performance with natural expression, gestures, and lip-sync.
- Text or recorded audio
- 720p, 1080p, or 2K
- Up to 10 minutes
VisionStory model catalog
Choose the model for the output you need.
VisionStory combines our own V-Character model with selected partner models for talking avatars, cinematic video, image generation, and multimodal creative work.
Use this page to compare strengths. Use the model-list endpoints as the source of truth before sending a request.
Public API models
Three model families to start with.
These models map directly to public VisionStory API workflows. Availability and accepted parameters remain discoverable at runtime.
The latest Seedance family model for prompt-led and image-led video generation, longer single clips, coherent motion, and native audio.
- Text-to-video
- Image-to-video
- 4–30 second clips
A balanced image model for fast generation, high-resolution output, reliable text rendering, and conversational edits guided by reference images.
- Prompt-to-image
- Reference-guided editing
- 1K and 2K output
Model selection guide
Start from the job, then choose the ID.
The best model is the smallest reliable choice for your quality, control, duration, and cost requirements.
Talking avatar video
Recommended
vs_character_v4Use V-Character 4.0 for the best overall motion and stability. Use VS Talk 1 for speech-first clips with focused mouth movement and 480p support.
Frontier AI video
Recommended
seedance-2.5Choose 2.5 for the newest generation and longer clips; 2.0 for multimodal references; fast or mini when latency and credit use matter most.
AI image generation
Recommended
nano-banana-2Choose Nano Banana 2 for balanced generation and editing, the base model for economy, or Pro when maximum quality and 2K output matter.
Text to speech
Recommended
Automatic selectionThe TTS endpoint selects the synthesis strategy for the chosen public or cloned voice and input language; you do not pass a public model ID.
VisionStory workspace
A broader model catalog for creative work.
These model profiles are available in the VisionStory creative workspace. A profile name is not automatically a public API model ID.
Developer rule
Only send model IDs returned by the three public model-list endpoints. Do not derive an API ID from a workspace model name.
AI videoby Alibaba
Wan 3.0
A multimodal video model for longer 1080p stories, reference control, and synchronized audiovisual generation.
Open model profile
Language & multimodalby Moonshot AI
Kimi K3
A frontier-scale open model for long-context reasoning, coding, and multimodal work inside the VisionStory model catalog.
Open model profile
AI videoby MiniMax
MiniMax H3
An open-weight multimodal video model built for 2K clips, synchronized stereo audio, and reference-rich creative workflows.
Open model profile
AI videoby Kuaishou
Kling 3.0
A cinematic video model known for physically believable motion, camera control, multi-shot storytelling, and native audio.
Open model profileDiscover at runtime
Let the API tell you what is available.
Model catalogs evolve without changing the endpoint contract. Fetch the catalog when your application starts, validate user choices against the response, and store the returned stable ID with each generation job.
- Accepted aspect ratios, resolutions, and durations
- Reference-media limits and model defaults
- Image credit cost and current availability
Discover available models
# Talking-avatar rendering models
curl -s -H "X-API-Key: $VISIONSTORY_API_KEY" \
https://openapi.visionstory.ai/api/v1/models
# Frontier AI video models and parameter limits
curl -s -H "X-API-Key: $VISIONSTORY_API_KEY" \
https://openapi.visionstory.ai/api/v1/ai_video/models
# Image models, resolutions, reference limits, and credits
curl -s -H "X-API-Key: $VISIONSTORY_API_KEY" \
https://openapi.visionstory.ai/api/v1/image/models