Grok Imagine Video 720p vs Veo 4K

Grok Imagine Video starts at 24 credits, text-to-video, native audio, 720p, 6 / 10 / 15 / 20 / 30 seconds. Veo 3.1 starts at 64 credits, text-to-video, native audio, 4K, 4 / 6 / 8 seconds. Length versus resolution and dialogue. Grok is the cheap long hold. Veo is the short 4K cinematic. Do not prompt Veo as if 30 seconds were a sentence. Do not prompt Grok as if 4K were a quality token.

The Grok card: audio true, category text-to-video, no image required, 720p, 24 credits. The duration list is the point of the row: 6s, then 10, 15, 20, 30. There is no 4s. There is no 8s. There is no 1080p token. Do not invent one.

The Veo card: audio true, text-to-video, 4 / 6 / 8 seconds, 4K, 64 credits. Same native-audio flag. Different ceiling. Different sticker. The model does not change because you launched it from Flow. Google Flow is a Veo surface, not a studio. This page is the row.

Twenty-four credits is the catalog sticker. Sixty-four credits is the catalog sticker. Both are the shortest job, so do not divide 24 by 30 and 64 by 8 to invent a per-second race. The catalog lists the real rates: 4 to 6 credits a second on Grok, 16 silent or 32 with sound on Veo. The race is which constraint is binding: time, or 4K dialogue.

Grok: from 24 credits, up to 30 seconds, 720p

When Grok is the pick:

The brief is a single take that has to last past eight seconds. A 15-second hold, a 20-second walk, a 30-second product sentence that must not cut.
The feed is 720p, or you will accept 720p as the master. Many social cuts never need 4K.
You are drafting camera language before you spend 64. Twenty-four credits is how you find out whether the sentence wants to be a clip.
Native audio is required, and Veo’s eight seconds will not hold the line.

When Grok is the wrong pick:

The insertion order is 4K.
The shot is a four-second or eight-second cinematic you already know is Veo-shaped: physics, always-on dialogue, a plate you will crop from 4K.
You have a still that must remain frame one. Grok Imagine Video is T2V in this catalog. A still contract is I2V on another row.
You need 1080p identity on a pack. 720p will soften type. Lock the pack on an I2V row at 1080p, not on a 24-credit sentence.

Native audio is on. Write whether anyone speaks. A 30-second surprise monologue is a long error if the brief was foley. A 30-second line that actually needs to be said is the reason the row exists. Spell names. Then caption them anyway; native audio is not typesetting.

Sora 2 text-to-video was 32 credits, 720p, 4 / 8 / 12 / 16 / 20 seconds, native audio. The Sora API shuts 24 September 2026 (OpenAI Help Centre, September 2026). A leftover Sora brief that lived at 720p and wanted length is closer to Grok’s 30s at 4 to 6 credits a second than to Veo’s 8s at 16 to 32 a second. Grok’s menu does not include 4s, 8s, 12s, or 16s. It includes 6, 10, 15, 20, 30. Pick the nearest length. Do not prompt “20s like Sora” into Veo.

Seedance 2.5 is the other long native-audio T2V: 34 credits (was 67), 720p, 4–30 seconds. That is not this comparison. If 24 credits at 720p is enough, Grok is the cheap door. If you already know you wanted Seedance’s family, go there. Do not treat Grok as a 34-credit discount.

Grok Imagine Extend is 13 credits, extend-video, audio false, 2–10 seconds. It is not a way to turn Veo into 30s. It is not a way to add a soundtrack to a mute extend. If you need to grow a Grok take, read the extend card. Do not assume the 30-second T2V menu is an extend slider.

Veo: from 64 credits, 4K, 4 / 6 / 8 seconds

When Veo is the pick:

4K is the master, or you will crop a 4K plate and still want the detail.
The shot is four, six, or eight seconds and you would still send it to Veo if Grok were free.
Dialogue and physics are the job, not duration.
You are willing to cut a longer script into eight-second scenes rather than stretch one take.

When Veo is the wrong pick:

The brief is “one file, 20 seconds, no cut.” Veo will return 8. Writing “continue for thirty seconds” into the prompt still returns four, six, or eight.
You are look-devving a sentence you will throw away. Sixty-four credits is a poor sketchbook. Sketch on 24. Promote the keeper.
You needed a still as frame one. Veo 3.1 T2V will invent the plate. VEO First Last Frame is 64 credits, I2V, 4K, 4 / 6 / 8s, two stills. VEO 3.1 Reference to Video is 128 credits, 8s, up to three stills. Those are other doors. This page is T2V versus T2V.

Google made the visible watermark optional on Gemini app and Flow on 14 August 2026 (Fastlane, 24 August 2026). SynthID and C2PA stay embedded. Optional chrome is not 4K. It is not 30 seconds.

Flow Plus/Pro/Ultra is a launcher. A studio is a roster you can leave. When the next shot is a 30-second hold, you change the row, not the vendor relationship with Google.

Length versus resolution and dialogue

The honest splits:

Binding constraint is duration past 8s. Grok. 10 / 15 / 20 / 30 at 4 to 6 credits a second, 720p.
Binding constraint is 4K. Veo. 4 / 6 / 8 at 16 credits a second silent, 32 with sound.
Binding constraint is a four-second clip. Veo. Grok’s floor is 6s.
Binding constraint is an eight-second clip at 720p with a cheap sticker. Neither is perfect. Grok has no 8s; 6s or 10s. Veo has 8s at 128 credits silent (256 with sound) and 4K you may not use. If 10s at 4 to 6 credits a second is tolerable, Grok. If the eight is sacred and 4K is free extra, Veo.
Binding constraint is 1080p pack identity. Neither T2V row. I2V at 1080p.

Do not upscale a 720p Grok take and call it Veo. Do not cut a Veo eight into a 30-second timeline with freeze frames and call it Grok. The fail-over is a different row, not an edit trick.

Article 50 applies to photoreal people in either take. Label at first exposure for EU audiences since 2 August 2026. YouTube’s 27 May 2026 labelling pass (YouTube Blog) overlays photoreal or meaningfully AI-altered Shorts. Labels alone do not impact recommendations or monetization. They also do not add seconds to Veo.

Captions after the take. A 30-second Grok line that autoplays mute is a caption job. The AI caption generator is the burn-in. Captions for silent autoplay if the feed will never unmute. Veo’s eight-second line needs the same burn-in. Native audio is not a subtitle file.

Production order

Write the length on the brief before the model name. If the number is 15, 20, or 30, start on Grok. If the number is 4, 6, or 8 and the plate is 4K dialogue, start on Veo.

If you do not know the length, you do not have a brief. “Make it cinematic” is not a duration. Default Grok at 6s for 24 credits is a cheap question. Default Veo at 8s for 128, or 256 with sound, is an expensive question.

If Grok’s 720p keeper is good enough for the feed, ship it. “We have a 64-credit 4K row” is not a reason to rerun. If the keeper is the motion but the insertion order is 4K, you still cannot promote Grok to Veo by pressing upscale on this card. Rerun the sentence on Veo only if it will fit in eight seconds. If it will not fit, split the script.

The AI video generator is the launcher. One bill. Two rows. Honesty about the slate is the whole decision.

Twenty-four credits is the length door. Sixty-four is the 4K dialogue door. Pick the constraint that will actually kill the take.

Run the length you can actually ship in Versely.

FAQ

Can Veo 3.1 make a 30-second clip?

No. 4, 6, or 8 seconds. For a 30-second native-audio T2V take at 720p, use Grok Imagine Video; the card starts at 24 credits and a full job runs to 168.

Can Grok Imagine Video output 4K?

The published card is 720p. Veo 3.1 is the 4K T2V row, from 64 credits.

Is 24 credits always the correct first try?

When you are drafting a sentence and 720p is tolerable, yes. When you already know the shot is an eight-second 4K Veo plate, skip the 720p detour.