Soutine AI
Official website: Soutine AI
Soutine AI is a browser-based workspace for creating images and videos with multiple AI models. Start from a written description, a reference image, or a still frame, then choose the workflow and compatible model for the task.
The website brings generation tools, prompt libraries, credits, and creation history together. This Hub card introduces that workspace and points to its live resources; it does not distribute a Soutine model.
Product description
Soutine supports several starting points: creating a new image from text, editing or restyling an existing image, describing a video shot, and using a still image as the starting point for motion. The selected workflow determines which models, inputs, and settings are available.
The website also includes enhancement tools and examples for exploring a direction before generating. Model availability and limits can change, so use the controls displayed in the current generator rather than assuming every workflow supports the same options.
Main workflows
| Workflow | Starting point | Typical purpose |
|---|---|---|
| Text to Image | A written visual description | Explore a concept, composition, or visual style. |
| Image to Image | Reference images and editing instructions | Restyle an image or change selected aspects of a scene. |
| Text to Video | A description of a scene, action, and camera | Develop a short motion concept or visual sequence. |
| Image to Video | A still image | Explore movement from an existing frame. |
| Enhancement tools | Existing media | Improve or upscale media using the available tool settings. |
These are use cases, not guarantees of visual accuracy or production readiness.
Explore the prompt libraries
| Library | What to explore |
|---|---|
| Nano Banana prompts | Image ideas organized around portraits, photo edits, product staging, characters, and visual styles. |
| Seedance video prompts | Prompt cards with video previews, camera directions, motion, and frame-ratio choices. |
| Qwen Image prompts | Examples for portraits, products, transparent assets, editing, and assigning roles to multiple reference images. |
Use the libraries to find a starting point, inspect examples, and adapt the wording to your own subject. Preview images and clips show examples rather than a promise that a new generation will reproduce them.
Intended use
The workspace can support creators looking for visual ideas, marketers preparing draft campaign assets, designers exploring styles, and teams building image or video concepts. A useful sequence is to settle on a still-image direction first, then explore motion when the chosen workflow supports it.
For a client deliverable or public campaign, review the output, permissions for uploaded material, and the service's current commercial-use terms before use.
How to use Soutine
- Choose the task. Decide whether you need a new image, a reference-guided edit, a video from text, or motion from a still image.
- Describe or upload. Write the desired result and add reference images when the selected model supports them.
- Choose a compatible model. Review the input limits and settings shown for that model and workflow.
- Refine the instructions. Specify the subject, composition, lighting, style, or motion that matters. A library prompt can help you start.
- Review the settings and credit quote. Sign in and confirm generation when ready. Copying a prompt does not itself produce a result.
- Inspect the output. Find the task in the generator or My Creations, then download or refine it. If a task remains pending, check its status before submitting a duplicate.
An illustrative project
For a fictional coffee campaign, first explore a still image of a cup on a simple studio background. Review the product appearance and layout. If you want motion, take an appropriate still into an image-to-video workflow and describe a restrained camera move. This is a suggested workflow, not a tested result or a promise of automatic continuity between models.
Limitations and review
Different models can interpret the same prompt differently. Generated text, anatomy, object details, identity, and motion may need correction. Reference images guide a result but do not guarantee that every feature stays unchanged.
Generation time and cost vary with the task, model, media, and settings. The website's current quote and terms govern use; this card makes no speed or quality benchmark claims.
Check output visibility, watermark settings, and commercial-use eligibility in your account before publishing. Free prompt browsing does not imply that every generated output is private, watermark-free, or licensed for commercial use.
Frequently asked questions
Where does generation happen?
On Soutine AI, using its workspace and credits. This Hugging Face repository has no inference endpoint.
Which model should I choose?
Start with the compatible options shown for your task. Compare their current settings and limits rather than assuming a single model is best for every image or video.
Can I use prompts from the libraries?
You can browse and adapt the examples on their landing pages. Check whether a prompt expects reference images, a particular frame ratio, or a specific generation mode.
Are prompt browsing and generation the same service action?
No. Browsing gives you wording and examples; generating submits a task and uses the service's current credit requirements.
Does this card contain downloadable model weights?
No. It is a reference and navigation card for the live product.
Can I use an output commercially?
Check the current plan or credit-pack terms and the rights to your inputs. Public access to this card does not grant a license to every generated asset.
Get started
Open Soutine AI to choose a workflow, or begin with one of the three prompt libraries above.
About this repository
This is a README-only resource card linking to the Soutine website and prompt libraries. It contains no model weights, training code, or hosted inference endpoint. Generation takes place on Soutine and is subject to the service's current credit requirements and terms.