PRACTICAL GUIDE · PROMPTTOVISUAL
Reka Rho-1: Image and Video Workflows, Research Preview and Limits
Rho-1 is Reka’s 19B research preview, announced October 5, 2026. It combines understanding, generation and editing within one context. It is not available in our model selector. This independent guide explains the creative workflow and what you still need to verify before planning to use it.
Read Reka’s official research announcement →
What changes for an image-to-video workflow?
Reka describes text, images, video and actions in one model. Its lighthouse demonstration creates an image, locates an object, animates the scene, changes the weather and explains the edit within one conversation. This is a vendor demonstration, not a PromptToVisual test. The announcement does not establish native audio or rap lip-sync.
A creative evaluation plan, not an API tutorial
Write down one scene and one change you want to make. Save the original composition and specify what must remain stable, such as the product shape or camera direction. If access becomes available, assess the image, animation and edited clip separately. Ask whether an edit preserves your intended subject rather than treating a coherent explanation as proof that the pixels are correct.
Limits that matter before promising a deliverable
Reka reports native video rollouts capped at 672×384, structural drift in long sequences, unreliable grounding across video and brittle editing across prompts. Base-model results and distilled/Flash demonstrations are different settings. Published latency is internal evidence, not our benchmark or a production SLA; the comparison pipeline graphic is illustrative.
Can you use it yourself today?
As checked October 6, exact public self-service Rho-1 API access, prices and downloadable weights are NOT_VERIFIED. Research preview does not establish general availability or open weights. Follow the official announcement for access information.
What to confirm before a project depends on it
Obtain the exact access route and account eligibility, then confirm permitted inputs, output dimensions, usage terms, delivery format and total cost. For a deadline-driven job, an announcement or selected demo is insufficient evidence of a repeatable workflow. Avoid promising a client an export or turnaround time before those details are confirmed.
Rho-1 is not RIDM
Reka’s separate Inverse Dynamics Model infers control signals from video and publishes weights under Apache 2.0. That separate release does not grant the same license or weight availability to Rho-1. Robotics examples do not establish a ready-made creative video service.
Choose your next step
For Rho-1, start with Reka’s announcement below. For a present-day single-image task on PromptToVisual, the linked Seedance 2.5 workflow is a separate four-second, 40-paid-credit alternative. It does not reproduce Rho-1’s conversational workflow. No Rho-1 generation, integration or matched comparison was performed for this guide.
Sources and next steps
Sources checked October 6, 2026. Independent documentation review; no integration or model generation test.
Utopai X access guide · Keyframe control guide · FLUX 3 Video guide · Seedance limits · PTV single-first-frame alternative · All guides