24 things to ask once eroq is connected to Claude or ChatGPT

Once eroq is connected to Claude or ChatGPT, the question is what to say. An assistant turns your sentence into a tool call — generate_image, generate_video, generate_speech — and the sentence decides which tool, which engine, which format and how much it costs. These twenty-four asks are the ones that reliably produce the thing you meant in one turn, grouped by what comes back, each with the call it fires and the price.

Prices are the studio's, from the billing function: 10 credits for a Krea 2 image, 45 for five seconds on the default video engine, 3 per 100 characters of speech. Failed renders refund themselves.

Pictures

1. « Make a 16:9 image of a neon alley in the rain, wet asphalt, one figure under an umbrella, 35mm. » — generate_image, aspect: "16:9". Naming the aspect is the single most useful habit; without it you get a square.
2. « A 3:4 portrait of a woman in her thirties, short dark hair, navy blazer, studio grey background, soft key light. » — generate_image. Concrete nouns; no adjectives the engine cannot draw.
3. « Same scene at night. » — a second generate_image; the assistant rewrites the previous prompt. Say « same scene » and it keeps the composition words.
4. « Enhance this first: "cozy cabin". Then render it in 4:3. » — enhance_prompt (free) then generate_image. Ask to see the enhanced prompt if you want to edit it.
5. « Three variations of that, different palettes. » — three generate_image calls, 30 credits. The assistant varies the color words.
6. « A 9:16 product shot of a matte black bottle on black marble, one beam of light. » — generate_image. Vertical for stories.
7. « A 1:1 avatar for the persona "Rook", a grizzled ship mechanic, warm hangar light. » — generate_image. Square is the default; say it anyway.
8. « What did that cost, and what's my balance? » — get_account, free. Ask any time.

Clips

9. « Animate that image — slow push-in, 5 seconds. » — generate_video with the image's link as reference_image_url, then get_video_status. This is how a clip matches the still: the picture becomes the first frame.
10. « A 5-second clip on Kling 3.0: a woman walks toward camera down a graffiti alley, tracking shot, 9:16. » — generate_video, model: "kling-3", 170 credits. Naming the engine changes the look and the price.
11. « 8 seconds on Veo 3.1 Fast, with sound: rain on a tin roof, slow tilt down to a puddle. » — generate_video, model: "veo-3-1-fast", seconds: 8, audio: true, 216 credits. Veo renders in steps of two seconds.
12. « The cheapest 5-second clip of a drone orbit over a lighthouse at dusk. » — generate_video on the default engine, 45 credits. « Cheapest » makes the assistant check list_models.
13. « Same clip, 1080p. » — generate_video with resolution: "1080p" on an engine that serves it; the assistant switches if the current one does not.
14. « Check the render. » — get_video_status, free. If the assistant stopped waiting, this is the nudge.
15. « A 10-second clip on Seedance 2.5 of a detective under a flickering neon sign, slow push-in, no sound. » — generate_video, model: "seedance-2-5", audio: false, 620 credits (a Creator engine; the assistant says so if your plan lacks it).
16. « Which engines can I use on my plan, and what do they cost for 5 seconds? » — list_models, free. The answer is a table.

Voice

17. « Say "You're late." in a low male voice. » — list_voices then generate_speech. The assistant picks from the roster.
18. « Read this paragraph as Aria, warm and slow. » — generate_speech with a named roster voice. Direction words go into the text's punctuation more than the engine's settings; the punctuation guide explains.
19. « Have Mina say "It's still lit." — whispering. » — list_characters then generate_speech with char:<id>; the line comes out in the voice assigned to your character.
20. « A 20-second voice-over for a perfume ad, female, confident, from this script: … » — generate_speech; about 9 credits for 300 characters.

Combos

21. « Storyboard a 3-shot teaser for a horror short, then render shot one as an image. » — the assistant writes three shots in text (free) and calls generate_image once. Render the other two when you like the first.
22. « Make the hero image for a landing page about ceramic kettles, then a 5-second loop of it. » — generate_image then generate_video from it, 55 credits.
23. « Give me a one-line prompt I can paste into the eroq studio for this idea. » — enhance_prompt, free. No render; take the prompt to Cinema where cast, look and continuity live.
24. « Show me the prompt you used. » — no tool. The assistant has it; edit it and ask for the render again.

Phrasing that steers

• Aspect first. « 16:9 », « 9:16 », « 3:4 » — same price, different picture.
• Engine by name for clips: Kling 3.0, Seedance 2.5, Veo 3.1 Fast, Hailuo 3. Otherwise the default, cheapest engine.
• « Animate that » ties the clip to the last image. « A clip of … » starts fresh.
• « Cheapest » / « best » makes the assistant consult list_models instead of guessing.
• « With sound » / « no sound » — sound costs more on some engines; say which you want.
• Characters by name if you keep a cast on eroq; the assistant resolves them.

What an assistant will not do

Render a whole film (that is Cinema), hold a cast across shots, or edit. It renders one thing per call, well, and hands you the link. For anything with continuity, take the prompt to the studio.

FAQ

Do these work the same in Claude and ChatGPT?

Yes — both call the same ten tools on the same server. The phrasing that steers one steers the other.

Why did the assistant render a square?

You did not name an aspect. Add « 16:9 » or « 9:16 » and ask again; the price is the same.

Can I set a budget?

Say it: « stay under 200 credits ». The assistant calls get_account (free) and stops before exceeding it.

Where do the renders go?

Every one is in your eroq Library with its recipe, and the link in the chat is valid thirty minutes.

Connect first on the MCP page, then paste number 1.