Most video models on 8frame start from a text prompt alone: Veo 3.1 (Lite, Fast and Standard), Wan 3.0, Wan 3.0 Prime and Wan 2.5, Seedance 2.5 and 2.0, Kling 2.6 Pro, Gemini Omni Flash, Hailuo H3, Happy Horse, Grok Imagine, Pruna p-video and Creatify Boreal. A smaller group will not run without a start image: Kling v3 Standard and Pro, Kling O3, Runway Gen-4 Turbo and Wan 2.2. The one that catches people is Kling. The Kling tool opens on Kling v3 Standard, which needs an image, so a text-only Kling prompt means switching the node to Kling 2.6 Pro: 95 credits for 5 seconds with sound, 47 silent. Everything below was read from 8frame's own model code on 2026-10-01.
TL;DR
- Text only works on: Veo 3.1 (all three tiers), Wan 3.0, 3.0 Prime and 2.5, Seedance 2.5 and 2.0, Kling 2.6 Pro, 2.5 Turbo Pro and 2.1 Master, Gemini Omni Flash, Hailuo H3 and H3 Max Turbo, Happy Horse, Grok Imagine, every Pruna p-video variant, Creatify Boreal
- Needs a start image: Kling v3 Standard (the Kling tool's default), Kling v3 Pro, Kling O3, Kling O1 Image, Kling 2.1 Pro, Runway Gen-4 Turbo, Wan 2.2, Hailuo H3 Max Camera Controls
- Kling from text: switch to Kling 2.6 Pro, 95 credits for 5 seconds with sound, 47 silent
- Veo from an image is always 8 seconds. The 4 and 6 second options exist only from text
- Cheapest text-only clip: Pruna p-video Draft, 4 credits for 5 seconds at 720p. Cheapest that always has sound: Creatify Boreal, 7
- A start image costs 6 to 11 credits: Seedream 5 Lite 6, Nano Banana 2 at 1K 11
Every video model: what it needs to start
"Needs" is what the model refuses to run without. "Optional" is what the 8frame node lets you add. Prices are credits for a 5-second clip at the resolutions shown; a credit is about $0.01 at pack rate.
| Model on 8frame | Text only? | Needs | Optional inputs | 5 s price |
|---|---|---|---|---|
| Veo 3.1 Lite | Yes | nothing | start image (fixes the clip at 8 s), end frame with a start image | no 5 s option: 27 for 4 s, 54 for 8 s with sound (17 / 33 silent) |
| Veo 3.1 Fast (the Veo tool's default) | Yes | nothing | start image (8 s), end frame with a start image | no 5 s option: 84 for 4 s, 168 for 8 s with sound (56 / 112 silent) |
| Veo 3.1 | Yes | nothing | start image (8 s), end frame with a start image, or 1 to 3 reference images on the multi-reference option | no 5 s option: 224 for 4 s, 448 for 8 s with sound (112 / 224 silent) |
| Kling v3 Standard (the Kling tool's default) | No | start image | sound on or off | 86 with sound, 57 silent |
| Kling v3 Pro | No | start image | sound on or off | 114 with sound, 76 silent |
| Kling O3 | No | start image | end frame, sound (off by default), multi-shot prompts | 57 silent, 76 with sound |
| Kling 2.6 Pro | Yes | nothing | start image, sound on or off | 95 with sound, 47 silent |
| Kling 2.5 Turbo Pro | Yes | nothing | start image, end frame with a start image | 48 |
| Kling 2.1 Master | Yes | nothing | start image | 189 |
| Kling 2.1 Pro | No | start image | end frame | 67 |
| Kling O1 Image | No | start image | end frame | 73 |
| Kling O1 Reference | No | at least one reference image or element (up to 7) | 73 | |
| Wan 3.0 (the Wan tool's default) | Yes | nothing | start image | 24 / 48 / 95 (480p / 720p / 1080p) |
| Wan 3.0 Prime | Yes | nothing | start image, or a reference video | 46 / 95 / 189 (480p / 720p / 1080p) |
| Wan 2.5 | Yes | nothing | start image | 102 (1080p) |
| Wan 2.2 a14b | No | start image | 50 (720p) | |
| Seedance 2.5 (the Seedance tool's default) | Yes | nothing | start image and end frame, or up to 9 reference images; reference video; reference audio | 70 / 157 (480p / 720p); 291 / 654 with a reference video |
| Seedance 2.0 | Yes | nothing | same as 2.5 | 48 / 108 (480p / 720p); 60 / 132 with a reference video |
| Happy Horse | Yes | nothing | up to 9 images, or a video to edit | 95 / 189 (720p / 1080p) |
| Gemini Omni Flash | Yes | nothing | start image, or 1 to 7 reference images (not both), or a video to edit | 85, sound always on |
| Hailuo H3 | Yes | nothing | start image | 88 (2K) |
| Hailuo H3 Max Turbo | Yes | nothing | start image | 17 / 27 / 54 (480p / 768p / 1080p) |
| Hailuo H3 Max Camera Controls | No | start image | 34 / 54 / 108 (480p / 768p / 1080p) | |
| Grok Imagine | Yes | nothing | start image; a separate reference option takes 1 to 10 images | 34 / 48 (480p / 720p), sound always on |
| Grok Imagine 1.5 (older) | Yes | nothing | start image | 54 / 95 (480p / 720p) |
| Runway Gen-4 Turbo | No | start image | a prompt (optional here) | 40 |
| Pruna p-video | Yes | nothing | start image, an audio track (the track sets the length) | 14 / 27 (720p / 1080p) |
| Pruna p-video Draft | Yes | nothing | start image, an audio track | 4 / 7 (720p / 1080p) |
| Pruna p-video 2 Pro | Yes | nothing | start image | Speed 14 / 24, Quality 27 / 51 (480p / 768p) |
| Creatify Boreal | Yes | nothing | reference image, driving audio | 7 / 21 / 81 (720p / 1080p / 2K), sound always on |
Watch three inputs that change the bill. A start image on Veo fixes the clip at 8 seconds, so the 4 and 6 second prices are text-only. A reference video moves Seedance to its video-input rate: 654 credits for 5 seconds at 720p on 2.5, 132 on 2.0. An audio track on Pruna p-video makes the clip as long as the track. On every other model in the table, a start image leaves the per-second price where it was.
Pruna and Creatify Boreal are new to 8frame, so we quote their prices, not their quality.
The Kling tool opens on a model that needs an image
A new Kling node starts on Kling v3 Standard, 5 seconds, sound on: 86 credits. On 8frame, Kling v3 Standard and Pro run only through Kling's image-to-video endpoint, so a prompt with nothing connected does not run; the node asks for a start image first. Kling O3 works the same way. There are three routes from there.
- Switch the node to Kling 2.6 Pro. It runs from text: 95 credits for 5 seconds with sound, 47 silent; 189 and 95 for 10 seconds. Its shapes from text are in AI video aspect ratios by model.
- Make a start frame and keep Kling v3 Standard. A Seedream 5 Lite frame (6) plus a v3 Standard clip with sound (86) is 92 credits; silent, 63. With an image attached, Kling v3 takes the image's shape, so make the frame vertical if you need a vertical clip. Upload limits are in Kling image-to-video upload requirements.
- Use an older Kling from text. Kling 2.5 Turbo Pro is 48 for 5 seconds and Kling 2.1 Master 189. Neither has a sound setting.
The same split applies through the MCP connector: its Kling tool runs Kling v3 Standard and requires an image URL. More on prompting the newer model in how to use Kling 3.
Getting a start image cheaply
When the model you want needs an image, the image is the cheap part.
- Seedream 5 Lite: 6 credits per image. The Seedream node opens on 5 Pro (7 at 1K), so switch it to 5 Lite.
- Nano Banana 2: 11 credits at 1K, the Nano Banana node's default. 17 at 2K.
Connect the image node to the video node and the image becomes its start frame. What a few combinations cost:
| Route | Credits |
|---|---|
| 1 Seedream 5 Lite frame + Kling v3 Standard 5 s with sound | 6 + 86 = 92 |
| 4 Seedream 5 Lite options + Kling v3 Standard 5 s with sound on the one you keep | 24 + 86 = 110 |
| 1 Nano Banana 2 frame (1K) + Kling v3 Standard 5 s with sound | 11 + 86 = 97 |
| 1 Seedream 5 Lite frame + Runway Gen-4 Turbo 5 s | 6 + 40 = 46 |
| 1 Seedream 5 Lite frame + Wan 2.2 5 s | 6 + 50 = 56 |
Choosing the frame first settles the composition at image prices, before you pay for motion. Generate the frame in the final shape: Kling v3, Wan, Hailuo H3 and Grok Imagine follow the start image's shape, and Veo 3.1 takes only 16:9 and 9:16. Which image model to use for the frame is compared in Nano Banana 2 vs Seedream 5, and the full chain is in the image-to-video workflow.
Other inputs, in one place
- End frames: the table shows which models take one. Rules and transition prices are in AI video from a first and last frame.
- Reference images: Seedance (up to 9, instead of a start frame), Gemini Omni Flash (1 to 7, instead of a start image), Happy Horse (up to 9), Veo 3.1 multi-reference (1 to 3), the Grok Imagine reference option (1 to 10) and Kling O1 Reference (up to 7). Limits across all models: reference image limits by model.
- Reference video: Seedance 2.5 and 2.0 (at the video-input rate) and Wan 3.0 Prime (same price as from text).
- Audio in: Seedance takes a reference audio track, Pruna p-video and Draft take a track that sets the clip length, and Creatify Boreal takes driving audio. Which models put sound in the output is a separate question: AI video models with audio.
Modes that start from a video, a portrait or a sound
These are not text-to-video at all, but they sit in the same tools:
- Edit or extend a clip: Grok Imagine extend (54 to add 5 seconds to a 5-second clip) and edit (41 at 480p, 54 at 720p for a 5-second clip), Gemini Omni Flash video edit (85 for a 5-second clip, inputs up to 10 seconds), Happy Horse video edit (189 at 720p, 378 at 1080p for 5 seconds), Wan 2.7 VideoEdit (68 per 5 seconds of output), Wan 2.2 video-to-video (50), Kling O1 Video (109).
- Motion transfer: Kling 2.6 Pro Motion Control needs a character image and a motion video; 95 credits for a 5-second motion video.
- Talking portraits: Hailuo H3 Max Lip Sync, OmniHuman 1.5 and Pruna Avatar need a portrait and an audio track, and Runway Act Two needs a character and a performance video. Prices: AI lip sync price by model.
FAQ
Can Kling v3 make a video from text only? Not on 8frame. Kling v3 Standard and Pro, and Kling O3, need a start image. For a text-only Kling clip, switch the node to Kling 2.6 Pro: 95 credits for 5 seconds with sound, 47 silent.
Why won't my Kling node run without an image? The Kling tool opens on Kling v3 Standard, which is image-to-video only here. Connect an image, or change the model to Kling 2.6 Pro, 2.5 Turbo Pro or 2.1 Master.
Does Wan 2.2 do text-to-video? Not on 8frame. Wan 2.2 needs a start image, or a video in its video-to-video option. Wan 3.0 (24, 48 or 95 for 5 seconds by resolution) and Wan 2.5 (102) run from text.
What is the cheapest AI video from text only on 8frame? Pruna p-video Draft, 4 credits for 5 seconds at 720p. Creatify Boreal is 7 and always generates sound.
Can I make a 4-second Veo clip from an image? No. With a start image the Veo node is fixed at 8 seconds, so the cheapest Veo clip from an image is Veo 3.1 Lite at 54 with sound or 33 silent.
Is image-to-video more expensive than text-to-video? On models that do both, the per-second price is the same with or without a start image; Veo is the exception in practice, because an image fixes the length at 8 seconds. Image-to-video adds the frame itself: 6 credits on Seedream 5 Lite, 11 on Nano Banana 2 at 1K.
Does this work the same through Claude or Cursor? Through the MCP connector, the Kling tool (v3 Standard) and the Wan tool require an image URL; the Veo, Seedance, Gemini, Hailuo, Grok, Pruna and Boreal tools take a prompt alone. Setup is in generate AI video from Claude.
Sources
- Which inputs each model requires and accepts: 8frame per-model generation code (the strategy files for Kling, Veo 3.1, Wan, Seedance, Gemini Omni Flash, Hailuo H3, Happy Horse, Grok Imagine, Pruna Video, Creatify Boreal and Runway Gen-4 Turbo), read 2026-10-01
- Which inputs each canvas node shows, default models, and the Veo 8-second lock with an image: 8frame tool configuration and model compatibility map, read 2026-10-01
- Prices: 8frame tool registry credit functions, read 2026-10-01
- MCP connector inputs: 8frame MCP tool definitions, read 2026-10-01
Try it without signing up: the 8frame canvas opens on a real board. The canvas is free and unlimited; generation is paid from $19/month.