Mux Robots playground vs Sume: test a video job before paying

Mux launched a Robots playground on October 1, 2026, and it is not a free sandbox: jobs run on the real Robots API and consume the same units as API jobs. Sume's closest equivalents are three unbilled checks: the Timeline 1.0 plan, the probe and stills half of video inspect, and the video filter contract check.

Mux's facts are from its playground announcement and the Mux changelog; Sume's are from the docs.

What is the Mux Robots playground?

Mux describes it as a place in the dashboard, under Robots then Playground, to run Robots workflows against your own video and see what they return before you write code. You can tune workflow parameters and output steering, for example the selection_strategy of Find Key Moments between product_moments and story_beats.

The cost note matters. Mux says jobs in the playground consume the same units as API jobs, the page shows a running total in units and dollars, results are stored for 30 days, and nothing is a simulation. The first 100,000 units each month are free per the pricing page, so small tests may cost nothing.

What can you run on Sume without paying?

Sume does not have a playground UI for media tools, but three API calls are documented as unbilled or non-reserving.

• POST /v1/timeline-1.0/plan compiles a Timeline 1.0 document and returns duration_seconds, segment_count, billable_minutes, estimated_cost_usd_micros and a filtergraph summary. It creates no job, reserves no credits and downloads no media. It cannot predict short-source pad or loop warnings.
• POST /v1/video-inspect returns probe facts and sampled stills without billing. Only transcribe: true reserves, at $0.01 per audio minute.
• POST /v1/video-filter/check is an unbilled contract check for a filter program before the $0.02 encode job.
• POST /v1/video-frames extracts exact-time stills and is documented as unbilled.

curl -X POST https://api.sume.com/v1/timeline-1.0/plan \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "audio": { "mode": "silence", "duration_seconds": 12 },
    "video": [
      { "source_url": "https://media.sume.com/artifacts/artf_demo/intro.mp4",
        "start": 0, "duration": 12 }
    ]
  }'

What do those checks not tell you?

They validate structure and cost, not taste. Mux's playground shows what an AI workflow returned; Sume's plan shows what the compiler will do and what it will cost. Neither predicts whether a result looks good to your audience.

Sume's plan also checks that URLs are on the Sume media host, so the source must already be imported (POST /v1/media-imports). The plan docs say a short source that will be padded or looped shows up only as a warning in the real render, so budget for one test render on your actual footage.

Which workflow should you start with?

If you are evaluating Mux's AI workflows on assets already in Mux, use the playground and watch the unit counter. If you are evaluating a render pipeline on Sume, call the plan first, then run one real render on a short clip, and only then scale up with an Idempotency-Key on every submit so a retry does not bill twice.

What does a safe first test look like?

Start with three steps that cost nothing on Sume. Import a short clip, call video inspect with frames: false to read the probe, including whether the clip has audio, and call the Timeline plan with the segments you intend. Only then submit one real render.

Reading probe.has_audio first avoids a billed mistake: transcribe: true on a silent clip fails with inspect_source_has_no_audio. On the Mux side, the playground's unit counter performs the same job. Set a ceiling in your own code before moving to the API.

• Sume plan: no job, no reservation.
• Sume inspect probe: unbilled.
• Mux playground: same units as the API, first 100,000 a month free.