A creative platform for generated video, image references, footage edits, character performances, Agent, and connected production workflows.

Runway is a creative platform for generating images and video, transforming existing footage, creating character performances, and connecting production steps. Its workspace includes individual model controls, task-specific Apps, a conversational Agent, a timeline for assembling material, and node-based Workflows. It is useful for developing shots and sequences when part of the production involves generated or transformed media.
Filmmakers can explore a scene before a shoot, animators can develop a character performance, and marketing teams can create variations around a product concept. Runway also provides audio and image tools that support those tasks. The relevant choice is the workflow: generating a new shot, editing a supplied shot, transferring a performance, or combining several stages. Each has different inputs, capabilities, and credit requirements.
Gen-4.5 supports text-to-video and image-to-video generation. The documented model settings include short clip durations and specific output formats, with access on qualifying paid plans. Other available Runway and partner models have their own settings, so frame inputs, sound, duration, and resolution should be checked for the model selected rather than inferred from the platform as a whole.
A text-led shot needs a description of both the visible scene and its motion. An image-led shot already has a starting composition, so the motion request becomes especially important. For example, an approved still of a fictional shop can become a short shot with a slow camera approach and a small movement in the window display. That is a more focused instruction than asking the same clip to establish the shop, introduce a character, and complete an entire story.
Review the shot over time, not just as a thumbnail. A strong opening frame can be followed by an altered object, an unexpected camera movement, or a gesture that no longer supports the scene. Keep the usable portion only if it contributes to the sequence. Splitting a complicated scene into several deliberate shots can give you more control over the story than repeatedly adding instructions to one generation.
Gen-4 Image References uses supplied images to guide new images through characteristics such as a character, object, environment, or style. The documented workflow supports up to three active references, saved reference names, and sketch annotations. A temporary reference in an active session differs from one saved for later use.
Build the starting image around the shot you need. If the character must be recognizable in three locations, establish the important appearance details in a clear reference. If an object has to sit on a particular side of the frame, a sketch can communicate that placement more clearly than a long verbal description. Assign different references clear roles instead of expecting the model to preserve every element from each source.
Check the reference image before sending it to video. Clothing, props, proportions, and the camera angle should already support the intended movement. An uncertain still becomes an uncertain foundation for motion. Save approved references with meaningful names and keep experiments separate so the next shot uses the intended version. References help guide continuity, while side-by-side review determines whether the images actually form a coherent set.
Edit Studio, powered by Aleph 2.0, transforms traditional or generated footage through prompt-based edits. Documented applications include replacing objects or characters, changing backgrounds, adding effects, adjusting lighting, and restyling a scene. Its supported sequence length and input requirements differ from a model that creates a new clip from scratch.
This is useful when the original movement is already appropriate but a visual element needs to change. A short concept shot might have the right camera motion and performance while requiring a different environment. Preserve the source and make the requested transformation specific. Compare the resulting clip with the source around the moments where the subject crosses a background edge or interacts with an object.
The Aleph prompting guide recommends clear actions and descriptions of the desired transformation. A request such as changing the coat's color is easier to assess than a general request to improve the shot. Inspect unaffected details too. When a product must remain exact, compare its shape, label, and reflections throughout the clip before approving a generated edit.
Act-Two uses a driving performance video with a character image or video to transfer a performance. It is designed around movement, speech, and expression, with input-dependent controls such as gesture control for a character image. This provides a route for someone who wants to act a scene rather than describe every expression in a prompt.
Record the performance around the character's role in the scene. A brief spoken reaction with a clear expression is easier to judge than a recording containing unrelated movements. Keep the performer and the intended gesture visible, and use a character reference whose framing makes sense for that performance. A close portrait and a broad body gesture express different visual requirements.
Review the timing of the face, hands, speech, and body together. A performance can communicate the intended emotion while a particular movement still needs another take. Compare the output with the story's purpose, not merely with the input recording. For a dialogue sequence, retain room for the other character's response so the performance does not feel like a continuous isolated monologue.
Character Script to Video animates a character image using a typed script with a selected voice or a supplied audio file. It offers preset or uploaded characters and supported voice choices. The guide explicitly states that this App uses credits and does not support infinite generations in Explore Mode.
Use it when the dialogue is already written and the character needs to deliver that line. Keep the image clearly framed and the audio understandable. A short instruction, introduction, or story beat is a useful starting point. Listen to the words and inspect the performance together; the appropriate voice tone for a narrator may be inappropriate for a character reacting to an event.
Choose this route or a performance-driven route according to the creative requirement. A supplied acted take gives you a specific expression and timing to transfer. A written script with a chosen voice starts from the words and voice. Neither choice should be treated as interchangeable simply because both can produce a speaking character. The shot's emotional intention determines which input is more useful.
Runway Agent is a conversational creation interface that can plan and generate images, video, audio, and music. It can work on multi-shot projects and assemble clips and audio in Final Cut, where you can also trim, split, and arrange generated or uploaded media. Current documentation identifies Agent as the successor to Chat Mode.
Give Agent a bounded production brief. For a short announcement, describe the audience, the message, the approximate structure, and any supplied material that must be used. Review the proposed visual direction before asking for several completed shots. A conversation can help develop the idea, but approval still belongs at the important creative decisions: subject appearance, shot order, and the final spoken words.
Use Final Cut to inspect the sequence as a sequence. A set of individually convincing shots may repeat the same information or change visual direction halfway through. Remove a redundant clip, adjust the order, and make sure the ending serves the brief. Keeping the generated media accessible as separate assets also lets you refine one shot without discarding everything else.
Apps provide focused workflows using Runway's generative tools. The documented collection includes image, video, and audio tasks such as product reshoots, story panels, background changes, motion sketches, stitching clips, and sound effects. An App packages a particular task rather than exposing every possible model setting in one interface.
Start with the task you can name clearly. A product reshoot explores a supplied product image in a changed setting; a motion sketch expresses movement through annotations; stitching combines existing clips. Those are different operations even if they contribute to the same campaign. Check the input requirements before preparing material so the asset is suitable for the chosen route.
For a small team, a focused App can make a recurring step easier to explain. Keep an example input and an approved output with the project notes. That makes it possible to compare the next result against an actual requirement. If the task grows more complex or needs branches, a custom Workflow may express the process more clearly than a chain of unrelated App runs.
Workflows use a node canvas to connect text, media, generation, and utility steps. Links pass compatible outputs into later inputs. You can create branches to compare different models or prompts and build reusable production patterns. This is useful when the same multi-stage process needs to be repeated with new material.
A simple example is an image prompt feeding an image model, with the selected image feeding a video model and a separate motion prompt. Keep the still-image instruction distinct from the motion instruction. That makes the cause of a problem easier to identify: the composition may need revision while the motion request is already suitable, or the reverse.
The workflow tutorial documents running individual nodes as well as the full graph and reviewing node execution history. Reuse a strong earlier output instead of automatically rerunning every stage. Clearly label the branches and keep the graph small enough to understand. Automation is most useful after you know which stages deserve a human approval point.
Runway provides audio creation through supported tools and models, including speech generation. Agent and compatible Workflows can also participate in audio production. Some video models generate native audio; others need a separate soundtrack. Check the selected model rather than assuming every output contains speech or ambience.
Write narration to match the edit and confirm pronunciation before using it across several shots. If a generated clip already includes sound, listen for details that contradict the visual action. A quiet interior does not necessarily need a busy soundtrack. A product demonstration may need a clear operating sound rather than music throughout.
Make the final mix a separate review. A voice that works alone can become difficult to follow beneath music, and an effect may land on the wrong action after trimming the video. Keep narration, ambience, and music as distinct creative decisions wherever the workflow allows. That gives you more flexibility when the pacing or message changes.
Imagine a team making a concept film for an invented desk lamp. Prepare an accurate reference image of the intended design and a three-shot outline: the lamp in a workspace, a close detail, and a closing composition with room for text. Use image references to approve each still before animation. Describe a single useful movement for each shot rather than asking for the complete film in one generation.
If a supplied shot has the right movement but the wrong environment, explore a targeted Edit Studio change. Review the lamp's geometry and reflections in every frame that will be used. Create a brief narration separately and assemble the accepted clips in Final Cut or the finishing editor you prefer. Keep exact title text editable in the production layout.
Save the reference set, chosen outputs, prompts, and relevant Workflow together. If the second shot needs a change, return to that stage rather than rebuilding the whole sequence. The resulting process uses Runway for visual development, motion, footage transformation, and assembly while preserving clear review points.
The pricing page offers free entry with a limited one-time credit allocation and paid plans with different credits, storage, concurrency, and features. Model generation cost depends on the selected operation and settings. Explore Mode applies to eligible models and workflows under qualifying access, while some Apps explicitly require credits even when the account has broader unlimited-generation features.
Budget for revisions rather than only the duration of the final film. A usable shot may require several candidates, and transformed footage needs review for temporal artifacts, object changes, and continuity. Higher-resolution processing does not make every generated detail correct. Enterprise, team, and API requirements should be checked separately when the workflow involves organization-wide use or integration into another application.
Yes. Edit Studio transforms supplied footage, while the timeline can assemble and trim uploaded material. Choose the route that matches a visual transformation or an ordinary editorial change.
No. Free access is limited, and model or App requirements vary. Check the selected tool and current plan comparison before budgeting a project.
No. They connect and repeat production steps. Review the important outputs and use execution history to keep a suitable version before proceeding.
A reference guides generation. Inspect the product throughout the resulting shot, especially small markings, geometry, and interactions that must remain accurate.