Preview the first frame before an avatar render: Sume vs credits
Synthesia and HeyGen sell avatar video from monthly credit pools. Sume previews the first frame of an avatar scene before you commit to the full render.
If you want to approve how an avatar scene looks before paying for the whole video, Sume has a separate preview stage: POST /v1/avatar-video-previews creates first-frame stills only, and you call generate-video on the one you like. The Synthesia and HeyGen pricing pages I read describe credit allowances, not a preview-first workflow, so I make no claim about whether either offers one.
This matters most for ads, where a wrong product placement or composition costs a full render to find out.
| Item | Synthesia | HeyGen | Sume |
|---|---|---|---|
| Allowance | Starter 1,250 credits ($29); Pro 6,000 ($89) | Creator 600 credits ($29); Pro 1,000 ($49) | USD balance |
| Longest video | Not stated on the page I read | 30 minutes (Creator, Pro); 60 (Business) | 4 to 60 seconds estimated |
| Resolution | Full HD | 1080p; Pro and up to 4K | 720p |
| Preview before full render | Not stated | Not stated | Yes: first-frame stills |
How the Sume preview works
A preview takes the same body as an avatar video: a script or video_inputs, not both, plus optional product_image, scene, quality, aspect_ratio, title and captions. The default quality is plus. The response gives job URLs and an avatar_video_preview_id, and the resource later holds preview_image_url and one scene_previews entry per input scene.
You can regenerate the stills, then call generate-video. Caption intent can be stored at create time; Sume never burns captions into preview stills and applies them only when the final video generates.
- Multi-scene plans get one still per scene.
- Poll the job like any other generation, or use a webhook.
- Prefer resource_status and job_status over the legacy status field.
Where this saves money
The preview exists so you do not pay for a failed composition. Compare that with a credit pool: when a clip is wrong on a plan, you spend credits to find out, and you can run out before month end. On a balance, the same mistake is a charge you can see in the usage ledger.
The catch is that a preview is its own job. Use it when the first frame is where the risk is, such as a product shot or a scene with a specific background. For a plain talking head on a studio backdrop, skip it and render directly.
Verdict
Plan-based tools are simpler to budget if you always generate several variations a month. A preview stage is better when each video is a separate decision. If you already have a plan with credits in hand, keep using it for that; if you submit avatar jobs from code, add the preview call to your pipeline.
If you want a rule of thumb, preview when the scene is new and skip it when it is a repeat. A first run of a product ad with a specific setting deserves a preview, and so does any multi-scene plan where each scene has a different backdrop. A weekly update with the same avatar, scene and framing does not need one. Keep the preview id with the project so you can tell later which still was approved.
Sources
Related posts
More in Comparisons
- Qwen Image Max or Ideogram 4.5: 100 images, price and limits
100 images cost $9.375 on Qwen Image Max and $3.75, $7.50 or $27.50 on Ideogram 4.5 by quality. Ratios, references and output formats side by side on Sume.
- Recraft V4 or ChatGPT Image 2.5 for marketing graphics on Sume
Recraft V4 bills $0.05, text-only, webp only. ChatGPT Image 2.5 bills by tokens, makes transparent PNGs and takes a mask. Which row for which job.
- Regenerate or recast a 15-second ad? $3.00 vs $8.44 at 1080p
New presenter in a 15-second ad: regenerate with H3 Max (from $1.50 at 768p) or swap with H3 Max Recast ($5.63 at 768p). What each keeps and what it costs.
- Rev AI forced alignment $0.003/min vs Sume caption script_text
Rev AI lists forced alignment at $0.003 a minute. Sume aligns your script to speech inside a $0.20 caption job. Which to use for subtitles from a script.
Written by Sume