Runway GWM Worlds 2 (reported): Sume has no world model
Releasebot reports a GWM Worlds 2 research preview from Runway. Sume has no world model or live scene. Sume offers finished video clips from a prompt.

Sume does not offer a world model, and it has no interactive or real-time scene. If you are looking for something like Runway's GWM Worlds 2, Sume's nearest tool is the Videos API, which returns a finished clip for a prompt. Details about GWM Worlds 2 here are reported by Releasebot's Runway update feed, a third-party aggregator; this post has not checked Runway's own page for them.
What is reported about GWM Worlds 2?
Releasebot lists a GWM Worlds 2 research preview dated 2026-09-03, with 720p, 24 fps and 48 kHz audio. It also lists a Solaris world model entry dated 2026-09-01. Treat these as reported, not confirmed, and read Runway's own announcement before building on them.
What does Sume do instead?
The Videos API at POST /v1/videos is OpenRouter-style and asynchronous. You submit a prompt and a model id, wait for the job and read the clip. Seedance 2.5 takes 4 to 30 seconds and Gemini Omni Flash 1.1 takes 3 to 10 seconds, with a video_url edit mode, per the docs.
| Property | GWM Worlds 2 (reported) | Sume Videos |
|---|---|---|
| Output | Research preview of a world model | A finished MP4 clip |
| Interaction | Not covered by this post | None, submit then read |
| Resolution and rate | 720p, 24 fps, reported | Per model, see the docs |
| Audio | 48 kHz, reported | Per model |
Can I fake a world with several clips?
You can chain clips: take a still from the end of one clip with Video frames and use it as the start image of the next. That gives you a sequence of separate clips. It is not a persistent 3D space, and nothing in Sume keeps scene state between jobs.
When should I use something else?
If you need a navigable scene or a live session, Sume is not the tool. If you need a clip for an ad or a social post, it is.
How do I keep a look across several clips?
Use the same reference images and the same prompt skeleton in each job, and change only the action. Extract the final still of each clip with Video frames and compare it with the next clip's first still. This gives visual continuity between separate clips, which is the most Sume can offer in this area. Sume's docs list the exact limits and error codes for each call, so read the page for the endpoint you use before you build on it, and keep a short note of which limits applied to your run, so a later change is easy to spot. If a call is refused, the error code names the cause, and fixing that one input is usually enough to retry safely with the same Idempotency-Key.
Sources
Related posts
More in Comparisons
- Runway voice dubbing API: 1 credit per 2 seconds vs Sume's dub steps
Runway prices voice dubbing at 1 credit per 2 seconds of audio: a 10-minute video is 300 credits. Sume has no dubbing endpoint, so it takes three steps.
- Medical transcription API: Scribe v2 Medical GA, and what Sume STT is
ElevenLabs reports Scribe v2 Medical as GA on 2026-09-11. Sume's STT is general speech-to-text with word timings, not a medical product. Where that matters.
- Seedance 2.0 4K on Hedra: what Sume's seedance-2 lists
Hedra lists Seedance 2.0 at 4K for about $9.07 a minute. Sume's docs list seedance-2 at 480p to 1080p; here is where 4K does and does not work on Sume.
- Segmind API vs Sume: PixelFlow workflows or saved Formats
Segmind turns visual PixelFlow graphs into API endpoints. Sume saves an agent thread as a Format you call over the API. How the two reuse recipes.
Written by Sume