# Frames by Runway Guide to Gen-4 Image Prompts and API Access > Published 2026-04-06, updated 2026-09-06 · https://www.promptzone.com/zuzanna_choi/frames-ai-launches-for-video-generation-45ef --- title: Frames by Runway Guide to Gen-4 Image Prompts and API Access published: true description: Understand Runway Frames, its transition to Gen-4 Image, and a practical workflow for creating and reviewing visual concepts. tags: ai,imagegeneration --- Frames is Runway's model for still images with control over visual style. Runway now identifies Frames as Gen-4 Images, accessed through its hosted creative tools and developer API. [Announcement][launch] For practical work, use the current Gen-4 Image documentation while keeping the historical Frames release distinct from Runway's video models. ## What are the key facts about Runway Frames? | Field | Verified detail | | --- | --- | | Developer | Runway. [Announcement][launch] | | Released | Announced November 25, 2024; the changelog records availability for Unlimited and Enterprise plans on January 17, 2025. [Announcement][launch], [changelog][changelog] | | Type | Hosted image generation; the announcement now labels Frames as Gen-4 Images. [Announcement][launch] | | Size or parameters | Not published in the announcement or model API reference. [Announcement][launch], [API reference][api] | | License and access | Proprietary hosted access; no open weights are supplied through the documented access paths. [Announcement][launch], [API guide][apiguide] | | Where it runs | Runway's service, reached through the browser interface or HTTP API. [Creation guide][create], [API guide][apiguide] | ## How can Frames help you plan a consistent visual style? Runway introduced Frames around stylistic consistency across variations. Its examples include photography, illustration, collage, and still life. [Announcement][launch] A useful application is planning the look of a project before producing motion. Create a set of still-image candidates that share the same subject, palette, and lighting intention, then select the visual direction you want to develop. Write the art brief before the prompt. Specify what the viewer should notice first, what should remain in the background, and which features need to remain consistent across the set. For a small editorial series, that brief might require a centered ceramic object, a warm background, and shadows falling in the same direction. These are example design choices, not special Frames commands. Runway's current prompting guide recommends descriptive language about subject, scene, composition, lighting, and style. Full sentences can make the relationships between those elements clearer. [Prompting guide][prompting] Try an original description such as: “A blue ceramic teapot on a pale wooden table. Soft window light enters from the left. The camera looks straight across the table, with empty space above the teapot.” ## What are the limits of Frames and Gen-4 Image? Frames produces still imagery. Creating motion requires a separate video-generation operation; a stylistically consistent image series does not by itself demonstrate temporal consistency. [Announcement][launch], [API reference][api] The current Gen-4 Image guide says negative prompting is unsupported. Describe a visible desired state, such as an empty tabletop, instead of relying on a list of objects that must be excluded. [Prompting guide][prompting] Treat reference-based consistency as something to inspect. Check the object outline, material, and colors across candidate images rather than assuming every variation preserves all the details in your brief. Runway documents image generation settings and credit usage in its creation guide. Confirm the selected model, output settings, and displayed cost before requesting a batch. [Creation guide][create] No local GPU requirement applies to calling a hosted endpoint. If your goal is to retain and run model weights yourself, evaluate a downloadable model whose card documents local inference. [API guide][apiguide], [FLUX card][flux] ## How do you use Gen-4 Image on the web and through the API? ### Generate an initial image in the web interface Open Runway's dashboard and select **Generate Image**. In the model selector, open the **Runway** group and choose **Gen-4 Image**, then enter the prompt and adjust the image settings. [Creation guide][create] Start with one clear subject and the output shape your layout requires. Inspect the first result at its intended size before spending time refining details that may disappear in the final placement. Use the [PromptZone prompt collection](/ai-prompts) to develop variations in subject and composition. Keep the structure of your own brief consistent so you can compare those variations deliberately. ### Call the current image API Create a Runway Dev account and configure an API key. The official API guide documents the authorization header, API version header, and asynchronous task workflow. [API guide][apiguide] This example requests Gen-4 Image, the current image access path associated with Frames. It uses the documented text-to-image endpoint and a supported square output setting. [API reference][api] ```bash curl https://api.dev.runwayml.com/v1/text_to_image \ -H "Authorization: Bearer $RUNWAYML_API_SECRET" \ -H "X-Runway-Version: 2024-11-06" \ -H "Content-Type: application/json" \ -d '{ "model": "gen4_image", "promptText": "A blue ceramic teapot on pale wood, soft window light from the left, centered composition.", "ratio": "1024:1024" }' ``` Read the returned task identifier, then retrieve its status with `GET /v1/tasks/{id}`. Wait for success before using the output; handle a failed task separately from a completed image. [API guide][apiguide] Store the model identifier, prompt, task identifier, and selected result together. This is especially helpful when an editor asks which settings produced an approved composition. Before automating a series, review one complete request and output in the developer environment. Confirm that your application handles both the waiting state and a task failure without showing an unfinished asset as ready. ### Review a coherent set Arrange the candidate images in their intended order. Ask whether the light, color, and framing tell the same visual story, then identify the single image that breaks that pattern most clearly. Revise that image against the approved brief. Keep rejected candidates in your working notes long enough to explain the decision, then archive the final selection with its generation record. ## How does Runway Frames compare with FLUX.1 dev? | Option | Practical distinction | | --- | --- | | Runway Frames / Gen-4 Image | Hosted image generation through Runway's creative tools and API. [Creation guide][create], [API reference][api] | | FLUX.1 [dev] | Downloadable image model with documented local inference and a model-specific non-commercial license. [FLUX card][flux] | The [ComfyUI pillar](/tomas_novak/comfyui-2026-the-complete-guide-to-power-user-ai-image-generation-1g17) is relevant when evaluating a local image workflow. Compare the effort required to operate that workflow with the hosted route you would otherwise use. ## What else should you know about Runway Frames? ### Is Frames a video-generation model? Runway introduced Frames as an image model. Video generation is a separate operation, even when the resulting still image becomes the starting point for a clip. [Announcement][launch], [API reference][api] ### Where can I find Frames now? The official announcement says Frames is now Gen-4 Images. Follow Runway's current Gen-4 Image creation guide or developer documentation for access. [Announcement][launch], [creation guide][create] ### Can I download Frames for local inference? Runway documents hosted access and does not provide open weights through those access paths. A local workflow therefore needs a different model with explicitly downloadable weights. [API guide][apiguide] ### Does the API immediately return the finished image? Runway's `POST /v1/text_to_image` endpoint returns a task ID. Poll `GET /v1/tasks/{id}` and use the output only after its status becomes `SUCCEEDED`; handle `FAILED` as an unsuccessful generation. [API guide][apiguide] ## Sources - [Runway Frames announcement][launch] - [Runway product changelog][changelog] - [Creating with Gen-4 Image][create] - [Gen-4 Image prompting guide][prompting] - [Runway API reference][api] - [Runway API getting-started guide][apiguide] - [FLUX.1 dev model card][flux] [launch]: https://runway.com/research/introducing-frames [changelog]: https://runway.com/changelog [create]: https://help.runwayml.com/hc/en-us/articles/37053594806419-Creating-with-Gen-4-Image [prompting]: https://help.runwayml.com/hc/en-us/articles/35694045317139-Gen-4-Image-Prompting-Guide [api]: https://docs.dev.runwayml.com/api.md [apiguide]: https://docs.dev.runwayml.com/guides/using-the-api.md [flux]: https://huggingface.co/black-forest-labs/FLUX.1-dev ## Related guides on PromptZone - [AI Model Releases Timeline](/ai-model-releases)